TRACKING CONTROL OF NONLINEAR MECHANICAL SYSTEMS

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Universiteit Twente, op gezag van de rector magnificus, prof.dr. F.A. van Vught, volgens besluit van het College voor Promoties in het openbaar te verdedigen op vrijdag 14 april 2000 te 15.00 uur.

door Adriaan Arie Johannes Lefeber geboren op 7 december 1972 te Beverwijk

Dit proefschrift is goedgekeurd door de promotoren prof. dr. ir. H. Kwakernaak, prof. dr. H. Nijmeijer.

Ik droomde eens en zie ik liep aan ’t strand bij lage tij. Ik was daar niet alleen, want ook de Heer liep aan mijn zij. We liepen saam het leven door en lieten in het zand, een spoor van stappen, twee aan twee; de Heer liep aan mijn hand. Ik stopte en keek achter mij en zag mijn levensloop, in tijden van geluk en vreugd van diepe smart en hoop. Maar als ik goed het spoor bekeek, zag ik langs heel de baan, daar waar het juist het moeilijkst was, maar een paar stappen staan ´´ Ik zei toen: “Heer, waarom dan toch?” Juist toen ’k U nodig had, juist toen ik zelf geen uitkomst zag op ’t zwaarste deel van ’t pad De Heer keek toen vol liefd’ mij aan en antwoordd’ op mijn vragen: “Mijn lieve kind, toen ’t moeilijk was, toen heb Ik jou gedragen ”

opg. Box 513 5600 MB Eindhoven The Netherlands CIP-DATA KONINKLIJKE BIBLIOTHEEK. 165 p. 2000. Enschede.O. 24.: ill. of Mechanical Engineering Eindhoven University of Technology P.: p. — vi. — Proefschrift Universiteit Twente.. ISBN 90-365-1426-6 Ook verschenen als Online Resource .153–159 — Met samenvatting in het Nederlands.Erjen Lefeber Systems Engineering Dept. DEN HAAG Tracking control of nonlinear mechanical systems / Adriaan Arie Johannes Lefeber.5 cm Auteursnaam op omslag: Erjen Lefeber. — Lit. — Enschede : Universiteit Twente.

Henk Nijmeijer. I have benefited from our (lively) discussions and his willingness to listen to me when I tried to settle my thoughts. colleagues. Arjan van der Schaft of Twente. I am grateful for the opportunity of working with him. I would like to thank Guido Blankenstein for being more than just a colleague. Koos Rooda of Eindhoven. whereas it is one of the last pages I wrote. as a period of my life is about to finish. Kristin. I would like to express my gratitude to Anders Robertsson. by writing down this question I also found an answer to it. Furthermore. I can not avoid looking back at it. and Prof. Huibert Kwakernaak. Janusz ı Jakubiak and Rafael Kelly. I could elaborate more on these paradoxes. Romeo Ortega. Paradoxically. First of all I would like to thank my supervisor and promotor Henk Nijmeijer. and for making me put things in the right perspective. Guido. It is funny to realize that this probably is the first page that people read (except from the title. and Prof. this preface is full of paradoxes. but especially for all the playing on words in English during our conversations. and for Dutch readers also the poem. Claude Samson of Sophia Antipolis (France). While typing this sentence. They all are people who. I am really looking forward to our collaboration in Eindhoven in the near future. as they all contributed to this thesis. These include Henk. I am grateful to the members of my promotion committee for thorough reading my manuscript: Prof. not only for our working together and his hospitality during my visit to Lund. Prof. Prof. Antonio Lor´a. I would also like to thank Prof. amongst others. Guy Campion of v . my thoughts went out to several people. Ben Jonker. Zhong-Ping Jiang. e-mails and phone calls. I presume). Anders. Prof. In addition I am indebted to Kristin Pettersen. If one comes to think about it. for the possibilities of visiting several colleagues. Prof. I enjoyed the moments we spent together outside of office and his special sense of humor. During the last five years we have come to know each other quite well. To˜ o & Lena.Preface How to start a preface to a thesis that should give an overview of what I have been doing for almost four years? That is not an easy question to answer. Arun Bagchi. Elena Panteley. my committee. deserve credit for their contributions (in one way or the other) to this thesis. but that would make this first page so philosophical and paradoxical that even this first page will not be read At a time like this. parn ents and not least of all Wieke (thinking about periods of my life ). and looking forward too.

and especially those of the Systems. I would finally like to thank all my friends. Kristin Pettersen. They all contributed to this thesis with their valuable comments. A special word of thanks goes to my fianc´ e Wieke Fikse for all e her support and help. and Phil Chimento who sacrificed themselves for going through (parts of) a draft version of this thesis. A special word of thanks goes to Henk Ernst Blok. and all the people whose name I did not mention explicitly. my beloved family. Special thanks also go to the Systems Engineering group in Eindhoven (my current colleagues) for giving me the opportunity to finish my thesis quietly and showing me a glimpse of the challenges I will face in the near future. I would like to take the opportunity to thank all my (former) colleagues at the Faculty of Mathematical Sciences in Twente. Signals and Control group for having provided me with such a creative and friendly atmosphere to work in. Erjen Lefeber .vi Preface Louvain-la-Neuve (Belgium) for his feedback during the meetings of my advisory committee. Paul Huijnen.

. . . . . . .2 27 29 Cascaded design . . 37 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 State-feedback . . . . . . . . . . . . . . . . . . . . . . 13 Lyapunov stability . . . . . . . . . . . . . . . .4 2. 25 I A cascaded approach to tracking 3 Introduction to Part I 3. . . . . . . . . . . . . . . . 37 A cascaded design . . . . . . . . . . . .2 Previous results . Non-holonomic systems . . . . . . . . . . . . . . . . . . . . . . . . .2 Introduction . . . . . . . . . . . . . . . . . . . . . 15 Linear time-varying systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Cascaded systems . . . . . . Outline of the thesis . . . .2. .2 2. . . .3 2. . . . . . . . . . . .1 4. . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v 5 6 8 9 13 2 Preliminaries 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Backstepping . . . . . . . . . . . 39 1 . . . . . . .1 2. . . . . 31 35 4 Tracking of a mobile robot 4.5 Mathematical preliminaries . . . . . . . . . . . 29 An introductory example: tracking of a rotating rigid body . . . . . . . .1 1. . . .3 Formulation of the tracking control problem . . . . . . . . . . . . . .1 4. . . . . . . . . . . .Contents Preface 1 Introduction 1. . . . . . . . . . .1 3.2 1. .

. . . . . 45 Saturated control . 75 State-feedback: a cascaded approach . . . . . . . . . . . . . . . . 55 57 5 Tracking of non-holonomic systems in chained form 5. . . . . . . .4. . . .2 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 4. . .4 Unmeasured Ü . . . . . . . . . . . . . . . . . . . . . .3. 57 The search for a cascaded structure . . . . . . . . . . . . . . . . 92 Discussion . . . . . . . . . . . . . . . . . . . . . . 70 73 6 Tracking control of an under-actuated ship 6. . . .4 89 91 Paper I: Global asymptotic stability of robot manipulators with linear PID and PI¾ D control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Contents Dynamic output-feedback . . . 88 II Robot manipulators and adaptive control 7 Introduction to Part II 7. . . . . . . . . . . . 73 The tracking error dynamics . . . .1 6. . . . . . . . . . . . . . . . . . . . . . . . .5 4. . . . . . . . . . . . . . . .4 5. 65 Simulations . 59 State-feedback . . . . . . . . . .6 Simulations . . . . . 46 4. . . . . . . . . . . 93 . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . .7 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 A cascaded design . . . . . . .1 7. . . . . . . . . . . . . . .2 6. . . . . . .3 5. . . . . . . . . .1 5. . . . . . . . . . . . . . . . 79 Concluding remarks . . . . .6 5. . . . . . . .2 4. . . .1 4. 42 Unmeasured Ý . . . . . . . . . . . . 92 Paper III: Adaptive tracking control of non-holonomic systems: an example . . . . . . . . . . . . . . . . . .3 7. . . . . . . . . . . . . . 76 Experimental results . . . . . . . . . . . . . . . 66 Concluding remarks .1 4. 62 Saturated control . . . . . . . . . . . . . . .4 6. . . . . . . . .5 5. . . . . .2 A Lyapunov design . . . . . . . .3 6. . . 42 4. . . . . . . . . . . . . . . . . .2 5. . . . . . . . 60 Dynamic output-feedback . .2 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Paper II: Adaptive and filtered visual servoing of planar robots . . 49 Concluding remarks . . . . . . . . . . . . .5 Introduction . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .3 Main results . . . . . . . . .1 8. . . . . . . .1 8. . . . . . . . . . . . . . . . . . . . .2. . . . . .3 9. .3. . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . . .2. . . . . . . . . . . . . . . 121 9.2 First case: measurable velocities . 123 Filtered visual servoing . . . . . . . . . . . . . . . . . . . 100 Second case: unmeasurable velocities . . . .2. . . . . 121 Output equation . . . . . . . . . .2 Preliminary results .4 Discussion . . . . 116 9 Adaptive and filtered visual servoing of planar robots 9. . . . . . . . . . . . . .6 A class of stable visual servo controllers . . . . . . . . . . . . . . . . . . . . . . . . 125 9. .1 9. . . . . . . . . . . . . . . . . . . . . . . . .6 Simulation results .1 9. . . . . . . . .1 8. 100 8. . .2. . . . . . . . . . 124 Acknowledgements . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 119 Introduction . . . . . . . . . 103 8. . . . . . . . . . . . . . . . . . . . .Contents 8 GAS of robot manipulators with linear PID and PI¾ D control 8. . . . . . 105 Second case: unmeasurable velocities . . . . . . . . . . . . . . .3. . . . preliminaries and notation . . . . . . . . . . . . . . . . . . . . . . . . . .2 The normalized PID of Kelly . . . . . . . . . . . . 96 Model and problem formulation . . . . . . . . . . . . . . 122 Adaptive visual servoing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 8. . . 122 9. . . 112 Concluding remarks . . . . 124 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . .5 9.2. 108 8. . . . . . . . .2 First case: measurable velocities . . . . . . . . . . . . . . . . 105 8. . . . . . . . . .1 8. . . . . . . . . . . . . . . . . . . . . . .1. . . . . .5 8. . 110 The saturated PID controller of Arimoto . . . . 109 8. . . . . . . . . .2 Literature review . . . .1 3 95 Introduction . . .4 Robot dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4. . .3 9. . . . . . . . . . .7 References . . . . . . . 121 Notation . . . . . . . . . . . . . . . . . . . . . . . . 121 Problem formulation . . 125 . . . .4 9. . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . 120 Problem formulation. . . . . 96 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 References . . . .2 9. .

. . .7 . . . . . 143 A. . . . . . . .4 A tracking controller . . . . . . 130 10. . . . . . . . . . . . . . . . . . . . . . . . 128 10. . . . . . . . . . . . . . . . . . . 133 References . . . . . . . . . . . . . 133 11 Conclusions 135 11. . . . . . . . . . 128 10. . . . . . . . . . . . . . . . . . 146 B Backstepping controller for an under-actuated ship Bibliography Summary Samenvatting Curriculum Vitae 149 153 161 163 165 .1 Tracking control problem . . . . . . . . . .5 An adaptive controller . . . . . . . . .2 Problem formulation . . . .1 Discussion . . . . . . .2 Adaptive tracking control problem . . . . . . . . . . . . . . . . . . . .2 Further research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . 135 11.2 Proof of Theorem 2. .4 10 Adaptive tracking control of non-holonomic systems: an example Contents 127 10. . . .1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Preliminaries .3. . . . . . . . . . . . . . . . . . . . . . 137 A Proofs 141 A.1 Proof of Theorem 2. . . . . . . . . . . . . . . . . . 130 10. . . . . 128 10. 131 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . .1 Introduction . .3. . . . . . . . . . . . . . . . . 145 A. .3 Proof of Proposition 6. . . . . . . 129 10. . . .6 Concluding remarks . . . . .

like a car. is presented. control systems are inevitable. Besides the growing usage of control systems. automobiles. the cascaded approach. This thesis is also concerned with the control of special classes of nonlinear systems. systematic approaches and mathematical tools are only available for certain classes of nonlinear systems. Nonlinear control is an important area in control. and more complex control systems. ships).g. robots. spacecrafts. as virtually all physical systems are nonlinear in nature. resulting in more. It turns out that this may simplify the controller design. as part of the nonlinear dynamics can be forgotten. As a result nonlinear control theory has to be used for these systems. in industry (e.. In case a system does not deviate too much from the nominal set of operating conditions. CD-player). In nonlinear control theory a large variety of approaches and mathematical tools for analysis exists. and in numerous other places and situations.Chapter 1 Introduction Nowadays. They appear almost everywhere: in our homes (in e. often linear models can be used for describing the system and designing controllers. in telecommunications. in several types of vehicles (in e. in biomedical engineering. whose linearization around any equilibrium point is uncontrollable. In the first part a new design approach.. video. radio.. In mathematical control theory the basic principles underlying the analysis and design of control systems are studied. airplanes. The cascaded design approach aims at arriving at a specific structure for the closed-loop system. Nonlinear controllers are capable of handling the nonlinearities directly in large operating ranges. As a result. However. The main advantages of this new approach are that the expressions for the resulting control laws are not complex and that transforming the system is not necessary: all analysis can be done in the original co-ordinates. a linear(ized) model usually is inadequate or inaccurate. The main reason for this variety is that no tool or methodology is universally applicable in nonlinear systems analysis.g. process control). television. The tracking prob5 .g. And even when the operation range is small enough. Then nonlinear control comes into play. In order to be able to design more complex control systems for a larger variety of systems. The thesis consists of two parts. as controllable systems exist. when the required operation range is large. a good understanding of control systems is crucial. linearization does not always work. the requirements for a control system increase considerably.

a ship. Thirdly the tracking control problem for nonlinear systems is considered in the presence of unknown parameters.1a) (1. This conflict is illustrated by means of an example. experiments on a scale model of an offshore supply vessel have been performed. focal length and center offset). a feasible reference state trajectory ÜÖ Ø is assumed to be given for the system to track. viewed on top with a camera. Secondly the visual servoing problem for a rigid robot manipulator is considered. The problem is solved by using an adaptive controller. i. Ù ¾ ÊÑ denotes the input by means of which the system can be controlled. this thesis is not concerned with it and a reference state ÜÖ Ø as well as a reference input ÙÖ Ø which satisfy (1. Feasible means that once being on the reference trajectory it is possible to stay on that trajectory. This means that also a reference input ÙÖ Ø is assumed to exist. The goal is to regulate the tip of the robot manipulator to a specified point at this screen using only position measurements. a robot manipulator is considered operating in the plane. which is such that ´Ø Ü Ùµ ´Ø Ü Ùµ (1.2) The problem of generating such a feasible reference trajectory for a system is a challenging problem. In case of the under-actuated ship. Extra difficulties are that both the camera position and orientation are assumed to be unknown. ´µ . like a mobile robot. The applicability of the method is illustrated by means of simulations.6 Chapter 1. Furthermore. but also solved. In the second part of this thesis three specific problems are considered. 1.1 Formulation of the tracking control problem In this thesis the tracking control problem for nonlinear systems is considered.e. That is. a rigid spacecraft. First. the regulation problem for a rigid robot manipulator under a constant disturbance is considered.2) are assumed to be given. or a rigid robot manipulator. Although motion planning (including obstacle avoidance) is an interesting problem. as not knowing certain parameters and specifying a reference trajectory are in conflict with each other. known as the motion planning problem. for which an adaptive tracking control problem is not only formulated. It turns out that finding a suitable problem formulation is a problem in itself. as well as certain intrinsic camera parameters (like scale factors. An accurate mathematical model is assumed to be given for the system under consideration.1b) ´µ ´µ ÜÖ ´Ø ÜÖ ÙÖ µ ´µ (1. An image of the robot manipulator is displayed at a screen. It is shown that the common practice of using a PID-controller is guaranteed to work globally in case the integral action is turned on only after some time. and Ý ¾ Ê denotes the output of the system which represents the measurements. which is of the form: Ü Ý Here Ü ¾ ÊÒ denotes the state of the system. Introduction lem is first studied for mobile robots. the adaptive tracking control problem is considered.. a car towing multiple trailers. then for the class of so called chained-form systems (including cars towing multiple trailers and a rigid spacecraft with two torque actuators) and finally for an under-actuated ship.

Consider the system (1. it might happen that the mobile robot or ship turns around and follows the reference trajectory backwards. 1. also the constraint ٴص Ùmax Ø ¼ (1. finding a control law for the input Ù which is such that as Ø tends to infinity Ü Ø converges to ÜÖ Ø .3) Ø Ð Ñ Ü´Øµ   ÜÖ ´Øµ ½ ´ µ Ù´Ø ÜÖ ÙÖ Ý Þ µ ´Ø ÜÖ ÙÖ Ý Þµ ¼ Problem 1. more is desired.3) such that for the resulting closed-loop system (1. When the only focus is on controlling the position.4) Ø Ð Ñ Ü´Øµ   ÜÖ ´Øµ ½ ¼ Remark 1. the following two problems can be distinguished: ´µ ´µ Problem 1.4b) such that for the resulting closed-loop system (1. i. Consider the system (1.e. a trajectory satisfying (1. but it is not all what is really of interest.1. no constraint on the size of the input is given. This is not the problem this thesis deals with. Find an appropriate dynamic control law Ù Þ (1. Assume that a feasible reference trajectory ÜÖ ÙÖ is given (i.1).. Find an appropriate control law ´ µ Ù Ù´Ø ÜÖ ÙÖ Üµ (1. Formulation of the tracking control problem Once a reference state trajectory ÜÖ Ø and a reference input ÙÖ reference output ÝÖ Ø can be defined by means of 7 ´µ ´µ ´Øµ are given. the tracking problem can also (actually: better) be formulated as finding an appropriate control law that renders the zero tracking error equilibrium asymptotically stable.1.e..2)). Therefore. To be more precise. In case of the first problem the entire state can be used for feedback.2)).1). that is.. the measured output typically is the position. Notice that the assumption of tracking a feasible trajectory is the same as requiring the zero tracking error to be an equilibrium. Two major state trajectory tracking problems can be distinguished.2 (Output-feedback state-tracking problem). In general. 1. whereas in practice the input that can be supplied to the system is limited.5) . that is the problem of finding a control law for the input Ù such that as Ø tends to infinity Ý Ø converges to ÝÖ Ø . also the resulting ÝÖ ´Ø ÜÖ ÙÖ µ An often studied problem is the problem of output tracking.3.1 (State-feedback state-tracking problem). Tracking of the position might seem an interesting problem. Assume that a feasible reference trajectory ÜÖ ÙÖ is given (i.e. ´µ ´µ This is one of the reasons for insisting on state-tracking. a trajectory satisfying (1.1.1.1. namely the state-feedback problem as well as the output-feedback problem. In the problem formulations as presented.1. For systems like a mobile robot or a ship.1. whereas for the latter only the output can be used.4a) (1.

where Ùmax is a given constant. i. all systems studied in this thesis have so-called nonholonomic constraints.8) . ½ ¾ turns out to be only a first order model.8 Chapter 1.6) contains a constraint on the velocities: ¼ (1.6) which seems to be a second order model. like the state-tracking control problem. since it can be integrated to obtain ½ ܾ · ½ ܾ ¾ ½ ¾ ¾ Ö constant This teaches us that model (1. ܽ ܽ · ܾ ܾ Notice that model (1. with the additional constraint (1.6) ´Ü½ ܾ µ is the state and Ù is the input. 1.7) This constraint is a so-called holonomic constraint. Introduction has to be met. which results in the additional assumption that Ùmax ×ÙÔ Ø ¼ ÙÖ ´Øµ These are the control problems studied for several types of systems in this thesis. Clearly. consider the simple model ܽ ܾ where Ùܾ  Ùܽ (1. The change of co-ordinates ܾ · ܾ ½ ¾ ÖØÒ Ü½ ܾ Ö Ö´¼µ leads to the “new model” Ù As ܾ ܾ is a conserved quantity. In that case the state-tracking problem under input constraints can be formulated in a similar way.2 Non-holonomic systems Except for the rigid robot manipulator. · Things become different when considering the model ܽ ܾ Ü¿ Ù½ Ù¾ ܽ Ù¾   ܾ Ù½ (1. for obtaining a solvable problem it has to be assumed that the reference satisfies the input constraints. the model (1. What does this mean? To make this more clear..e.5).6) can be reduced.

1.3 Outline of the thesis This thesis consists of two parts. . that asymptotically stabilizes the system (1. Nijmeijer.-P. Third Portuguese Conference on Automatic Control. Lefeber. Jakubiak. Model (1.3. general chained-form systems in Chapter 5 and an under-actuated ship in Chapter 6. Tch´ n. preceded by a chapter with preliminaries.9) can not be integrated.7) the constraint (1. Jiang. The applicability of the method is illustrated by means of simulations. contrary to (1. and H. Lefeber. As it turns out. conditions on the reference input ÙÖ have to be imposed in order to circumvent this problem.9) is called a non-holonomic constraint.7. This chapter is included for making the thesis more or less self-contained. 315– 320.9) However. Portugal. This approach is illustrated by means of several examples: mobile robots in Chapter 4.” in Proceedings of CONTROLO’98. Part I In the first part a cascaded design approach to the tracking problem for nonlinear systems is presented.4 is fundamental for Part I.” 2000. The main contributions of Chapter 2 are Theorem 2. K.8 and Lemma 2. i. Z.9) is inherently part of the dynamics. Australia. “Observer based tracking cono trollers for a mobile car. Outline of the thesis where ܽ ܾ Ü¿ is the state and on the velocities. a control law of the form Ù Ù Ü .9) can not be written as time derivative of some function of the state. Nijmeijer. 1998. pp. ´µ Since in this thesis the tracking problem for non-holonomic systems is dealt with and stabilization is a specific case of tracking. In case of the under-actuated ship also experiments have been performed..8) also contains a constraint ܽ ܾ   ܾ ܽ   Ü¿ ¼ (1. Theorem 1) that are necessary conditions for the existence of a continuous static state-feedback law. Sydney. namely 9 ´ µ ´Ù½ Ù¾µ is the input. Theorem 2. i. This first part is a composition of the papers ¯ ¯ J. the constraint (1.5.8).1..3. E. The constraint (1. in one way or the other this difficulty should be taken into account. which means that the constraint (1.8) fails to meet the conditions of Brockett (1983. Section 2. E. “Stabilization and tracking of a nonholonomic mobile robot with saturating actuators. 1. vol.3. Chapter 2 provides an overview of notions and results that are used throughout the thesis.e. and H.e. As a result the system (1. It turns out that it is really necessary to use three variables for modeling this system. Coimbra.4. Submitted to the 39th Conference on Decision and Control.

France. in preparation. K. Nijmeijer.” 1999a. 2. Lefeber and H. 246 in Lecture Notes in Control and Information Sciences. pp. The Netherlands. 10. no. E. Lefeber. “Adaptive tracking control of nonholonomic systems: an example. 1998. no. also under (partial) input saturation. Lor´a. Lor´a. R. A. Lefeber. “Output feedback tracking of nonholonomic systems in chained form. Nijmeijer. respectively ¯ ¯ ¯ A. United Kingdom: Springer-Verlag. Introduction ¯ ¯ ¯ Z. E.” in Proceedings of the 38th IEEE Conference on Decision and Control. Aeyels. 183–199. E.” 1999. Ortega.and output-feedback control problems are considered. chained-form systems and under-actuated ships. and H. van der Schaft. Lamnabhi-Lagarrigue.” in Proceedings of the Fourth IFAC Symposium on Nonlinear Control Systems Design (NOLCOS’98). Lefeber. and H.-P. pp. and H. Nijmeijer. vol. Robertsson. and H. No transformations are needed. Lefeber. Lefeber. “Linear controllers for tracking chainedform systems. vol. Karlsruhe. 563–568. and H. Nijmeijer. Panteley. R. Enschede. D. Y. 1999a. . ¯ ¯ ¯ and some additional unpublished material.” in Proceedings of the IFAC Workshop on Motion Control. 2094–2099. Grenoble. pp. This part consists of three papers. 2000b. J. 243–264.10 Chapter 1. Nijmeijer. all analysis is done in the original error co-ordinates. “Saturated stabilization and tracking of a nonholonomic mobile robot. and H. E. paper 772. and H. A. E. Germany. The main contribution of this part is the introduction of the cascaded design approach. pp. Nijmeijer. “Linear controllers for exponential tracking of systems in chained form. E. Jiang. E. Part II In the second part solutions to three specific problems are presented. The state. “Exponential tracking control of ı a mobile car using a cascaded approach. Nijmeijer. Kelly. 4. “Global asymptotic stability of robot manipuı lators with linear PID and PI¾ D control. USA. New and simple time-varying state-feedback controllers are presented that achieve global and uniform tracking results for tracking mobile robots. and H. Nijmeijer. and A.” 2000a. Submitted to Stability and Control: Theory and Applications. “Adaptive and filtered visual servoing of planar robots. Arizona. A. Phoenix. London. E. pp. A. 1999. 221–226. 1999b.” International Journal on Robust and Nonlinear Control. Lefeber.” in Proceedings of the 5th European Control Conference. Lefeber. “Tracking control of an under-actuated ship. Robertsson. Nijmeijer. Robertsson. E. Lefeber.. Eds. E. 1998. Submitted to Systems and Control Letters. Pettersen.” in Stability and Stabilization of Nonlinear Systems. F.

1.3. Outline of the thesis The contributions of this part consist of

11

¯ ¯ ¯

Global asymptotic stability of linear PID controllers is shown by delaying the integral action; Classes of controllers are introduced that solve the visual servoing of planar robots under a fixed camera position for both the state- and output-feedback problem. These classes also contain saturated controllers. In case of unknown camera orientation a class of adaptive controllers is presented; Difficulties in formulating the adaptive state-tracking problem for nonlinear systems with unknown parameters are illustrated by means of an example. For this example a suitable problem formulation of the adaptive state-tracking problem is given and a solution is presented.

Chapter 11 contains the conclusions of this thesis and some recommendations for further research. Appendix A contains the proofs of some theorems presented in this thesis. Appendix B contains a backstepping control law for tracking an under-actuated ship. This expression which was too long to be incorporated in the text of Chapter 6.

12

Chapter 1. Introduction

Chapter 2

Preliminaries
In this chapter we recall a few notions and results that we use throughout this thesis. First, we consider some fundamental mathematical definitions. Next the concept of Lyapunov stability and some lemmas useful for showing stability are given. We review some basic notions for linear time-varying systems, introduce some crucial theorems, present results on (timevarying) cascaded systems, and briefly illustrate the method of backstepping.

2.1 Mathematical preliminaries
Definition 2.1.1. A norm Ü of an Ò-dimensional vector Ü valued function with the properties

´Ü½

ÜÒ µÌ

¾ ÊÒ is a real

¯ Ü ¼ for all Ü ¾ ÊÒ , with Ü ¼ if and only if Ü ¼; ¯ Ü·Ý Ü · Ý , for all Ü Ý ¾ ÊÒ ; ¯ «Ü « Ü , for all « ¾ Ê and Ü ¾ ÊÒ .
Some commonly used norms are

Ü Ô ¸ ´ ܽ Ô · ¡ ¡ ¡ · ÜÒ Ô µ Ô
½

½

Ô

½

and

Ü ½ ¸ Ñ ÜÒ Ü ½
Definition 2.1.2. We denote a sphere of radius Ö by

Ö , i.e.,

Ö

¸ Ü ¾ ÊÒ Ü
13

Ö

10. Often uniform continuity of a function lemma.7.1. ÊÒ ÊÑ is said to be continuous at a point Ü if given an Ü Ý Æµ ´Üµ   ´Ýµ ÊÒ ÊÑ ¯ Ü Ý ¾ ÊÒ Ë (2.8. Remark 2. we define the ÄÔ norm Ü´ Ü Ô¸ for Ô ¾ ½ and say that norm is defined as ½ µ ¼ µ Ôd Ô ½ Ü ¾ ÄÔ when Ü Ô exists (i.6.1. A function ¯ a constant Æ exists such that ¼ ¼ Ü ½ exists. Notice that the function Ü Ü is uniformly continuous not uniformly continuous on Ê.e.1.14 Definition 2. Consider a differentiable function such that Ê.. Definition 2. the function Ü on any compact set Ë Ê.1. A function it is continuous on Ë .5. For functions of time Ü Chapter 2.3. We denote the class of Ò times continuously differentiable functions by Ò . except for a finite number of points. Ü is continuous on Ê. Ü¾Ê ¬ ¬d ¬ ¬ dÜ ´Üµ ¬ ¬ ¬ ¬ Definition 2. As a result. However. is said to be continuous on a set ÊÒ ÊÑ is said to be piecewise continuous on a set Ë if Definition 2.1.1.1. ¼ ÊÒ ¼ exists (depending only on ¯) such that (2. Preliminaries Ê· ½ ÊÒ . uniform continuity implies continuity.9. for uniform continuity the same Æ “works” for all points of the set. A function if given an ¯ a constant Æ Ü Ý ¾ Ë.1) holds for all ´µ ´µ ÊÑ is said to be uniformly continuous on a set Ë Notice that uniform continuity is defined on a set. From the fact that × ÒÜ Ü   Ü¿ · Ç´Ü ½ µ and Ó× Ü ½   ½ ܾ · Ç´Ü µ ¾ . Furthermore. but but not necessarily vice versa. when Ü Ô is finite). The Ľ Ü ½ ¸ ×ÙÔ Ü´Øµ Ø ¼ and we say that Ü ¾ Ľ when Definition 2.1.1) if it is Definition 2. Ê Ê can be verified by means of the following Ê Å Lemma 2.4. If a constant Å ¾ Ê exists ×ÙÔ then is uniformly continuous on Ê. A function continuous at every point in Ë .

1.3) respectively.13. .3) For simplicity of notation we use the expressions × Ò Ü and ½  ÜÓ× Ü throughout this thesis.4) which is expressed by Ü ´Ø ܵ (2.1. (Khalil 1996)).2. The same holds true for similar expressions that at first glance seem not to be defined for Ü .2) and (2.4) and locally Lipschitz in Ü on . Ü whereas it would be more precise to use (2. « ¼ µ ¼ ½µ is said to belong to class à ´¼µ ¼.11.12. ½ function ¯ Ê Ü ¯ ´Üµ d ¯ dÜ ¼ Ü ¼ and ´¼µ ¼ 2.2 Lyapunov stability Consider a non-autonomous system described by Ê· ¢ ÊÒ is piecewise continuous on Ê· ¢ where Ò is a domain that contains the origin Ü Ê· ¢ . and Ê is an equilibrium point for (2.2.g.2) and ¾ ´Üµ are continuous. A continuous function ¬ ¼ µ ¢ ¼ ½µ ¼ ½µ is said to belong to class ÃÄ (¬ ¾ ÃÄ) if for each fixed × the mapping ¬ ´Ö ×µ belongs to class à with respect to Ö and if for each fixed Ö the mapping ¬ ´Ö ×µ is decreasing with respect to × and ¬ ´Ö ×µ ¼ Definition 2.1.. Definition 2. Lyapunov stability we can conclude that 15 РѼ × Ò Ü ½ Ü Ü ½ ´Üµ ½ ¼ ½ ¼ and ´ РѼ ½   ÜÓ× Ü ¼ Ü ½ ´ As a result the functions Ó×´Ü×µd× × Ò´Ü×µd× × ÒÜ Ü for Ü for Ü ¼ ¼ ¼ ¼ (2. A saturation function with saturation level ¯ is a  ¯ ¯ that satisfies ¼ ½. We assume that the origin ¼ ´Ø ¼µ ¼ Ø ¼ For studying the stability of the origin we introduce the following notions (see e. ¼ ½  Ó× Ü Ü for Ü for Ü (2. A continuous function (« ¾ Ã) if it is strictly increasing and « as × Definition 2.

4) is said to be (locally) exponentially Definition 2.4) to be semi-globally exponentially stable or globally exponentially stable (GES).2.6) ¯ semi-globally asymptotically stable if for each constant Ö ¼ and for all pairs ´Ø¼ ܴؼ µµ ¾ Ê· ¢ Ö a function ¬ ¾ ÃÄ exists such that (2.3.4) with ı ´Ø ܵ where ½   ½·Ø × Ò´Üµ  Ü if if Ü Ü ½ ½·Ø ½ ½·Ø × Ò´Üµ Then for each Ö and Ü Ø¼ Ö  ½ ´ µ ¼ ¼ and ؼ ¼ there exist constants ܴص ܴؼ µ  ­ ´Ø Ø¼ µ ¼ ½ if Ü if Ü if Ü ¼ ¼ ¼ ¼ and ­ ¼ such that for all Ø Ø Ø¼ ؼ (2. The equilibrium point Ü ¼ of (2. ¼ ¼­ ¼ ´ µ ´ · Example 2. i. Consider the system (2. The equilibrium point Ü stable if it is (locally) asymptotically stable and (2.4) is said to be ¯ (locally) asymptotically stable if a constant Ö ¼ exists such that for all pairs ´Ø¼ ܴؼ µµ ¾ Ê· ¢ Ö a function ¬ ¾ ÃÄ exists such that ܴص ¬ ´ ܴؼ µ Ø   ؼ µ Ø Ø¼ ¼ ܴؼµ ¾ Ö (2.2.2.4) is said to be (locally) stable (in the ¼ exists such that for all ´Ø¼ ܴؼµµ ¾ Ê· ¢ Ö Ø Ø¼ If the bound (2.6) is satisfied with globally asymptotically stable (GAS) if a function pairs ؼ Ü Ø¼ ¾ Ê· ¢ ÊÒ (2. ¯ of (2.6) holds.6) holds.16 Definition 2.5) ´ ´ µµ ¬ ¾ ÃÄ exists such that for all ¼ ¬ ´Ö ×µ Ö  ­× In a similar way we can define the equilibrium point Ü of (2. Lor´a and Teel 1999)).e..5) holds for all ´Ø¼ ܴؼ µµ ¾ Ê· ¢ ÊÒ . Preliminaries ¼ of (2. then the origin is globally stable. ܴص «´ ܴؼ µ µ ¼ ܴؼµ ¾ Ö (2.2.7) However.4 (see (Panteley. always a bounded (arbitrarily small) additive perturbation Æ Ø¼ exist such that the trajectories of the perturbed system Ü unbounded. The equilibrium point Ü sense of Lyapunov) if a positive constant Ö a function « ¾ à exists such that Chapter 2.1.2. Unfortunately this is in general not true for non-autonomous systems. ´Ø ܵ and a constant ´Ø ܵ · Æ´Ø Üµ are ¼ . under a uniformly bounded additional perturbation Æ Ü Ø solutions of the system Ü Ü Æ remain bounded. For linear time-invariant systems Ü Ü it is well-known that asymptotic stability is equivalent to GES and robustness with respect to perturbations is guaranteed. Definition 2.

1999).9) ¯ ¯ ¯ (locally) uniformly asymptotically stable if a constant Ö exist. for each value of ؼ different constants and ­ may be chosen. Lemma 5. The equilibrium point Ü of (2.5. such that «´ ܴؼ µ µ Ø Ø¼ ¼ ܴؼ µ ¾ Ö (2.2. where ´Æµ is a class à function of Æ.9) holds. ´ ´ µµ ¾ ÃÄ exists such ¼ ´ µ ¼ ¼ Having these definitions of uniform stability we are now able to formulate the following robustness result for uniformly asymptotically stable systems: be a uniformly asymptotically stable Lemma 2. Then one can and Ê such that for all perturbations Æ Ø Ü that satisfy the determine constants uniform bound Æ Ø Ü Æ and all initial conditions Ü Ø¼ Ê.4) is said to be uniformly stable if a positive constant Ö and an « ¾ à exist. ´ ´ µµ ¼ and for all Definition 2. Therefore. such that ܴص ¬ ´ ܴؼ µ Ø   ؼ µ Ø Ø¼ semi-globally uniformly asymptotically stable if for each constant Ö Ø¼ Ü Ø¼ ¾ Ê· ¢ Ö a function ¬ ¾ ÃÄ exists such that (2. we can Furthermore.2.9) holds. if Ü large enough. Let Ü equilibrium point of the nominal system Ü £ Ø Ü where Ê· ¢ Ö ÊÒ is continu¢ ously differentiable. One of the reasons for this negative result is that in (2. allow for arbitrarily large Æ by choosing Ê ¾ ÃÄ and some finite time ؽ . we introduce the notion of uniform stability. the solution Ü Ø of the perturbed system Ü Ø Ü Æ Ø Ü satisfies ´ µ ¼ ¡ ¼ ´ µ ܴص ܴص ¼ ¡ ´ µ· ´ µ ¬ ´ ܴؼ µ Ø   ؼ µ ´Æµ ´ µ ؼ Ø ´ µ ؽ ´µ Ø Ø½ and for some ¬ is a globally uniformly exponentially stable equilibrium point.8 ((Khalil 1996.4) is said to be (locally) uniformly exponentially stable/semi-globally uniformly exponentially stable/globally uniformly exponentially stable (GUES) if it is (locally) uniformly asymptotically stable/semi-globally uniformly asymptotically stable/globally uniformly asymptotically stable respectively and (2. i. Definition 2.6. then the origin is globally uniformly stable. Lyapunov stability 17 More details concerning the proof of the claims made in this example can be found in (Panteley et al.2.7) the constants and ­ are allowed to depend on ؼ . both independent of ؼ .2.8) holds for all ´Ø¼ ܴؼ µµ ¾ Ê· ¢ ÊÒ . and the Jacobian Ü is bounded on Ö . The equilibrium point Ü Ü´Øµ ¼ ¼ ¼ of (2. The equilibrium point Ü of (2.7.8) If the bound (2.2.4) is said to be ¼ and a function ¬ ¾ ÃÄ ¼ ܴؼµ ¾ Ö (2..3)).2. ¼ ¼ . Definition 2.e. uniformly in Ø. both independent of ؼ .9) is satisfied with ¬Ö× Ö  ­× ­ globally uniformly asymptotically stable (GUAS) if a function ¬ that for all ؼ Ü Ø¼ ¾ Ê· ¢ ÊÒ (2.

Let ½ and its derivative satisfies entiable function. but for uniform asymptotic stability instead. Example 2. If ¾ Ľ .10) . then ´Øµ and ¼ 2. then Ø Ø Ð Ñ ´Øµ ¼ ½ Ð Ñ ´Øµ ¼ ½ An extension of Barb˘ lat’s Lemma to functions that are not uniformly continuous (but can a be written as the sum of a uniformly continuous function and a piecewise continuous function that decays to zero) was presented in (Micaelli and Samson 1993): Ê· Ê be any differLemma 2. The equilibrium point Ü of (2.2. This explains why in this thesis we do not aim for asymptotic stability. one needs uniformity.8 it is also clear why exponential stability is a most favorable property. Notice that for autonomous systems solution depends only on Ø   ؼ .11. Then Suppose that Ø ½ ¼ ÐÑ ´µ Corollary 2.4 clearly shows that exponential convergence in itself does not guarantee robustness. However. ¾ Ľ .2.18 Chapter 2.2. If Ø converges to zero as Ø ´µ ´Øµ ¼ ´Øµ · ´Øµ Ø where ¼ is a uniformly continuous function and ¼ Ø tend to zero as Ø ½. Let a ÊØ d exists and is finite.12 ((Micaelli and Samson 1993.4) is said to be globally Ã-exponentially stable if a function ¾ à and a constant ­ exist such that for all ؼ Ü Ø¼ ¾ Ê· ¢ ÊÒ we have  ­ ´Ø Ø¼ µ ÜØ ÜØ Ø Ø ¼ ¼ ´µ ´ ´ µµ ´ ´ ¼µ µ ¼ ¼ A useful tool for showing asymptotic stability of a certain signal is: Ê· Ê be a uniformly continuous function. which could be an explanation for all the different notions of exponential stability that are available in literature. ´µ ´Øµ tends to zero as Ø ½. global uniform exponential stability can not always be achieved. Definition 2. Definition 2)). Lemma 2. Ü ´Üµ we may drop the word “uniform” as the From Lemma 2. Lemma 1)). and ¾ ÄÔ for some Ô ¾ ½ ½µ.10 (see (Barb˘ lat 1959)).3 Linear time-varying systems Consider the linear time-varying (LTV) system Ü Ý ´ØµÜ · ´ØµÙ ´ØµÜ (2. Preliminaries This implies that uniform asymptotic stability gives rise to some robustness that is not guaranteed by asymptotic stability.2. A notion that is equivalent to having both global uniform asymptotic stability and local uniform exponential stability (GUAS+LUES) is the following.2. Unfortunately.9 ((Sørdalen and Egeland 1995.2.

Let Ø Ø¼ denote the statetransition matrix for the system Ü Ø Ü. ¼ «½ ÁÒ Ï ´Ø   Æ Øµ for all Ø ¼ Corollary 2. (Kailath 1980. Lipschitz and constants Æ and ¯ exist such that ´µ ´ ´µ µ ´¼µ ¼ Ø ¼ ¼ ¼ ´ ´µ µ ´×µ × ¼ ´Øµ is × Ø Æ × Ø such that ¯ then the system (2. Theorem 2)). ´µ ´µ ´µ ´µ ¨´ µ Definition 2.3. Consider the linear time-varying system (2.e.. The pair Ø Ø is uniformly completely observable (UCO) if constants Æ ¯½ ¯¾ exist such that for all Ø : ¼ ´ ´ µ ´ µµ ¯½ ÁÒ Ø Ø Æ ¨Ì ´ Ø   Ƶ Ì ´ µ ´ µ¨´ Ø   Ƶd ¯¾ ÁÒ A very helpful theorem for showing UCC or UCO is Theorem 2. Rugh 1996)). . We recall two definitions from linear control theory (see e.g. Consider the system Ü Ý ´ ´ØµµÜ · Ü Ù (2.2.11) is continuous.10). Ù ¾ ÊÑ .3.3. If bounded.3.1. Linear time-varying systems 19 where Ü ¾ ÊÒ . . Ø . Assume that for all where the pair × is controllable (respectively the pair × is observable).10) is uniformly completely controllable if a constant Æ with Ø   Æ × Ø exist such that the matrix function Ï Ø   Æ Ø defined by ´ µ ¼ and an × Ø½ ؼ ´×µ´Ø   µ ½ ´ µ Ì´ µ Ì ´×µ´Ø   µd ½ where «½ ´Æ µ is a constant..2.11) is uniformly completely controllable (respectively observable). constants à and Ä exist such that ´µ ´µ ´µ ´Øµ ´Øµ ´Øµ   ´Ø¼ µ Ï ´Ø¼ ؽ µ satisfies Ã Ã Ä Ø   ؼ for all Ø for all Ø for all Ø Ø¼ ¼ ¼ ¼ Then the system (2. Ø are matrices of appropriate dimensions whose elements are piecewise continuous functions.3. Ý ¾ Ê .3 ((Kern 1982. i. and Ø .4. Suppose that Ø and Ø are bounded and that Ø is Lipschitz. The pair Ø Ø is uniformly completely controllable (UCC) if constants Æ ¯½ ¯¾ exist such that for all Ø : ¼ ´ ´ µ ´ µµ Ø·Æ Ø ¼ ¯½ ÁÒ ¨´Ø µ ´ µ Ì ´ µ¨Ì ´Ø µd ¼ ¯¾ÁÒ Definition 2. Ê Ê continuous.

no matter where we put this window on the graph.. A continuous function (PE) if all of the following conditions hold: ´µ Ê· Ê is said to be persistently exciting ¯ ¯ ¯ a constant à a constant Ä constants Æ ¼ exists such that ´Øµ à for all Ø ¼. . and Æ such that for all for Ø ¼ ¯½ ÁÒ Ø·Æ Ø ´ µ Ì ´ µd ´µ ´µ ¯¾ ÁÒ Furthermore. . Preliminaries The condition imposed on Ø in Corollary 2. Notice that in the common definition of persistence of excitation usually the first two assumptions on Ø are made implicitly.12) ¼ is globally uniformly exponentially stable (GUES) if ( Ò) are such that the polynomial ½ ´Øµ ¼ ´Øµ is persistently exciting and the · Ò Ò · ½ Ò ½ · ¡ ¡ ¡ · Ò ½ is Hurwitz (i.6.. See Appendix A. assuming the existence of positive constants ¯½ .3. . all its roots have negative real parts). . Ü (2. notice that the third condition on Ø as in Definition 2. It is known as the “persistence of excitation condition”. ¼ . . and ¼ and ¯ ¼ exist such that Ø ¼ × Ø   Æ × Ø such that ´×µ ¯ ´µ Remark 2. .3. ¼ . . Then. but also in identification and adaptive control systems. Theorem 2.e.5 can be interpreted as follows: assume that we plot the graph of Ø and look at this plot through a window of width Æ . not only in this thesis.7. . Definition 2.20 Chapter 2. The third condition is in general formulated Ê ÊÒ .. ¯¾.5.3. . The system ¾   ½   ¾ ´Øµ   ¿   ´Øµ ¿ Ü ´Øµ ¼ ¼ . ¼ ´µ ¼ The following are some useful results. . always a time instant × exists where × is at least ¯ .. .. . . .3. ¼ exists such that ´Øµ   ´Ø¼ µ Ä Ø   ؼ for all Ø Ø¼ ¼. .4 plays an important role.3. Proof. . ..

. .13. .. . . ¼ ¼ ´Øµ ¼ ¼ . .10. · н · ¡ ¡ ¡ · ÐÒ ½ · ÐÒ (2. . Consider the system ܽ ܾ   ¯ ´Ü½ µ · ´ØµÜ¾   ´ØµÜ½ (2. If Ø is persistently exciting (PE). . . then the system (2.. .4 Cascaded systems Consider a system Þ ´Ø Þµ that can be written as Þ½ ½ ´Ø Þ½ µ · ´Ø Þ½ Þ¾ µÞ¾ Þ¾ ¾ ´Ø Þ¾ µ (2.. . Ü ¼ ´Øµ .8. . The linear time-varying system (2... . Theorem 3.16b) . . .16a) (2. . . .4. (2. .4. . . . .2. . . . . ..15) where ¯ is a saturation function with saturation level ¯ as defined in Definition 2. . . . . . .. . .9 ((Ioannou and Sun 1996. . ¼ .3. .10) is globally uniformly exponentially stable (GUES) if and only if it is globally uniformly asymptotically stable (GUAS). . . . ¼ . . . .15) is globally Ã-exponentially stable. . . Cascaded systems Theorem 2. .3. .14a) Ò are Hurwitz.1. . . . ½ ¾ ´Øµ ¿ ´Øµ ¼ ¼ . ¼ ½ ¼ ¼  Ð ´Øµ  Ð¿ ¼  Ð¾ ´Øµ ´Øµ  Ð½ . . . . .14b) Theorem 2.13) ¼ Ü is globally uniformly exponentially stable (GUES) if ( Ò) are such that the polynomials ´Øµ is persistently exciting and the · Ò Ð Ò and · ½ Ò ½ · ¡ ¡ ¡ · Ò ½ Ò ½ (2. ¼ ¼ ¼ . The system ¾ 21   ½   ¾ ´Øµ   ¿   ´Øµ ´Øµ ¼ ¼ ¼ ... . . See Appendix A. . . . .6 v)). Proof.3. ´µ 2.. ¿ . . . Proposition 2. .. . .. . ..

19a) (2. Þ½ ¾ ÊÒ . i. Preliminaries spectively.. this is not true in general as can be seen from the following example. It is obvious that in that case also for (2. Þ¾ ¾ ÊÑ .e.19a) converge to (2. Example 2.16) become asymptotically stable. as a result the solution of Þ½ ´Øµ goes to infinity even quicker! ¾Þ½´¼µ Þ½ ´¼µÞ¾ ´¼µ  ­Ø · ¾   Þ½ ´¼µÞ¾ ´¼µ ­Ø Þ¾ ´¼µ  ­Ø (2.18) we have Ð ÑØ ½ Þ¾´Øµ ¼.22 where Chapter 2.18) Assume that the systems ¦½ and ¦¾ are asymptotically stable.16a) reduces to Þ½ ½ ´Ø Þ½ µ Þ½ Therefore. ½ ´Ø Þ½ µ is continuously differentiable in ´Ø Þ½ µ and ¾ ´Ø Þ¾ µ.19b) which can be seen as the system  Þ½  ­Þ¾ ­ (2.16a) reduces to the Unfortunately. One could consider increasing the gain ­ to make Þ¾ ´Øµ converge to zero faster and have the dynamics (2. ´Ø Þ½ Þ¾µ are continuous in their arguments. One would expect the system (2.16b) Þ¾ ´Øµ tends to zero.19) to be asymptotically stable.20a) that is perturbed by the output of the system ¼ (2. It seems plausible that therefore also (2.20b) are globally exponentially stable (GES). Þ½ Þ¾ Þ½ Þ¾ ¾  Þ½ · Þ½ Þ¾  ­Þ¾ ­ ¼ (2.17) ¦¾ Þ¾ ¾´Ø Þ¾ µ (2.21b) . Unfortunately. and locally Lipschitz in Þ¾ and ´Þ½ Þ¾µ re- Notice that if Þ¾ ¼ (2.16a) as the system ¦½ that is perturbed by the output of the system ½ ´Ø Þ½ µ (2.1.20b) Both (2.17) we know Ð ÑØ ½ Þ½´Øµ ¼ and for (2.17).20a) faster. However.21a) becomes zero at ½ Þ½ ´¼µÞ¾ ´¼µ Øesc ¾­ ÐÒ Þ½´¼µÞ¾´¼µ   ¾ so the solution of Þ½ ´Øµ goes to infinity in finite time.16a) and as a result the cascaded system (2. we can view (2. solving the differential equations (2. In that case the dynamics (2. for (2. Consider the system dynamics (2.19) yields Þ½ ´Øµ Þ¾ ´Øµ Notice that if Þ½ ´¼µÞ¾´¼µ ¾ the denominator of (2.4.21a) (2.20a) and (2.

16) when both ½ and ¾ are asymptotically stable: ¦ ¦ Lemma 2. For each fixed Þ¾ a continuous function and such that ­ ­ ­ ­ Ê· Ê· exists with Ð Ñ× ½ ´×µ Þ½ µÏ ´Þ½ µ ¼ with Î and Ï as in Assumption A1.17) and (2.16) are globally uniformly bounded. «½ «¾ ¾ ý .16) is globally uniformly asymptotically stable (GUAS). 2. The systems (2.18) are ı globally uniformly asymptotically stable (GUAS) and solutions of the cascaded system (2.3 ((Panteley and Lor´a 1999.4. Continuous functions ­ Î ´Ø Þ½ Þ¾µ­ ­ Þ½ ­ ´ ½ Ê· Ê· and « Ê· Ê· exist such that ´Ø Þ½ Þ¾µ ½´ Þ¾ µ« ´ Þ½ µ and a continuous non-decreasing function « Ê· Ê· and a constant such that (2. we can use the following: Theorem 2.24) Þ¾ Ö ¼. A4. « ¾ à and a positive semi-definite function Ï Þ½ such that ´ µ «½ ´ Þ½ µ Î ´Ø Þ½ µ «¾ ´ Þ½ µ Î Î · Þ½ ½´Ø Þ½µ  Ï ´Þ½µ Ø ­ ­ ­ ­ ´ µ (2. constants ¼ and Ï ´Þ½ µ ¼ exist such that for all Ø ¼ and all Þ½ ­ ­ ­ ­ ­ Î ´Ø Þ½ Þ¾µ­ ­ Þ½ ­ .17) and (2.23) ¼ exist « ´×µ and   ¡   ¡ « « ½ ´×µ « « ½ ´×µ ½ ½ ½ with «½ .18) are both globally uniformly asymptotically stable (GUAS) and we know explicitly a ½ Lyapunov function candidate Î Ø Þ½ .2. under certain conditions it is possible to conclude asymptotic stability of (2. A3.22b) (2.22c) Î Þ½ ­ ­ ­ ­ « ´ Þ½ µ A2.4. Theorems 1.22a) (2. The question that remains is when solutions of (2.4.2 ((Panteley and Lor´a 1999.16) are globally uniformly bounded. To answer that question. Lemma 1)). 4)). « as in Assumption A1. For each Ö « ´×µ d× ½ (2. If the systems (2. Consider the following ası sumptions A1. Cascaded systems 23 However. then the system (2.

24 A5. Lemma 2. then the cascaded system (2. then the cascaded system (2. A function Chapter 2. Proof.4. A3 and A4 hold. ´ µ ´ µ ´ ´ µ µ .25) and Assumption A5 is satisfied.23) is satisfied with « Then we have ½ ´ Þ¾ ´ Þ½ µ Ñ Ü´ ½ ´ µ ½ · Þ½ Þ¾ µ ¾´ Þ¾ µµ   ¡   ¡ « « ½ ´×µ « « ½ ´×µ ½ ½ Ô Ô Ô ½ × ½ · Ô½ ½ × ½ so that we can take « ´×µ If we take Ô Ô× · × ½ ½ in (2.16) is globally Ã-exponentially stable.17) and (2. If Assumption A1 is satisfied with ı «½ ´ Þ½ « ´ Þ½ continuous functions µ µ ½ Þ½ ¾ Þ½ Ê exist such that ½ Ê· Ê and ¾ Ê· ´Ø Þ½ Þ¾µ ½ ´ Þ¾ µ · ¾´ Þ¾ µ Þ½ (2.16) is globally uniformly asymptotically stable (GUAS). Then the cascaded system (2.18) satisfies ½ ؼ Þ¾´Øµ dØ ´ Þ¾´Ø¼µ µ Then we can conclude ¯ ¯ ¯ If Assumptions A1 and A2 hold. If Assumptions A1.24) we have that Assumption A3 is satisfied.16) is globally uniformly asymptotically stable (GUAS). « Þ½ Þ½ and positive semi-definite Ï and that Ø Þ½ Þ¾ satisfies (2.25). Lor´a and Nijmeijer 1998)). «¾ ¾ ý .4.22) with «½ Þ½ ½ Þ½ ¾ . Corollary 2. Lefeber.16) is globally uniformly asymptotically stable (GUAS).18) are globally Ã-exponentially stable. then the cascaded system (2. Preliminaries ¾ à exists such that the solution Þ¾ ´Øµ of (2. we know explicitly a ½ Lyapunov function candidate Î Ø Þ½ that satisfies (2. Assume that both subı systems (2.5 (see (Panteley. then the cascaded system (2. We have that (2.4 (see (Panteley and Lor´a 1998)).16) is globally uniformly asymptotically stable (GUAS). If Assumptions A1. A3 and A5 hold.

we could prescribe its des   ½ Ü   Ó× Ü ¸ «´Üµ   «´Üµ Next.26a) (2.5.6.4.5 Backstepping A commonly used method of nonlinear controller design is backstepping.27) To obtain a Lyapunov function candidate we simply augment the Lyapunov function with a quadratic term in Þ : Î ´Ü Î ´Ü Þ Ùµ µ Î ´Üµ · Þ ¾   The derivative of Î along the solutions of (2.27) becomes ½ ¾ ½ ܾ · ½ ´ · ¾ ¾ ½Ü · Ó× Üµ¾ ¡   ½ ܾ   Ü · Þ Ü · Ù · ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ .12)). Backstepping 25 Corollary 2.26) in the new co-ordinates ´Ü Þ µ: Ü   ½ Ü   Ü¿ · Þ Þ Ù · ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ Þ   des · (2. since the existence of a suitable Lyapunov function candidate is guaranteed from converse Lyapunov theory (see (Khalil 1996. c Example 2. A Unfortunately desired value is not the control but a state variable.26b) where Ü Ì ¾ ʾ is the state and Ù ¾ Ê is the input.25).18) is globally Ã-exponentially stable and that Ø Þ½ Þ¾ satisfies (2.16) is globally Ã-exponentially stable. that (2. Theorem 3.26a) can easily be stabilized by means of ½ ܾ . c Kanellakopoulos and Kokotovi´ 1995). we define Þ to be the difference between and its desired value: ½ Ü · Ó× Ü We can now write the system (2. 2. Krsti´ .5. This follows immediately from Lemma 2.4. Nevertheless.1. ´ µ Proof. We want to design a state-feedback controller to render the equilibrium point Ü Ì   Ì globally asymptotically stable (GAS). Assume that (2. Lyapunov function would be Î Ü ¾ ´µ   ½ Ü   Ó× Ü. Consider the second order system Ü Ù Ó× Ü   Ü¿ · ¼ ½ (2. Then the cascaded system (2. If were the input. We illustrate this method by means of a simple example considering the special case of integrator backstepping. For a more detailed explanation the reader is referred to (Marino and Tomei 1995.5.17) is globally uniformly exponentially stable (GUES). then (2.2.

26 The simplest way to arrive at a negative definite Î is to choose

Chapter 2. Preliminaries

Ù Ù

  ¾ Þ   Ü   ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ
Ü Ì
becomes

which in the original co-ordinates

 ´ ½ · ¾ µ   ´½ · ½ ¾ µÜ   ´ ½ · ¾ µ Ó× Ü · ½ Ü¿   Ü¿ × Ò Ü · × Ò Ü · × Ò Ü Ó× Ü

(2.28)

Usually is called a virtual control, « variable.

´Üµ a stabilizing function and Þ the corresponding error

From this example it is not difficult to see that the more general class of “triangular” nonlinear systems

Ü ½ ¾
. . .

´Üµ · ´Üµ ½ ½´Ü ½ µ · ½ ´Ü ½ µ ¾ ¾´Ü ½ ¾ µ · ¾ ´Ü ½ ¾ µ ¿
Ò ½ Ò

Ò ½ Ò

´Ü

´Ü
½

½

µ· Ò µ · Ò ´Ü
Ò ½

Ò ½

´Ü

½

½
Ò

µÙ

Ò ½ Ò

µ

can be stabilized in a similar way. First consider ½ as a virtual input to stabilize the first subsystem, define the error variable Þ½ , consider ¾ as a virtual input to stabilize the Ü Þ½ Ì subsystem, etc. Proceeding step by step along these lines one finally arrives at a control law for Ù. One of the advantages of backstepping is that it provides a constructive systematic method to arrive at globally stabilizing control laws. Unfortunately, one usually obtains complex expressions (in the original co-ordinates) for the control law, as already can be seen from (2.28).

Part I

A cascaded approach to tracking

27

Consider the second order system of Example 2.Chapter 3 Introduction to Part I 3.1.1a) (3. 1995) and forwarding for upper c triangular systems (Mazenc and Praly 1994.4. Marino and Tomei 1995. ´ µ ´ ¾ ´Ø Þ¾ µ.1b) can be concluded from the stability of the systems Þ½ ½ Ø Þ½ and Þ¾ implies that in the analysis we can simply “forget” about the term Ø Þ½ Þ¾ certain conditions) it does not play a crucial role. Mazenc and Praly 1996. Jankovi´ et al. This µÞ¾ since (under Example 3. We use the results on cascaded systems (Ortega 1991. Jankovi´ Sepulchre and Kokotovi´ 1996). associated Lyapunov functions for showing global stabilization are derived. Byrnes and Isidori 1989. Our goal is to arrive at less complex expressions and to gain more insight in the control laws. This is why we follow a different approach. However.5.1 Cascaded design In recent years recursive design methods for global stabilization of nonlinear systems have been developed. Two major design techniques can be distinguished: backstepping for lower triangular systems (Koditschek 1987.1. we can summarize Theorem 2.5. or to be more precise the result for timec varying systems as initially presented by Panteley and Lor´a (1998) and further developed in ı (Panteley and Lor´a 1999). a disadvantage is that the resulting control laws usually are complex expressions like in Example 2. ı Roughly speaking.1 after the change of co29 . Krsti´ et al.1.3 by saying that under certain conditions the stability of the system Þ½ Þ¾ ½ ´Ø Þ½ µ · ´Ø Þ½ Þ¾ µÞ¾ ¾ ´Ø Þ¾ µ (3. c One of the advantages of these methods is that they provide a systematic way of recursively designing feedback laws. For applying these methods the nonlinear system has to have (or should be transformed into) a certain triangular form. c. Tsinias 1989. 1996). Furthermore.

2a) results in the extra term ÜÞ in the derivative of the Lyapunov function.4.. As a result we can also view the system (3.2b) The backstepping approach resulted into the control law   ¾ Þ   Ü   ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ Ü Þ (3.30 ordinates.5 is globally asymptotically stable (GAS).2a) (3.5). since Þ is precisely the difference between the virtual control and its desired value that would have stabilized the Ü-subsystem. Notice that for Þ the system (3.2a) is left out (the term Þ in (3. Ù   ¾Þ   ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ Ü Þ (3.4) globally asymptotically stable we can claim global asymptotic stability of the overall system from the theory on cascaded systems. if we are able to render (3.2a) is globally asymptotically stable (GAS).4) that is perturbed by the output Þ of the system As a result.e. i. It turns out that the “ Ø Þ½ Þ¾ Þ¾ -part” of the subsystem (3. which is accounted for by an additional  Ü in the control law).5) renders the system (3.1 is not that we are able to leave out the term  Ü in the control law.3) and (3. This is not surprising. Introduction to Part I Ü Þ Ù   ½ Ü   Ü¿ · Þ Ù · ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ (3. the main lesson that can be learned from Example 3. Notice that a slight difference exists between the control laws (3. but that by recognizing a cascaded structure while designing a controller one might reduce the complexity of the controller. ´ µ · · Although the difference is not remarkable. With this control law we arrive at the overall closed-loop system   ½ Ü   Ü¿ · Þ   ¾Þ which according to Lemma 2. .1. consider the system Chapter 3.2) as the (by means of the desired virtual control stabilized) system ¼ ¦½ ¦¾ It is clear that the control law Ü Þ   ½ Ü   Ü¿ Ù · ´ ½   × Ò Üµ´  ½ Ü   Ü¿ · Þ µ (3.3).4) globally exponentially stable (GES).3) which results in the globally asymptotically stable (GAS) closed-loop system   ½ Ü   Ü¿ · Þ  Ü   ¾ Þ We could also have taken a slightly different approach before applying the final step in the backstepping design to arrive at (3.

Assume that a feasible reference trajectory ½Ö ¾Ö ¿Ö When we define the tracking-error Á¾   Á¿ ·Ù Á½ ¾ Ö ¿ Ö ½ Ö Á¿   Á½ ·Ù Á¾ ¿ Ö ½ Ö ¾ Ö Á½   Á¾ Á¿ ½ Ö ¾ Ö ´ Ö ÙÖ µ is given. i. An introductory example: tracking of a rotating rigid body 31 3.. but only the dynamics of the velocities. Then the dynamics for a rotating rigid body with two controls can be expressed as: where ( principal moments of inertia.e. a trajectory satisfying ½ ¾ ¿ Á¾   Á¿ ´ ¾ ¿   ¾ Ö ¿ Ö µ · Ù½   Ù½ Ö Á½ Á¿   Á½ ´ ¿ ½   ¿ Ö ½ Ö µ · Ù¾   Ù¾ Ö Á¾ Á½   Á¾ ´ ½ ¾   ½ Ö ¾ Öµ Á¿   Ö we obtain the tracking error dynamics (3. For reasons of simplicity.1.7c) . Therefore.1 gives rise to the question how to recognize a cascaded structure while designing a control law and how to guarantee that the closed-loop control system can be written in the form (3. we consider as an introductory example the tracking problem for a rotating rigid body. ½ Á¾   Á¿ ·Ù Á½ ¾ ¿ ½ Á¿   Á½ ·Ù ¾ Á¾ ¿ ½ ¾ Á½   Á¾ ¿ Á¿ ½ ¾ ¿) are the angular velocities and Á½ ½ Á¾ ¼ and Á¿ ¼ are the Notice that we assume Á½ Á¾ in order to be able to control ¿ by means of Ù½ and Ù¾ .7a) (3.2 An introductory example: tracking of a rotating rigid body Example 3. ¦ ¦ ¦ To make this more clear. we can also start with designing ¾ .7b) (3.3. also other directions can be taken. Clearly. for instance a spacecraft. it might suffice to stabilize only the difference between the virtual control and its desired value. we consider not the entire model. Then the assumption Á½ Á¾ can be made without loss of generality. without taking into account the way this error enters the remaining dynamics. One possible answer has been given there: follow a backstepping design and notice before applying the final step that one has a globally asymptotically stable (GAS) subsystem together with a corresponding Lyapunov function. Instead of starting from a system ½ and designing ¾ .2.1).

10) ¾ ¿ Á½  Á¾ Á¿ ¼ ½Ö Á¿  Á½ Á¾ ¼ ½Ö ¾ ¿ · ½ ¼ Ù¾   Ù¾ Ö ) simplifies the remaining dynamics In general this assumption (i. ¼ For the example of the rotating body this boils down to substituting dynamics. In that light it is good to remark that we can use one input for stabilization of a subsystem of the control system (3. a subsystem (3.e.8) then the subsystem (3. 3.4. i. That is what we focus on in the controller design.7)..1a) at (3.1b). Now we still have one input left that should be chosen such that the overall closed-loop system is rendered asymptotically stable. we can conclude that it might suffice in the controller design for the remaining input to render the part (3. since part of it can be forgotten. What we are left with is the part of the closedloop system that is described by (3.3. To start with.e.32 Chapter 3. ´ µ So how to proceed? Notice that it is fairly easy to arrive from (3. the substitution Þ¾ considerably. for asymptotic stability of the system (3.9) is asymptotically stable.7a) is rendered globally uniformly exponentially stable (GUES).3 we furthermore know that it might be sufficient too! As a result. We aim for a closed-loop system of the form (3. ¼ .1) it is necessary that the part Þ½ ½ ´Ø Þ½ µ (3.1).11) is globally and ¿ with ¾ Á¾ uniformly asymptotically stable (GUAS). Introduction to Part I We are interested in obtaining a closed-loop system of the form (3.1b). i.. From Theorem 2..9) asymptotically stable and “forget” about the Ø Þ½ Þ¾ Þ¾ part. In the closed-loop system this stabilized subsystem can be considered as the system ¾ . the system (3.9) and the problem has reduced to finding a control law for the second input that is such that this remaining part becomes globally uniformly asymptotically stable (GUAS). which results into the linear system ½ ¼ in the remaining (3. Before we proceed with the controller design we assume that the stabilization of this subsystem worked out. From Theorem 2. provided that ½ Ö is persistently exciting.11) Á¿  Á½ makes that the closed-loop system (3.1a). If we for instance take ¦ Ù½ Á  Á Ù½ Ö   ¾ ¿ ´ ¾ ¿   ¾ Ö ¿ Ö µ   ½ ½ Á½ ½ ¼ ¦ (3.e. In the first step we designed a control law for one of the two inputs in such a way that in closed loop a subsystem was stabilized. It is mainly a matter of substituting Þ¾ in (3.9). For the rotating body this can be guaranteed by a proper choice of the remaining input Ù¾ . This is also the way to proceed in the controller design.1).7 we know that the control law ¼ Ù¾ Ù¾ Ö   ¾ ¾   ¿ ½Ö ¿ (3. This should is something that be guaranteed by the controller design. we look for a way to obtain in closed loop a subsystem ¾ . Besides.10.

then the resulting closed-loop system (3.3 to conclude asymptotic stability of the overall closed-loop dynamics. i.2. but does this enable us to conclude asymptotic stability of the overall closed-loop system? Fortunately the answer is: yes.12) is globally Ã-exponentially stable.2.. ´µ ´µ The example of tracking the kinematics of a rotating body learned us that another way of obtaining a closed-loop system of the form (3.7b) GUES and then use Ù½ to stabilize the dynamics that remain then. We study three different examples of systems with two inputs that can be stabilized using this procedure.7a) GUES and then Ù¾ to stabilize the remaining dynamics.11).25). This is similar to interchanging the indices ¡ ½ and ¡ ¾ in both (3. Instead of first using Ù½ to render the subsystem (3. .12b)   ßÞ ½ ½ ¾ ¾ ´Ø Þ µ which has a clear cascaded structure. apply Theorem 2. 3. Remark 3. this procedure). Proposition 3.e.7. That is.1. ´ µ ´ µ ´ µ We can summarize this result as follows. i. 3..1b).2.12a) (3.e.7) in closed loop with the control laws (3. Consider the tracking error dynamics (3. 3. This follows immediately when we assume that both ¾ Ö and ¿ Ö are bounded. assume that the stabilization of Þ¾ has worked out (as guaranteed by the first step of in the remaining system.12) once we have that Ø Þ½ Þ¾ satisfies (2. in the next three chapters.3.8) and (3.2. An introductory example: tracking of a rotating rigid body In that case the closed-loop system (3.1) for a system with two inputs is the following: ¯ ¯ ¯ ¯ use one input for stabilizing a subsystem of the dynamics.8.8. In the overall closed-loop system this is the system (3. If Ö is bounded and ½ Ö is persistently exciting (PE). as well as the “connecting term” Ø Þ½ Þ¾ . it follows from Corollary 2.4. we can clearly recognize the systems ½ Ø Þ½ and Þ¾ ¾ Ø Þ¾ . we can also first use Ù¾ to render the subsystem (3. substitute Þ¾ ¼ use the other input to stabilize the simplified remaining system.6 that we can conclude global Ã-exponential stability of the system (3.11) can be written as 33 ¾ ¿ ½ Á½  Á¾ Á¿  ¾ ½Ö   ¿   Á Á Á ¿ ½ ½ ´Ø Þ µ ½ ßÞ ¼ ¾ ½Ö ¾ ¿ · Á¿  Á½ Á¾ Á½  Á¾ Á¿ ´ ´ ¿ ¾ ´Ø Þ ßÞ ½ · · Þ¾ µ ¿ Öµ ¾ Öµ ½ (3.11). ´ µ ´ µ ´ µ Þ½ We now have found an overall closed-loop system with a cascaded structure. Since the systems Þ½ ½ Ø Þ½ and Þ¾ ¾ Ø Þ¾ both are globally uniformly exponentially stable (GUES). This is the approach that we follow in this part of the thesis.4.

Introduction to Part I .34 Chapter 3.

We assume that the wheels do not slide. shown in Figure 4.Chapter 4 Tracking of a mobile robot 4. It is assumed that the masses and inertias of the wheels are negligible Ý Ü Figure 4.1) 35 . and the angle between the heading direction and the Ü-axis. and that both the forward velocity Ú and angular velocity can be controlled independently by motors. which results in the following equations ´ µ Ü Ý where Ú and are considered as inputs.1.1: A two-wheel mobile robot. Let Ü Ý denote the co-ordinates of the center of mass. Ú Ó× Ú× Ò (4.1 Introduction In this chapter we study the tracking problem for a wheeled mobile robot of the unicycle type.

Consequently either discontinuous or time-varying (or both) controllers are needed for the stabilization problem. Tilbury. Tracking of a mobile robot Notice that the no-slip condition imposes the non-holonomic constraint Ü× Ò   Ý Ó× ¼ As a result. Assume that feasible reference dynamics ÜÖ ÝÖ Ö ÚÖ Ö Ì is given. The first global tracking control law that we are aware of was proposed by Samson and Ait-Abderrahim (1991). All these papers solve the local tracking problem.36 Chapter 4. Murray. the system (4.. which explains the interest of many researchers in this simple model. Sastry. i. Walsh. Murray and Laumond 1994. (1990) (cf. the errors are considered in a frame attached to the mobile robot. In (Kanayama.1) fails to meet Brockett’s necessary condition for feedback stabilization (Brockett 1983). This implies that no smooth (or even continuous) time-invariant static state-feedback law Ù Ù Ü exists which makes a specified equilibrium of the closedloop locally asymptotically stable. ´µ Although the stabilization problem for wheeled mobile robots is now well understood. As a matter of fact. it is not clear that the current stabilization methodologies can be extended easily to tracking problems.2): ¾ Ü Ý ¿ ¾ Ó×  × Ò ¼ × Ò ¼¿ ¾ÜÖ   Ü¿ Ó× ¼ ÝÖ   Ý ¼ ½ Ö  (4. Miyazaki and Noguchi 1990. Fierro and Lewis 1995) a linearization-based tracking control scheme was derived.3) . For an overview we refer to the survey paper of Kolmanovsky and McClamroch (1995) and references cited therein. Fliess. Kimura. Levine.2) This global change of co-ordinates from ÜÖ   Ü ÝÖ   Ý Ì to Ü Ý Ì makes that the errorvariables become independent from the choice of the inertial co-ordinate frame. Figure 4. The idea of inputoutput linearization was used by Oelen and van Amerongen (1994). Another global tracking result was derived by Jiang and Nijmeijer (1997) using integrator backstepping. dynamics that satisfies ´ µ ÜÖ ÝÖ Ö ÚÖ Ó× Ö ÚÖ × Ò Ö Ö For solving the tracking control problem the following global change of co-ordinates was proposed by Kanayama et al. Walsh and Sastry 1992.e. the tracking problem has received less attention. Martin and Rouchon (1995) dealt with the trajectory stabilization problem by means of a flatness approach. Micaelli and Samson 1993. In these new co-ordinates the tracking error dynamics becomes: Ü Ý   Ü · ÚÖ ´Øµ × Ò Ö ´Øµ   Ý   Ú · ÚÖ ´Øµ Ó× (4.

2. in Section 4. we recover these two results that both achieve global asymptotic stability (GAS) of the tracking error dynamics.2.6. By means of a cascaded design approach we derive a controller that achieves global Ã-exponential stability. .4 the statefeedback problem under input saturation. As mentioned in the introduction.4).1 Previous results We first summarize the available global tracking results.2 State-feedback In this section we study the state-feedback tracking control problem for a mobile robot. we are aware of two global tracking results.2: The new error co-ordinates.5. that the tracking error Ü Ý ´ µ such This is the problem studied in this chapter. The performance of the derived controllers is illustrated by means of simulations in Section 4. and in Section 4.2.3 the output-feedback problem. 4. State-feedback 37 Ü Ý ÝÖ Ý Ü ÜÖ Figure 4. We conclude this chapter with some final remarks in Section 4. Example 2.4. Subsequently we study in Section 4. The tracking control problem boils down to finding appropriate control laws for Ú and Ì converges to zero. For reasons of robustness we would like to be able to conclude global uniform asymptotic stability (GUAS) (cf. First.2 the statefeedback problem. 4.

backstepping based).4b) ¿ · ½ Ú ´ØµÝ × Ò   Ü µ · ´½   ½ µÝ ´ Ö ´Øµ · Ö ¾ ¾ ¾ ¾ .5a) Ó× · ¿Ü   ¾ Ö (4.3) in closed loop with the control law ¿ · ½ Ú ´ØµÝ × Ò   Ü (4. 4. ¿ . . Remark 4. 4. If ÚÖ Ø . ¾ . .2. Consider the tracking error dynamics (4.4) where we take ½ ½.38 Chapter 4.3) in closed loop with the control law Ú ¾ ´ØµÝ   ½ ¾ÚÖ ´ØµÝ¾ × Ò · ½ ¾ ¾Ý ¾ ¾   ½ ¾ ¿ Ý   ¾ ½ ¾ ÚÖ ´Øµ¾ Ý × Ò · ¿ ½ ¾ ÚÖ ´ØµÜ Ý × Ò · ¿ ½ ¾ ÚÖ ´Øµ Ö ´ØµÜ Ý × Ò   ½ ¾ ¿ ÚÖ ´ØµÝ¾ × Ò   ¾ ¿ Ö ´ØµÝ ¾ · ½ ÚÖ ´ØµÝ¾ Ó×   ¾ ÚÖ ´Øµ Ö ´Øµ × Ò   ½ ¾ ÚÖ ´Øµ × Ò ¾ ¾ ¾ · ¾ ½ ¾ Ö ´ØµÜ · ¾ Ö ´Øµ¾ Ü · ½ ¾ Ü ¾ · ¾ ½ ¾ ÚÖ ´Øµ¾Ü ݾ × Ò ¾ ¾ · ¾ ¾ ÚÖ ´Øµ¾Ý¿ × Ò Ó×   × Ò ÚÖ ´Øµ Ö ´Øµ · ½ · ½ ÚÖ ´ØµÝ ×Ò (4. ¿ ¼.6a) (4. This boils down to the controller (4. instead of differentiable with bounded derivative). and and are arbitrary constants.1 (Samson and Ait-Abderrahim (1991).6). then the closed-loop system (4.4) is globally asymptotically stable (GAS). then the closed-loop Ö ´Øµ · system (4.3.2. Consider the tracking error dynamics (4. Lyapunov based). ÚÖ Ø . Ö ´Øµ are bounded and either ÚÖ ´Øµ or Ö ´Øµ does not converge to zero.5b) ½ ¿ . ¼ ´µ ´µ ´µ ½ ´µ ´µ ¼ . . Tracking of a mobile robot Proposition 4. However. provided that ÚÖ Ø and Ö Ø are uniformly continuous and bounded. ¾ ¼.3.5) is globally asymptotically stable (GAS). 4.6b) can be shown to yield GAS of the closed-loop system (4. ¿ .3.2. Ö Ø and Ö Ø are bounded where ½ and either ÚÖ Ø or Ö Ø does not converge to zero. ¾ . If ÚÖ ´Øµ and where ½ ¼. Jiang and Nijmeijer (1997) remarked that by means of Lyapunov theory the control law ¼ ´µ ¼ ´µ ¼ ¼ ´µ ´µ ´µ ´µ Ú ´Øµ · ÚÖ ´ØµÝ × Ò · ÚÖ ´Øµ Ó× · ¾ Ü Ö ½ ½ ¾ ¼ ¼ ´µ ½ (4. and either ÚÖ Ø or Ö Ø does not converge to zero. ¾ . Proposition 4.2 (Jiang and Nijmeijer (1997).3.4a) Ö ¾ ¾ ¾ Ó×   ½ · µ Ú ÚÖ ´Øµ · ¿ Ü · ´¾ ¿ · ÚÖ ´Øµ (4. the assumption on ÚÖ Ø and Ö Ø is slightly weaker (uniform ½ continuous and bounded.

9) as the system ½ (4. As inputs we have the forward velocity Ú and the angular velocity . forward and angular velocity of the reference trajectory to be tracked. (4.2.2 A cascaded design In this section we derive a controller for the tracking control problem for a mobile robot.7a) (4.4). The control law ´µ ´µ Ö ´Øµ · ½ ½ ¼ ¼ (4. From Theorem 2. .7 we know that if Ö persistently exciting (PE).10) The remaining dynamics is then given by Ü Ý ´ØµÝ · ½ Ý   Ú · ÚÖ ´Øµ Ó×   Ö ´ØµÜ   ½ Ü · ÚÖ ´Øµ × Ò Ö We proceed by assuming that the stabilization of has been established. Section 2.11.9) ¦¾ (cf. As pointed out in Section 3. 4.9) globally uniformly asymptotically stable (GUAS). Recall that the tracking error dynamics for a mobile robot can be described by Ü Ý Ý   Ú · ÚÖ ´Øµ Ó×   Ü · ÚÖ ´Øµ × Ò Ö ´Øµ   (4.7c) is the orientation error and ÚÖ Ø and Ö Ø are the where Ü and Ý are position errors.4. then the control law ´Øµ is Ú ÚÖ ´Øµ · ¾ Ü   ¿ Ö ´ØµÝ ¾ ¼ ¿  ½ (4.3.7c) can easily be stabilized.1a) and focus on rendering (3.2 we first use one input for stabilization of a subsystem. By means of the input the dynamics (4.2.7b) (4.8) results into the globally uniformly exponentially stable (GUES) subsystem  ½ We can think of (4. As a result we obtain the following.2. which results into ´ µ ´µ ¼ Ü Ý   Ö ´Øµ ¼ Ö ´Øµ ¼ Ü Ý · ½ ¼ ÚÖ ´Øµ   Ú (4. to achieve globally uniformly asymptotically stable (GUAS) tracking error dynamics. The next step therefore is substituting Ø in the remaining dynamics (4. State-feedback 39 4.12) globally uniformly exponentially stable (GUES). For that we use the cascaded systems based approach sketched in Section 3. What we do is to “forget” about the Ø Þ½ Þ¾ Þ¾ part of the dynamics (3.11) which simply is a linear time-varying system.10).12) renders the resulting closed-loop system (4.

2. This result was originally presented by Panteley et al. ´µ ´µ Notice that due to the cascaded design approach we were able to reduce the problem of stabilizing the nonlinear tracking error dynamics (4.4) or (4. the controllers (4.5) versus (4. The price we pay is that (4. Applying Corollary 2. 1 Most likely the control laws (4.5) also yield GUAS of the resulting closed-loop tracking error dynamics.7) in closed loop with the control law Ú Ö ´Øµ · ½ ÚÖ ´Øµ · ¾ Ü   ¿ Ö ´ØµÝ ½ ¾ ¼ ¼ ¿  ½ (4.6 completes the proof.2.4. Tracking of a mobile robot Proposition 4. whereas for (4.4) and (4.13b) If ÚÖ Ø is bounded and Ö Ø is persistently exciting (PE) then the closed-loop system (4.7) to the problem of stabilizing the linear systems Ö and ´Øµ   ¼ Ö (4.13) is globally Ã-exponentially stable.5) were only shown to yield globally asymptotically stable (GAS) closed-loop tracking error dynamics.15) So. Due to the design we obtain a cascaded structure for the closed-loop system: ´µ ´µ Ü Ý   ¾ ´ ¿ · ½µ Ö ´Øµ Ü · ½ Ý · ÚÖ ´Øµ Ó×  ½   Ö ´Øµ ¼ Ý   ½ Ü · ÚÖ ´Øµ × Ò ßÞ ½   ßÞ ½ ¾ ´Ø Þ µ ½ ´Ø Þ ßÞ ½ Þ¾ µ ´Ø Þ µ ¾ Notice that the systems Þ½ ½ Ø Þ½ and Þ¾ ¾ Ø Þ¾ are globally uniformly exponentially stable (GUES).2.5). where was used.5.2.7. .4. the boundedness of ÚÖ Ø guarantees that the assumption on Ø Þ½ Þ¾ is met.13a) (4. Furthermore. Consider the tracking error dynamics (4.2.4) and (4.1 and 4. a sense we reduced the nonlinear problem into two easy-to-solve linear problems. Proof.2. Since Ó×  ½ and × Ò are bounded. a difference in complexity can be noticed in (4.13) makes it impossible to track a reference for which Ö Ø tends to zero but ÚÖ Ø does not. (1998). ´ µ ´ µ ´ µ ´µ Remark 4. ¿ ¼ When we compare the result of Proposition 4.13).40 Chapter 4. which is something that can be dealt with using (4.4) and (4. 4.13) we were able to show the more desirable property of global uniform asymptotic stability (GUAS)1 .14) Ü Ý   Ö ´Øµ ´Øµ ¼ Ü Ý · ½ ¼ ÚÖ ´Øµ   Ú (4.4 with the results as presented in Propositions 4.

7a. instead of going through a nonlinear À½ design. Remark 4.6. while this is the first thing . In case we have noisy measurements. 4. From the cascaded design we obtained for ¿ subsystem ¼ ¦ Ü Ý   ¾ Ü · Ö ´ØµÝ   Ö ´ØµÜ (4. Since the tracking error dynamics are nonlinear.4. Therefore.2.6).5). it suffices to solve the problem for a first order and a second order linear system.7b) stabilized by means of the input Ú ÚÖ Ø . since it makes it impossible to follow straight lines.15). Similar reasoning can be used in case of more general additive disturbances. this approach is in general not guaranteed to work. it is common practice to filter the measurements and use the filtered state for feedback. The only difficulty is to recognize the simpler structure for backstepping.15). Corollary 2.15) are rendered asymptotically stable. This structure became clear from the cascaded controller design. We can show global asymptotic stability ¾ Ü and the virtual control (GAS) of the system (4. Assume for instance that a constant disturbance is perturbing the system (4.4.14) and (4. Then one can for instance use À½ control techniques for arriving at robust controllers for the two linear systems. the robust controller design for the nonlinear system (4.2. Both can easily be solved by adding integral action.7) into two linear ones.6 then guarantees Ã-exponential stability of the nonlinear tracking error dynamics.7. The requirement that Ö has to be persistently exciting (PE)is a serious practical limitation.7) under constant disturbances simply reduces to the robust controller design for the linear system (4. State-feedback 41 This observation can be very helpful in finding answers to questions that come from a more practical point of view. it also provides an eye-opener to recognizing a the ½ simpler structure for backstepping. In practice we also have to deal with disturbances due to errors in the model or due to imperfect state measurements. backstepping not necessarily has to lead to complex expressions for control laws as (4. we simply design filters such that the linear systems (4.14) and the linear time-varying system (4. we obtain the control law Ú ´Øµ · ÚÖ ´ØµÝ × Ò · ÚÖ ´Øµ Ó× · ¾ Ü Ö ¼ ¼ which is exactly the controller (4.6).16) by means of the Lyapunov function candidate ´ µ Ó× · ¼ Î ½ ܾ · ½ ݾ ¾ ¾ ½ ½ ¾ and some additional standard Lyapunov techniques.16) which can be seen as the subsystem (4. For design purposes we can simply assume that this constant disturbance is perturbing the systems (4. for linear systems this approach can be applied successfully. but can also result in more simple expressions as (4. Not only can the cascaded design reduce the nonlinear controller design problem for the system (4. Remark 4. If we now ‘step back’ the virtual control to the true input .2. Therefore.7). However. Instead of solving the problem for a third order nonlinear system. Therefore.14) and (4.

19b) It is clear that we can still use the control law (4.20a) ݽ   ݽ (4. 4.18a) (4. we assume that we are unable to measure Ü . the control laws derived in the previous section can easily be extended. from Theorem 2. However.17) From a cascaded design point of view. but also with stabilization.19). Therefore. Panteley and Teel (1999b).7) where we are only allowed to use the measured output for designing the control laws for Ú and .18).18b) Ü Ý Ý½ Ý   Ö ´Øµ ¼ Ö ´Øµ ¼ · ½ ¼ ÚÖ ´Øµ   Ú (4. In that ݽ ݾ Ö Ý (4.42 Chapter 4. we know that we only have to stabilize the systems ݾ and ´Øµ   Ü Ý (4. we consider in the following sections the cases where we measure two of the state components. which makes it impossible to reconstruct the state from the measurements.1 Unmeasured Ü First.3. we study the problem of stabilizing the tracking error dynamics (4. This weakened version of PE makes it not ı only possible to deal with tracking of straight lines.3 Dynamic output-feedback In this section we study the dynamic output-feedback tracking control problem for a mobile robot.20b) (4. With the cascaded design from the previous section in mind. One way is to overcome this difficultie is sketched in the previous remark. global uniform asymptotic stability (GUAS) can be shown. By using this concept.19a) (4. The only problem is to stabilize (4. 4. That is. An other idea is using the idea of uniform Æ -persistence of excitation as introduced by Lor´a. Tracking of a mobile robot one would like to do in practice.3.8) for stabilizing (4.20c) . but that we can measure Ý and case the available output is .8 we know that the dynamic output-feedback Ú Ü Ý Ý½ ÚÖ ´Øµ · ¾ Ü   Ö ´Øµ ¢ ¼ Ö ¼ ½£ Ü Ý ´Øµ ¼   ¿ Ö ´ØµÝ Ü Ý · ½ ¼ ÚÖ ´Øµ   Ú ·  Ð¾Ð½Ö ´Øµ (4. In case we are able to measure only one of the state-components we end up with an unobservable system.

a system of the form (3..4.18. As a result we obtain: Proposition 4. н .e. even though we can measure Ý we also have generated an estimate for Ý .20c) is a full order observer for the system (4.19). we define a new variable Þ as where Ø is a function still to be determined in order to guarantee asymptotic stability of the reduced order observer.7. Dynamic output-feedback 43 with ¾ . ¿   . i.4.e. and о   renders the closed-loop system (4.8. ¼ ½ ¼ ½ ´µ ´µ Proof.6 completes the proof.8 we know that the system Þ½   ½ Þ¾ ¾ ¼ ¼ ¼ ¼   Ö ´Øµ ¼ ¼ ¾ ¼ ´Ð¾ · ½µ Ö ´Øµ  Ð½   ¿ Ö ´Øµ ¿ Þ½ ´ ´µ ´ µ ´ µ µ Notice that (4. 4. 4. Consider the tracking error dynamics (4. The boundedness of ÚÖ Ø guarantees that the assumption on Ø Þ½ Þ¾ is met. 4.19) we try to estimate some linear combination of the measured and the unknown signals.17) in closed loop with the control laws (4. We can see the closed-loop system (4.20) as a cascaded system. It is also possible to use a reduced order observer. 4.17.20b. ¿ ½ Ý · ÚÖ ´Øµ Ó×  ½ Ú µ× Ò ´Ø Þ½ Þ¾µ   ݽܷ ·Ö ´ØÖµ´ØÓ×  ½ Ú ½   ½ Ü · ÚÖ ´Øµ × Ò From Theorem 2. In order to find a reduced observer for the system (4. 4.3. Differentiating Þ with respect to time along the dynamics (4.e. i. to reconstruct only the unknown signal Ü .1. i...3.7) with output (4. It is also clear that the system Þ¾ ¾ Ø Þ¾ is GUES. Applying Corollary 2.20) globally uniformly exponentially stable (GUES). To be precise.19) yields ´µ Þ Ü   ´ØµÝ½ Þ ´Øµ Ö ´Øµ´Ü   ´ØµÝ µ · ´Øµ¾ Ö ´ØµÝ · Ö ´ØµÝ · ÚÖ ´Øµ   Ú   d d´Øص Ý ´Øµ Ö ´ØµÞ · ´Øµ¾ Ö ´Øµ · Ö ´Øµ   d ´Øµ Ý · ÚÖ ´Øµ   Ú dØ Ö ´ØµÝ · ÚÖ ´Øµ   Ú   d d´Øص Ý · ´Øµ Ö ´ØµÜ .1) where ¢ £Ì Þ½ Þ¾ Ü Ý Ü  Ü Ý  Ý ¾ ½ ´Ø Þ½ µ ¾ ´Ø Þ¾ µ   ¾ ´ ¿ · ½µ Ö ´Øµ   Ö ´Øµ ¼ ½ Ø Þ½ is globally uniformly exponentially stable (GUES). Assume that Ö Ø is persistently exciting (PE) and that ÚÖ Ø is bounded. 4.20). Then the resulting closed-loop system is globally Ã-exponentially stable.8.3.

22a) (4.3. then the closed-loop system (4.44 Chapter 4.7.1) where Þ½ Þ¾ ¢ Ü Ý Ü  Ü £Ì ¾ ½ ´Ø Þ½ µ ¾ ´Ø Þ¾ µ ¾   ¾ ´ ¿ · ½µ Ö ´Øµ   Ö ´Øµ ¼   ¾ Þ¾ ¼ ¼  Ð Ö ´Øµ¾ ¼ ¾ ¿ Þ½ ¿ ´Ø Þ½ Þ¾µ ½ Ý · ÚÖ ´Øµ Ó×  ½   ½ Ü · ÚÖ ´Øµ × Ò ½ Ý · ÚÖ ´Øµ Ó×  ½ · Ð Ö ´Øµ   ½ Ü · ÚÖ ´Øµ × Ò ¾ Þ¾ Þ¾ µ (4.17) in closedloop with the control law Ú Ü Þ Ö ´Øµ · ½ ÚÖ ´Øµ · ¾ Ü   ¿ Ö ´ØµÝ ½ ¾ ¼ ¼ ¿  ½ (4. ´µ ´µ ´µ Þ ´Øµ Ö ´ØµÞ We can combine this reduced observer with the controller (4.e.22b) where Ü is generated by the reduced order observer 4.23a) ¾ Þ · о Ö ´Øµ¿ · Ö ´Øµ · Ð Ö ´Øµ Ý · ÚÖ ´Øµ   Ú  Ð Ö ´Øµ (4. Consider the tracking error dynamics (4. we are able to conclude global uniform exponential stability (GUES) of (4.23b) If ÚÖ ´Øµ is bounded and Ö ´Øµ is persistently exciting (PE). Proof.. which can also be expressed as Ü Ý Þ¾   ¾ ´ ¿ · ½µ Ö ´Øµ Ü ·   Ö ´Øµ ¼ Ý ßÞ ½ ´ µ  Ð Ö ´Øµ¾ Þ¾ ´Ø Þ µ ½ ´Ø Þ ¼ ßÞ ½ (4.23) is globally Ã-exponentially stable.12): Proposition 4.24a) . 4.21).22.21) we obtain for the observation-error Þ If we now take Ø  Ð Ö Ø with Ð a positive constant and we furthermore assume that Ö Ø is persistently exciting (PE). Þ   Ð Ö ´ØµÝ Ð ¼ (4. 4.7) with output (4.24b) To be able to apply Corollary 2.23) as a cascaded system. We can view the closed-loop system (4. 4.22. i. Tracking of a mobile robot In case we define the reduced order observer dynamics as Þ ´Øµ Ö ´ØµÞ · ´Øµ¾ Ö ´Øµ · Ö ´Øµ   d d´Øص Þ Þ Ý · ÚÖ ´Øµ   Ú (4.6 we need to verify global uniform exponential stability (GUES) of the system Þ½ ½ Ø Þ½ .7.4. a system of the form (3.2.

5) guarantees that the condition on Ø Þ½ Þ¾ is met. Ú Ö ´ØµÝ ¿  ½ where Ü and Ý are generated by the full order observer Ü ¼ ½ н Ö ´Øµ Ü   Ö ´Øµ ¼ Ý · ¼ ÚÖ ´Øµ   Ú · о Ö ´Øµ ݽ   ݽ Ý ¢ £ ݽ ½ ¼ Ü Ý where н ¼.3.7 the system Þ ½ ½ Ø Þ½ is GUES.9 enables us to conclude that Þ½ ½ Ø Þ½ is GUES.3. Proposition 4.2 Unmeasured Ý In case we assume that we are unable to measure Ý . we have the existence of constants such that for all Ø : Ø·Æ ¾ ½ ¾ Ö d Ø 45 ´µ ¼ Æ ½ ¾ ¼ ´µ Therefore.3.4.4.4. we have ݽ ݾ Ü (4. Corollary 2. о  ½.25) Since we can repeat the reasoning of the previous section.3.7) with output (4.26a) (4.3.3. Since also the system Þ¾ ¾ Ø Þ¾ is GUES and boundedness of both ÚÖ Ø and Ö Ø (cf.25) in closed loop with the control law bounded.27b) . we summarize this analysis in the following two propositions.25) in closedloop with the control law Ú Ý Þ Ö ÚÖ ´Øµ · ´Øµ · ½ ¾Ü   ¿ Ö ´ØµÝ ½ ¾ ¼ ¼ ¿  ½ Ð (4.26b) where Ý is generated by the reduced order observer Þ · Ð Ö ´ØµÜ  Ð Ö ´Øµ¾ Þ   о Ö ´Øµ¿ · Ö ´Øµ · Ð Ö ´Øµ Ü · Ð Ö ´Øµ Ú   ÚÖ ´Øµ ¼ (4. Consider the tracking error dynamics (4. Dynamic output-feedback Since Ö Ø is persistently exciting (PE).27a) (4.4. the subsystem (4.24b) is GUES. Since it is a linear time-varying system Theorem 2. ´ µ ´ µ ´ µ ´ µ ´µ ´µ ´ µ ´ µ 4.6 yields the desired result. Definition 2. Furthermore. Then the resulting closed-loop system is globally Ã-exponentially stable. Consider the tracking error dynamics (4. From Corollary 2. the term Ø Þ½ Þ¾ is bounded and according to Theorem 2. but we can measure Ü and the outputs .6 we can conclude that the system Þ½ Ø Þ½ is globally uniformly asymptotically stable ½ (GUAS).3. Assume that Ö ´Øµ is persistently exciting (PE) and that ÚÖ ´Øµ is Ö ÚÖ ´Øµ · ´Øµ · ½ ¾Ü   ¿ ½ ¾ ¼ ¼ Proposition 4.3.7) with output (4.

In practice. Therefore.28a) (4.2. 4..29a) (4.e. it is interesting to take these input constraints into account when designing control laws. In this section we study the global tracking control problem for a mobile robot under input saturation.7) yields Î Ü ½ · ܾ · ݾ ´ Ú · ÚÖ ´Øµ Ü ½ · ܾ · ݾ ´ Ú · ÚÖ ´Øµ Ú´ Ý Ó× µ · ½ ·ÖÜص· ݾ × Ò · ¯½½ ´ Ö ´Øµ   µ ¾ ¯½ ÚÖ ´ØµÝ × Ò Ó× µ · ¯½½   Ö ´Øµ · ½ · ܾ · ݾ . the larger the control.46 Chapter 4.4 Saturated control All control laws mentioned in this chapter have one thing in common: the larger the errors. Tracking of a mobile robot If ÚÖ Ø is bounded and Ö Ø is persistently exciting (PE).4. ´µ ´µ 4. however. In order to be able to do so.6).29b) ¼ Under these feasibility conditions we look for controllers for Ú and that always meet (4.26. in the controller design we take into account the constraints ڴص ´Øµ Úmax max Ø Ø ¼ ¼ (4.1 A Lyapunov design Our first approach is a Lyapunov design similar to the one which results in the (unsaturated) control law (4. the input is constrained: the mobile car has a maximum forward and angular velocity.30) Differentiating Î along the solutions of (4.27) is globally Ã-exponentially stable.28) and still guarantee global uniform asymptotic stability of the tracking error dynamics. 4. This means that the reference forward and angular velocity should not exceed the limits: ×ÙÔ ÚÖ ´Øµ Ø ¼ ×ÙÔ Ö ´Øµ Ø Úmax max (4. then the closed-loop system (4. Inspired by Remark 4.7. we assume that once we are exactly on the desired trajectory we can stay on it. i. 4.6 and (Jiang and Praly 1992) we consider the Lyapunov function candidate Î ½ ÐÓ ´½ · ܾ · ݾµ · ½ ¾ ¾¯½ ¾ ¯½ ¼ (4.28b) We would like to design controllers such that they never result into a forward and/or angular velocity exceeding the limits of the mobile car.

2.32) and (4.4.10) that a ´µ ´µ ´µ ´µ Ø Ü´ ¯ Ü´ Ð Ñ ½ · Üص´Øµ¾´· Ýصµµ¾ · ¯½ ´Øµ ¯ ´ ´Øµµ ¼ ½ ´Ø ½ ¿ ¾ which. results in ½ ¯¾ ´ µ ¼ ´ ´ µ ´ µ ´ µµ ´µ As a result.31b) a saturation function as defined in Definition 2.31a) (4.32) Next. using (4.2. ¯¿ ¼.12 with ´Øµ ½ Ö´ µ ´Øµ ½¯·ÚܾطÝݾ × Ò   ½ · ¯ ´ µ ¾ Ø Ð Ñ ´ Ü ´Øµ · ´Øµ µ ¼ ½ which enables us to conclude that Ø As a result also Ð Ñ ½ · Ü ´Øµ¯¾½ · Ý ´Øµ¾ ÚÖ ´ØµÝ ´Øµ ¼ ½ Ø Ð Ñ ÚÖ ´ØµÝ ´Øµ ¼ ½ (4. Saturated control Choosing 47 Ú with ½ Ö´ µ ´Øµ · ½¯·ÚܾطÝݾ × Ò · ¯ ´ µ ÚÖ ´Øµ Ó× · ¯ ´Ü µ Ö ¾ ¿ (4. .13 and ¯¾ Î ¯ Ü µ   ½Ü· ܾ ´· ݾ   ¯½ ¿ ¼. It follows by a direct application of Barb˘ lat’s Lemma (Lemma 2. implies that In order to show that Ý ´Øµ goes to zero as Ø ½ we use Lemma 2. provided that either ÚÖ ´Øµ or Ö ´Øµ does not converge to zero. we obtain that also Ü Ø . Ø We can summarize this result as follows. in turn.30) we can conclude that the trajectories Ü Ø Ý Ø Ø are uniformly bounded.4.1.33) we have that Ý ´Øµ tends to zero as Ø tends to infinity. Ý Ø and Ø are uniformly continuous.33) ½ As a result from (4. If we furthermore assume that ÚÖ Ø and Ö Ø are uniformly continuous.12 with ´Øµ ´Øµ in order to conclude that Ü ¯½ ÚÖ ´ØµÝ × Ò · ¯´ µ ½ · ܾ · Ý ¾ ¾ Ý   ¯ ´Ü µ ¿ Ð Ñ Ö ´ØµÝ ´Øµ ¼ (4.2. we apply Lemma 2.

2 the cascaded approach learned us that for the tracking problem of a mobile robot. 4.37) From Proposition 2.7) in closed loop with the control law (4.37) is globally Ã-exponentially stable.28) and the feasibility condition (4. Once we are able to find controllers for the systems (4.35) For stabilizing (4. As a result we obtain the following. Lefeber and Nijmeijer 1999)).31).7) in closed loop with the control laws (4.14) is given by Ö ´Øµ · ¯ ´ µ ½ (4. Furthermore. ¯¾ .36).15) that meet the constraints (4. A saturated controller for the system (4.2. 4. Assume that Ö Ø is persistently exciting (PE) and that ÚÖ Ø is bounded.14) and (4. Proposition 4. Consider the tracking error dynamics (4.1 (see (Jiang.4.7. For linear systems globally asymptotically stabilizing saturated controllers and several anti-windup controllers are available in literature and can be used. stabilization of the nonlinear tracking error dynamics (4.34. then the closedloop system (4. the same saturated controllers render the tracking error dynamics asymptotically stable too.2 A cascaded design As mentioned in Section 4. the system (4. So also the nonlinear tracking problem under input constraints reduces to two separated linear problems. Consider the tracking error dynamics (4.48 Chapter 4.36) Ü Ý   ¯ ´Ü µ · Ö ´ØµÝ   Ö ´ØµÜ ¾ (4. Then the resulting closed-loop system is globally Ã-exponentially stable.3.15) we can use Ú which results into ÚÖ ´Øµ · ¯¾ ´Ü µ (4.7) in a sense boils down to the separate stabilization of the linear systems (4. and either ÚÖ Ø or Ö Ø does not converge to zero. given the constraints (4.10 we know that if Ö Ø is persistently exciting (PE).28) are satisfied. If ÚÖ Ø and Ö Ø are uniformly continuous and bounded.4. Tracking of a mobile robot Proposition 4.28).14) and (4.29) it is always possible to choose ¯½ . The same holds true for the saturated controller design problem.34) which results in the globally Ã-exponentially stable closed-loop dynamics   ¯´ µ ½ (4.15). ´µ ´µ ´µ . ¯¿ such that the constraints (4.31) is GAS.4. ´µ ´µ ´µ ´µ 4.

We can write the closed-loop system (4.7.37) is  Ü ´Ü µ ¼ Applying Lemma 2. Let Ü . The simple non-holonomic example of a knife-edge moving on the plane was studied by Bloch.4.34. 4.37) satisfying the required properties. 4. Simulations Proof. Ý denote the co-ordinates of the center of mass of the knife-edge on the plane and let denote the heading angle measured from the Ü-axis. Á is the moment of inertia of the knife-edge. there is a non-holonomic constraint of the form Ü × Ò   Ý Ó× ¼ The controls are the pushing force ½ in the direction of the heading angle and the steering torque ¾ about the vertical axis through the center of mass. 4. and the scalar constraint multiplier.35) and (4.4. We only need to find the Lyapunov function candidate of Assumption A1 for the system (4. For that we can take ´ µ Î Î ½ ܾ · ½ ݾ ¾ ¾ ¯¾ whose time-derivative along solutions of (4.25). Reyhanoglu and McClamroch (1992). The d’Alembert’s formulation of the knife-edge dynamics provide Ü Ý Ü ½ × Ò · Ñ Ó× ½   Ñ Ó× · Ñ × Ò Ñ where Ñ is the mass of the knife-edge.4.36) as a system in cascade: 49 Ü Ý   ¯ ´Ü µ · Ö ´ØµÝ · ½ Ý ´ µ · ÚÖ ´Øµ Ó×  ½   ÖßÞØµÜ ´   ½ Ü ´ µ · ÚÖ ´Øµ × Ò ¾ ¯½ ¯½ ½   ¯ßÞ´ µ ¾ ´Ø Þ µ ½ ´Ø Þ ßÞ ½ Þ¾ µ ´Ø Þ µ ¾ ½ Almost all conditions of Lemma 2.37) are globally Ã-exponentially stable and Ø Þ½ Þ¾ satisfies (2.38) Ó× is Ó× · Ý × Ò . When we define the variables ×Ò Ú ¾ Á Ý Ü (4.5 are satisfied.5 Simulations This section is to illustrate that the cascaded approach as presented in this chapter can be easily extended.5 completes the proof. Since the velocity of the center of mass is always perpendicular to the runner.5. since both (4.

40a) (4. Kolmanovsky and McClamroch (1996) solved this problem by defining the change of coordinates ܽ ܾ Ü¿ Ü Ù½ Ù¾  Ü × Ò · Ý Ó× Ü Ó× · Ý × Ò  Ü × Ò · Ý Ó×   ´Ü × Ò   Ý Ó× µ ½ · ¾ ´ Ü Ñ Á ¾ Á × Ò · Ý Ó× µ   ´Ü Ó× · Ý × Ò µ ¾ and the tracking errors   Ü Ü Ü Ù Ù Ù The following hybrid controller was proposed Ù½ ´Øµ Ù¾ ´Øµ  Ü½   ܾ · ͽ ´« ص  Ü¿   Ü · ; ´« ص Ì Ì Ø Ø ´ · ½µÌ ´ · ½µÌ (4. Tracking of a mobile robot Ü Ý Ú Ú Ó× Ú×Ò ½ Ñ ¾ Á (4.39) which is simply the model of a mobile robot and two additional integrators.38) can also be expressed as Chapter 4.50 the dynamics (4.38) follow the reference dynamics Ü Ö ´Øµ Ý Ö ´Øµ Ö ´Øµ Ü Ö ´Øµ Ý Ö ´Øµ × ÒØ   Ó× Ø Ø Ó× Ø  × ÒØ This trajectory corresponds to the center of mass of the knife-edge moving along a circular path of unit radius with uniform angular rate. Kolmanovsky and McClamroch (1996) studied the problem of making the knife-edge (4.40b) .

3. Notice that it takes almost seconds for the knife-edge to converge to the reference trajectory. . ¾¼¼ ½ 2 position error [rad] 0 -2 -4 0 20 40 60 80 100 time [s] 120 140 160 180 Ü   ÜÖ Ý   ÝÖ 200 1 angular error [rad]   Ö 0 -1 0 20 40 60 80 100 time [s] 120 140 160 180 200 5 controls ½ ¾ 0 -5 0 20 40 60 80 100 time [s] 120 140 160 180 200 Figure 4.5. Starting from the initial condition Ü ´¼µ ½ Ý ´¼µ ´¼µ ´¼µ Ý ´¼µ ´¼µ Ì £ ¢ ½ ½ ½ ¼ ¼ ¼ £Ì (4. Simulations where Ì 51 ¾ .3: Tracking error and inputs for cascade controller (4. where we assumed that Ñ Á . and for a scalar parameter « ¿ « × Ò´¾Øµ   « Ó×´¾Øµ · « Ó× Ø ¾ ; ´« ص « × Ò Ø · « ͽ ´« ص is updated via the following scheme: Let ´ Ì µ. Then the parameter « «¼ « ´ « « « ¬ ½ ¬ ­ ¬«  ½ ¬ × Ò´ µ if « if «  ½  ½ ¼ or ¼ and ¼ ¼ where the values ­ ¢ ¼ Ü were proposed.4.40).41) and the resulting performance is depicted in Figure 4.

The following controller was proposed Ü Ó× · Ý × Ò ¾ Á ½ · ¾ ´ Ü × Ò Ñ Á and correspondingly the tracking errors Ü Ú½ Ú¾ where  ¾Ù½   Þ¿ · Þ¾ Ü   Þ½ Þ¿ ܾ · Þ½ Ú¾   ¾Ü  ¾Ù¾   ¾Þ¾   ¾Ü · Þ¿ Ü Ù½ Ù¾ Ü Ü (4.39) and using the change of co-ordinates (4.52 Chapter 4.42a) (4.42b)   ½Þ½Ü · ¾Þ¿ · ¾Þ¾ · Þ½ Ü¿ ܾ ܽ and Þ½ Þ¾ Þ¿   ܾ Þ½ Starting from the same initial condition (4. We can also use the cascaded approach presented in this chapter for solving the tracking problem.4. Tracking of a mobile robot Jiang and Nijmeijer (1999b) solved the same tracking problem by defining the global change of co-ordinates and preliminary feedback ܽ ܾ Ü¿ Ü Ü Ú½ Ú¾ Ü Ü Ó× · Ý × Ò × Ò   Ý Ó× · ´ Ü × Ò · Ý Ó× µ · Ý Ó× µ   ¾ ´Ü Ó× · Ý × Ò µ Ü Ü .2) solving the tracking problem boils down to stabilizing the linear time-invariant ( Ö ) systems: ½ ¾ Ü Ý Ú ¿ ¾ ¼ ½ ½¿ ¾Ü ¿ ¾ ¼ ¿  ½ ¼ ¼ Ý · ¼ ½ ¼ ¼ ¼ Ú Ñ · ¼ ½ Á ½Ö  ½ and ¼ ½ ¼ ¼ ¾Ö  ¾ .41) the resulting performance is depicted in Figure 4. Starting from the model (4.

42).5.4: Tracking error and inputs for backstepping controller (4. . Simulations 53 2 position error [rad] 0 -2 -4 0 1 2 3 4 5 time [s] 6 7 8 9 Ü   ÜÖ Ý   ÝÖ 10 1 angular error [rad]   Ö 0 -1 0 1 2 3 4 5 time [s] 6 7 8 9 10 5 controls ½ ¾ 0 -5 0 1 2 3 4 5 time [s] 6 7 8 9 10 Figure 4.4.

5: Tracking error and inputs for cascade controller (4. Both systems can easily be stabilized by using linear controllers. For tuning these controllers we defined the costs ½ ¾ ½ ¼ ¼ ½ Ü ´Øµ¾ · Ý ´Øµ¾ · Ú ´Øµ¾ · ¼ ¼½´ ½ Ö ´Øµ   ½ ´Øµµ¾ dØ ´Øµ¾ · ¼ ½´ ¾ Ö ´Øµ   ¾ ´Øµµ¾ dØ ¼ Ü   ¼ ¼¼½¾Ý ½Ö· · ½¼ ¼¼¼¼ · ½¼ ¾Ö · ¿ ¾½Ú and used optimal control to minimize these costs.43). controller of Jiang and Nijmeijer (1999b). As a result we obtained as controllers for the tracking control problem ½ ¾ (4.43b) Starting also from the initial condition (4.5. Tracking of a mobile robot where we defined Ú ÚÖ   Ú and Ö   .43a) (4. Notice that the resulting performance is comparable with the backstepping based Ü   ÜÖ Ý   ÝÖ 2 position error [rad] 0 -2 -4 0 1 2 3 4 5 time [s] 6 7 8 9 10 1 angular error [rad]   Ö 0 -1 0 1 2 3 4 5 time [s] 6 7 8 9 10 5 controls ½ ¾ 0 -5 0 1 2 3 4 5 time [s] 6 7 8 9 10 Figure 4.54 Chapter 4.41) the resulting performance is depicted in Figure 4. For comparison reasons we also considered the following quantities ÂÖ ¾¼¼ ¼ Ü ´Øµ   Ü Ö ´Øµ ¾ · Ý ´Øµ   Ý Ö ´Øµ ¾ dØ .

In that case global uniform asymptotic stability ı (GUAS) can be shown. We arrived at the results by means of the cascaded design approach as explained in Section 3. when additional integrators are added. output-feedback.6. For the three different controllers we considered we obtained the following values: hybrid controller (4. This persistence of excitation condition on the reference angular velocity makes that tracking of a straight line and stabilization is not possible with the “cascaded controllers” that have been derived. and under input saturation.341  63.566 4. This was illustrated by means of simulations using the example of a knife-edge.2.43) ÂÖ 103. (1999b). this problem can be overcome by weakening the persistence of excitation (PE) condition by assuming a so-called uniform Æ -persistence of excitation (uÆ -PE). . This is the case for both the state. However. as well as the control problem under input saturation and other interesting problems like incorporating robustness against uncertainties. This approach revealed a nice structure in the tracking error dynamics.42) cascade controller (4.558 3. All results yield globally Ã-exponentially stable closed-loop tracking error dynamics under a persistence of excitation condition on the reference angular velocity.789 8.and output-feedback problem. This simple structure is also maintained when we consider so called dynamic extensions of the model. We solved this problem using state-feedback.40) backstepping controller (4. which makes that the nonlinear tracking problem can be reduced to two linear problems. Concluding remarks and 55  ¾¼¼ ¼ ½ ´Øµ ¾ · ¾ ´Øµ ¾ dØ where ÂÖ can be thought of as an error measure and  as a measure of the control effort.752 3.6 Concluding remarks In this chapter we considered the tracking control problem for the kinematic model of a mobile robot. as recently introduced by Lor´a et al.4.796 4.

Tracking of a mobile robot .56 Chapter 4.

. where Ü Ù½ Ù¾ ܾ Ù½ ÜÒ ½ Ù½ (5. Many mechanical systems with non-holonomic constraints can be locally or globally converted to the chained form under co-ordinate change and preliminary feedback.2) When we define the global change of co-ordinates ܽ ܾ Ü¿ Ü Ó× Ü× Ò 57   Ý Ó× ·Ý× Ò . see (Murray and Sastry 1993). Chained-form systems of order Ò with two inputs can be expressed as ܽ ܾ Ü¿ . namely systems in chained form. Consider the kinematic model of a mobile robot that we studied in the previous chapter: Ü Ý Ú Ó× Ú× Ò (5. .1) ´Ü½ ÜÒ µÌ ÜÒ is the state.Chapter 5 Tracking of non-holonomic systems in chained form 5. and Ù½ and Ù¾ are two inputs.1 Introduction In this chapter we consider the tracking problem for a special class of non-holonomic systems.

¿ ´µ However.. Another global result for a chained-form system of order is given by Escobar. It is well-known that many mechanical systems with non-holonomic constraints can be locally or globally converted to the chained form under co-ordinate change and preliminary feedback. We first apply in Section 5. which implies that no smooth (or even continuous) timeinvariant static state-feedback Ù Ù Ü exists that renders a specified equilibrium of the closed-loop locally asymptotically stable.1) fails to meet Brockett’s necessary condition for smooth feedback stabilization.2). ¿ ¿ In this chapter we solve the global tracking problem for general chained-form systems by means of a cascaded systems based approach.3 the state-feedback tracking problem and in Section 5. where they introduced a field-oriented control approach for the tracking of the non-holonomic integrator. Section 5. Murray and Sastry 1993. We are not aware of any global results for general chained-form systems of order Ò. This results in a similar design separation principle as in the previous chapter for the mobile robot.7 contains some concluding remarks for this chapter. Tracking of non-holonomic systems in chained form and apply the preliminary feedback Ú Ù½ Ù¾ · Ü¿ ܽ ܾ Ü¿ Ù½ Ù¾ ܾ Ù½ where Ù½ and Ù¾ are new inputs.2 to the tracking error dynamics. the knife edge.6 by means of simulations.2 the idea explained in Section 3.g. the system (5.4 the output-feedback tracking problem. In Section 5. Jiang and Nijmeijer (1999b) derived semi-global tracking controllers for general chained-form systems by means of backstepping and they achieved global tracking results for some special cases. the tracking problem for systems in chained form has received little attention. Jiang 1996)). then the system (5. a vertical rolling wheel and a rigid spacecraft with two torque actuators (see e.58 Chapter 5. For this reason the stabilization problem has received a lot of attention (see e. Murray and Walsh 1995. which is a chained-form system of order (and dealt with in the previous chapter). Teel. The effectiveness of the derived controllers is illustrated in Section 5. Sørdalen and Egeland 1995. As for the mobile robot. Berghuis and Nijmeijer 1994. we reduce the problem of designing stabilizing controllers for the nonlinear tracking error dynamics to two linear controller design problems.5 we deal with both these tracking control problems under input saturation for a special class of reference trajectories. Most of the global tracking results we are aware of.. With this knowledge we tackle in Section 5. Interesting examples of such mechanical systems include not only the mobile robot. Kolmanovsky and McClamroch 1995) or Section 3. . Samson 1995. That is.g. (Murray and Sastry 1993. Canudas de Wit. Ortega and Reyhanoglu (1998). (Pomet 1992. are on the tracking control of a mobile robot. but also cars towing several trailers.2) is transformed to which is a chained-form system of order with two inputs.

3) ÜÒ ÜÒ ½ Ù½ Ö · ´ÜÒ ½ The tracking control problem boils down to finding appropriate control laws for Ù½ and Ù¾ such that the tracking error Ü converges to zero. . this subsystem has to be such that the remaining dynamics reduces considerably in case we assume that the stabilization of this subsystem has been established. Therefore. . . . .2 The search for a cascaded structure For studying the tracking control problem for systems in chained form.5.2. but this does not look too promising. ܽ ܾ Ü¿ . We look for a control law for one of the two inputs which is such that a subsystem of (5. we substitute ܽ Ø in the remaining dynamics. we could decide to use Ù¾ for stabilizing the ܾ dynamics.3). respectively. assume we are given a reference trajectory ÜÌ ÙÌ Ì satisfying Ö Ö ´ µ ܽ Ö Ü¾ Ö Ü¿ Ö .2. . Ù¾   Ù¾ Ö ´Øµ ܾ Ù½ Ö ´Øµ ÜÒ ½ Ù½ Ö ´Øµ ÜÒ . For that we like to use the cascaded design approach as proposed in Section 3. The search for a cascaded structure 59 5. Preferably. As the next step is to assume that the stabilization has been established. we assume that the stabilization of ܽ has been established.3) is asymptotically stabilized in closed loop. if we decide to use Ù½ for stabilizing the ܽ dynamics. On the other hand. . Ù½ Ö Ù¾ Ö Ü¾ Ö Ù½ Ö ÜÒ ½ Ö Ù½ Ö ÜÒ Ö We define the tracking error Ü Ü   ÜÖ and obtain as the tracking error dynamics Ù½   Ù½ Ö Ù¾   Ù¾ Ö Ü¾ Ù½ Ö · ´Ü¾ . Ù½   Ù½ Ö Ù¾   Ù¾ Ö Ü¾ Ù½   ܾ Ö Ù½ Ö ÜÒ ½ Ù½   ÜÒ ½ Ö Ù½ Ö · ܾ Ö µ´Ù½   Ù½ Ö µ · ÜÒ ½ Ö µ´Ù½   Ù½ Ö µ (5. Next. . that is. After this substitution the remaining dynamics becomes ´µ ¼ ´µ ´µ ¼ ܾ Ü¿ . the assumption that this stabilization has worked out simplifies almost all equations in (5. Notice that either the ܽ dynamics or the ܾ dynamics can be easily rendered asymptotically stable by choosing an appropriate control law for Ù½ or Ù¾ . Notice that as a result also Ù½ Ø   Ù½ Ö Ø . we decide to first use Ù½ for stabilizing the ܽ dynamics.

5) and finding one for the linear time-varying system ܾ Ü¿ ÜÒ . . . ¿ Ù½ Ö ´Øµ ¼ ¼ ¼ . ¼ .6) ¼ When we do so. ¼¿ ¾ ܾ ¼ Ü¿ . .. .3. · ½ ¼ .7) since then the resulting closed-loop system   ½ ܽ .. ¿ ¾ ¿ ¼ Ù½ Ö ´Øµ . . ¼ . ¿ ¾ Ù½ Ö ´Øµ ¼ ¼ ¼ .5) and (5. . . the system (5. . ¦ ¦ 5..3 State-feedback In this section we study the state-feedback tracking control problem for chained-form systems.. Tracking of non-holonomic systems in chained form which is the same as the linear time-varying system ¾ ܾ Ü¿ ÜÒ .4) is uniformly completely controllable. . we have to solve two linear ones. . . . Clearly..3) we might as well look at the two separate problems of finding a stabilizing control law for the linear system ܽ ¾ Ù½   Ù½ Ö ´Øµ ¾ (5. .60 Chapter 5.5) the role of the system ¾ .6).5) can easily be stabilized. the stabilized system (5. . Then we can use Theorem 2. .. . . . . . As derived in the previous section from a cascaded systems point of views the problem of stabilizing (5.3) reduces to stabilizing the linear systems (5.3 to conclude asymptotic stability of the entire nonlinear tracking error dynamics. Notice from Corollary 2. . So instead of solving a nonlinear control problem. . . . A possible control law for Ù½ is Ù½ ܽ Ù½ Ö ´Øµ   ½ ܽ ½ ½ ¼ ¼ (5. . . .6) plays the role of the system ½ in Section 2. . .4 and (5. . . ¼¿ ¾ ܾ ¼ Ü¿ . ..4. . ÜÒ .4) and hope that all conditions for applying the cascaded theorem (Theorem 2. . . .. . . . . . . . Ù¾   Ù¾ Ö ´Øµ (5. .3) are met. ..4) ¼ All we need to do is to find a feedback controller for Ù¾ that stabilizes the system (5. . .4 that if Ù½ Ö Ø is persistently exciting the system (5. . · ½ ¼ . Ù¾   Ù¾ Ö ´Øµ (5. ÜÒ . ´µ To summarize: instead of solving the problem of finding stabilizing control laws for the nonlinear tracking error dynamics (5. ¼ . . . . . ¼ . . ¿ ¾ ¿ ¼ Ù½ Ö ´Øµ .4. . . .

6 completes the proof.. State-feedback is globally uniformly exponentially stable (GUES). under the assumption that Ù½ Ö Ø is persistently exciting (which yields uniform complete controllability according to Corollary 2. ¼ ¼ ¼ . provided that Ù½ Ö ently exciting (PE)..3. 5. 5. ½ ܽ   ßÞ ½ ½Ü ¾ ¾ ¼ ßÞ Ù Ö ´Øµ ¼ ´Ø Þ µ ½ ½ Ü¿ . However. Therefore. . Due to the design.4).3).6) we can use the result of Theorem 2. 61 ´Øµ is persist- We can combine both results and solve the global state-feedback tracking control problem. ¼ ¿ ܽ ´Ø Þ ßÞ ½ Þ¾ µ ´Ø Þ µ From Theorem 2.7 we know that the system Þ½ ½ Ø Þ½ is globally uniformly exponentially stable (GUES). Also the system Þ¾ ÜÒ ½ Ö are ¾ Ø Þ¾ is GUES.4. ´ µ ´ µ ´ µ Remark 5. . . . . . . ÜÒ ¾ Ù½ Ö ´Øµ   ¾   ¿ Ù½ Ö ´Øµ     Ù½ Ö ´Øµ . . more control laws for Ù½ are available in literature that also guarantee GUES. is the fact that it is globally uniformly exponentially stable (GUES). ¿ ¾ ܾ ¿ .1. ...7. several other choices for Ù¾ can be made. As a result.6) (see (Lefeber. the proof still holds. Proposition 5. .5. Assume that Ù½ Ö Ø is persistently exciting (PE) and that ܾ Ö Ü¿ Ö ÜÒ ½ Ö are bounded.3.3. one might consider ´ µ ´µ ¯ a ‘standard’ linear control law (Rugh 1996) which involves using the state-transition matrix of the system (5. . Consider the chained-form tracking error dynamics (5. .3.8) globally Ã-exponentially stable.25). For instance. . . .3. Robertsson and Nijmeijer 1999a)). provided that ½ Ù½ Ù¾ ( Ù½ Ö ´Øµ   ½ ܽ Ù¾ Ö ´Øµ   ¾ ܾ (5.8): ¾ ܾ ¿ Ü¿ . Notice that the only property of the system Þ½ ½ Ø Þ½ that we need in this proof. . . This is something that (according to Theorem 2. . .. . . . For stabilizing (5. . Corollary 2. ´µ Then the control law   ¿ Ù½ Ö ´ØµÜ¿   Ü   Ù½ Ö ´ØµÜ   ¡ ¡ ¡ (5.3.8a) ¼ and ¾ Ò) are such that the polynomial Ò ½ · ¾ Ò ¾ · ¡ ¡ ¡ · Ò ½ · Ò is Hurwitz. Since ܾ Ö bounded we have that Ø Þ½ Þ¾ satisfies (2. . we can recognize a cascaded structure in the closed-loop system (5.3. ÜÒ ¾ ·   ½ ´Ü¾ ·Ü¾ Ö µ   ½ ´Ü¿ ·Ü¿ Ö µ   ½ ´ÜÒ ½ ·ÜÒ ½ Ö µ . .3. In case we replace Ù¾ with any of these. .3.7) is guaranteed by the choice for the input Ù¾ .8b) renders the closed-loop system (5. Proof.2. ¼ . . .

1) we are unable to reconstruct the variables ܽ and ÜÒ ..3. yielding convergence in terms of this distance. a pole-placement based control law. When ´¼ µ Ø . ´ ص is a quite complicated expression that satisfies ´¼ ص ¼. 5. In practice this has the advantage that the damping rate does not change with different values of Ù½ Ö Ø . this makes clear that one should avoid changing co-ordinates in order to be able to apply a standard control design technique.3. In case we translate the result of this chapter using the original error co-ordinates. we obtain Ú where Ö ÚÖ · · ½ ¾ Ü   ´ ¿   ½µÝ · ´ ص ½ ¼ ¾ ¿ ¼ we do so. (Astolfi 1995)). In case we interpret ´¼ ص as a part of the connecting term “ ´Ø Þ½ Þ¾µÞ¾” we can decide to forget about it. like for instance the one presented by Val´ sek and aˇ dÙ dÒ ¾ Ù½ Ö Olgac (1995) (which requires Ù½ Ö ¾ Ò ¾ and the signals d½ Ö Ø Ø to Ø dØÒ ¾ be available). it is possible to normalize the system’s equations in terms of the advancement velocity Ù½ Ö Ø . Furthermore. Furthermore.3. since it can lead to unnecessary complicated controllers. in order to replace time by the distance gone by the reference vehicle. the work of the previous chapter can not be considered redundant.13). The mobile robot we studied in the previous chapter is also a chained-form system. It would be interesting to compare the results of the previous chapter with the result derived here. the system (5. we study the problem of stabilizing the tracking error dynamics (5..6).4. in the variables Ü and Ý as defined by (4. ´µ ´µ Remark 5.e. Therefore. That is. One of the advantages of the cascaded control design approach is that all analysis can be done in the original co-ordinates. Remark 5. This “time normalization” makes the solutions “geometrically” unaffected by velocity changes. however.1) is locally observable at any Ü ¾ ÊÒ (see e. Robertsson and Nijmeijer 1999b)). Tracking of non-holonomic systems in chained form ¯ ¯ a less complicated control law (which also needs the state-transition matrix of the system (5.62 Chapter 5. can only be drawn The conclusion that we are allowed to leave out the term with the cascaded structure of the mobile robot in mind. . in case we are able to measure only ܽ and ÜÒ . no matter what output we have. As pointed out by Samson (1995).6)) as presented by Chen (1997) (see (Lefeber. we regain exactly the control law (4. i.g.3) where we are only allowed to use the measured output for designing the control laws for Ù½ and Ù¾ . It would be most reasonable to compare both results in the original error coordinates.2). ´µ ´µ or any other control law one prefers that guarantees GUES of the system (5. instead of time.4 Dynamic output-feedback In this section we study the dynamic output-feedback tracking control problem for chainedform systems. Notice that for the system (5.

. ¼¿ ¾Ü¾ ¼ Ü¿ . .. . . . . .. Uniform complete observability follows from duality.3.9) we can use Ù½ Ù½ Ö ´Øµ   ½ ܽ ½ ¼ ¼ (5.9b) ܾ Ü¿ .4 as mentioned in the previous section.5. ´µ By combining both results we obtain a solution for the dynamic output-feedback tracking control problem. . . . For this it suffices that Ù½ Ö Ø is persistently exciting.. Uniform complete controllability follows from Corollary 2. . . since in a sense it represents the least amount of components of the state vector that is required for being able to control the system (5. ÜÒ Ü¾ Ù½ Ö ´Øµ ¿ . . For stabilizing (5. ¼ . . ¿ ¾ Ù½ Ö ´Øµ ¼ ¼ ¼ .2 it is clear that the problem can be reduced to the problem of finding stabilizing dynamic output-feedback laws for the systems ܽ ݽ and ¾ Ù½   Ù½ Ö ´Øµ ܽ (5.4. Dynamic output-feedback 63 This is why we assume to have ܽ ÜÒ Ì available as an output for the chained-form system (see (Astolfi 1995.11) since then the resulting closed-loop system becomes ܽ   ½ ܽ ½ which is globally uniformly exponentially stable (GUES).3). .10b) For stabilizing (5.10a) (5.12a) (5. . .. .10) we can use Theorem 2.1). . ¼ . .12b) . ÜÒ . ¿ ¾ ¿ ÜÒ Ý¾ ¢ ¼ ¾ £ ¼ ¼ ½ . Jiang and Nijmeijer 1999a)). · ½ ¼ ¼ . . .10) by means of output-feedback we need both uniform complete controllability and uniform complete observability of the system (5.. . Proposition 5.9a) (5. . . ´µ Then the control law Ù½ Ù¾ Ù½ Ö ´Øµ   ½ ܽ Ù¾ Ö ´Øµ   ¾ ܾ   ¿ Ù½ Ö ´ØµÜ¿   Ü   Ù½ Ö ´ØµÜ   ¡ ¡ ¡ (5. . Consider the chained-form tracking error dynamics (5. In light of Section 5. .3.8.1. . .10). For stabilizing (5. Assume that Ù½ Ö Ø is persistently exciting (PE) and that ܾ Ö Ü¿ Ö ÜÒ ½ Ö are bounded. Ù¾   Ù¾ Ö ´Øµ (5. .4. .

.6 completes the proof. . . . .64 Chapter 5.. ¾ ¿ ¿ ¾ ¿ ¼ . . . . . ¾ .25)... . . . .. . . .4. ¾ ¿ Ù½ Ö ´Øµ Ù½ Ö ´Øµ ¼ ¼ .. From Theorem 2.8 we know that the system Þ½ ½ ´Ø Þ½ µ is globally uniformly exponentially stable (GUES).12. . ÜÒ ¿ ¾ Ù½ Ö ´Øµ ¼ ¼ . . . . . . .... . ¼ . . . . Since ܾ Ö ÜÒ ½ Ö are bounded ´Ø Þ½ Þ¾ µ satisfies (2. .12. .3. . . . ¼ . Ù½ Ö ´Øµ ¼ . ¼  Ð Ù½ Ö ´Øµ . .. . . . . . . . . . .  Ð .3. . . Ù½ Ö ´Øµ ¼ ¾´Ø Þ¾ µ ¼   ½ Þ¾ ¾ ¼ ¼ . .. . ¼ ¼ ¼ . . Application of   ½ ´ÜÒ ½ ·ÜÒ ½ Ö µ Ü  Ü ( ¾ Òµ. ÜÒ .. . a system of the form (3. 5.13) globally Ã-exponentially stable. ¿ Ù¾  Ù¾ Ö ´Øµ · Ð Ù½ Ö ´Øµ ÜÒ Ð Ð¿ Ù½ Ö ´Øµ о  ÜÒ (5. . . .. . . . . . provided and Ð ( Ò) are such that the polynomials that ½ Ò ½ ¾ Ò ¾ ¡ ¡ ¡ Ò ½ Ò ¼ ¾ · · · · and Ò ½ are Hurwitz. . .13) renders the closed-loop system (5. . . · о Ò ¾ · ¡ ¡ ¡ · ÐÒ ½ · ÐÒ Proof.. . . Þ½   ½ ´Ü¾ ·Ü¾ Ö µ   ½ ´ÜÒ ½ ·ÜÒ ½ Ö µ ¼   ½ ´Ü¾ ·Ü¾ Ö µ ´Ø Þ½ Þ¾µ and Ü Corollary 2. . . Tracking of non-holonomic systems in chained form where Ü is generated from the observer ¾ ܾ Ü¿ . . . . . Also the system Þ¾ ¾´Ø Þ¾ µ is GUES. ¼ ¼ ¼ . . . 5.13) as a cascaded system. . . ¼  Ð¿ Ù½ Ö ´Øµ ¼ Ù½ Ö ´Øµ  Ð¾ ¼ . . · ½ ¼ ¼ . .. .e. We can see the closed-loop system (5. . . 5. . . ¼ ¼ ½´Ø Þ½ µ ¼ ¼ .1) where ¢ £ ÜÒ Ü¾ ÜÒ Ì Þ½ ܾ Þ¾ ܽ ¾ Ù½ Ö ´Øµ   ¾   ¿ Ù½ Ö ´Øµ     Ù½ Ö ´Øµ . . . . 5. . . ¿ . .3.. . . . . . ¿ .. . . . . i. . ¼ ¼¿¾Ü ¼ Ü . Ù½ Ö ´Øµ .

15) where ¯ is a saturation function as defined in Definition 2.5) while meeting (5. Saturated control 65 5. As far as we know. no result on stabilizing linear time-varying systems by means of saturated state or output-feedback is known (yet).3) in closed loop with this control law for Ù¾ and (5.6) under the input constraints (5. Therefore.7) and (5.5. Therefore. for the reference trajectory the condition (5. Lemma 2.13 and Then the resulting closed-loop ܽ dynamics becomes which is globally Ã-exponentially stable. is to find a control law for Ù¾ that in closed loop with (5.5) and (5. we assume that the reference trajectory that we would like to track satisfies ×ÙÔ Ù½ Ö ´Øµ Ø ¼ ×ÙÔ Ù¾ Ö ´Øµ Ø Ùmax ½ Ùmax ¾ ¼ Instead of the nonlinear tracking control problem under input constraints.4. we only have to address the two linear problems of stabilizing the subsystems (5.3 is satisfied. that is.5 Saturated control In a similar way as in the previous sections. while meeting (5. we can study the problem of designing tracking controllers in such a way that we are always guaranteed to meet the constraints Ù½ ´Øµ Ù¾ ´Øµ Ùmax ½ Ùmax ¾ (5.14a) (5. Then Corollary 2. In addition.14b) Obviously. We can modify the control laws (5. if Ù¾ guarantees local uniform exponential stability (LUES) we can conclude global Ã-exponential stability of the closed-loop tracking error dynamics (cf.4. which deals with the stabilization under input constraints.14a) is not difficult.6) which is such that for the resulting closed-loop system Assumption A1 of Theorem 2. both by using state-feedback and . As can easily be seen.4.15).14b). For linear time-invariant systems we have the results of Sussmann. the only problem that remains. ܽ   ¯ ´Ü½ µ Assume that we have a control law for Ù¾ for the system (5.1.14) is met. That is.5). we need to assume that once we are on the reference trajectory we can stay on the trajectory.4 tells us that we have global uniform asymptotic stability of the tracking error dynamics (5.6) results in a globally uniformly asymptotically stable (GUAS) system (most preferable globally Ã-exponentially stable). stabilizing (5. we can study the tracking problem for a system in chained form under input saturation. Sontag and Yang (1994).5.11) into Ù½ ¯ Ù½ Ö ´Øµ   ¯ ´Ü½ Ùmax   ×ÙÔ Ù½ Ö ´Øµ ½ Ø ¼ µ (5.14).

we also have a solution to the general tracking problem for chained-form systems. the steering angle . see e.1 The rear wheels of the car and the trailer are aligned ´ µ Ý Ð Ý ½ ¼ ½ Ü Ü Figure 5. namely those reference trajectories for which Ù½ Ö Ø is a non-zero constant. as soon as we have a result on the stabilization of linear time-varying systems under input saturation. like for example the one proposed by Kapoor.66 Chapter 5. Teel and Daoutidis (1998). Tracking of non-holonomic systems in chained form output-feedback. see Figure 5. For linear time-invariant systems we can also think of using anti-windup controllers. and the steering velocity . ´µ 5. The two input signals are the driving velocity of the front wheels. The state configuration of the articulated vehicle consists of the position of the car Ü Ý .g. where the controls are the forward velocity Ú and the steering velocity of the tow car.1: Kinematic model of a car with a single trailer.6 Simulations In this section we apply the proposed state-feedback design for the tracking control of a well-known benchmark problem: a towing car with a single trailer. and the orientations ¼ respectively ½ of the car and the trailer with respect to the Ü-axis.and output-feedback) can (yet) only be solved for a special class of reference trajectories. . Samson 1995. Ú . (Murray and Sastry 1993. As a result. Jiang and Nijmeijer 1999b). the tracking problem under input constraints (5. However..14) for chained-form systems (under both state. with the chassis and are not allowed to slip sideways.

17) Ó× Ù½ ¼ ¼ ½ µÙ½ · ¬¾ ´ ¼ ½ µÙ¾ ¬½ Ð ¾ ¼ ¾   ¿ ×ÐÒ Ó×¾× Ò · × Ò Ó× Ó×   ¿ × Ò Ó× Ó×   ¾ Ó× × Ò´ ¼   ½ µ ¼ ½ ½ ½ ¼ ½ Ó× ½ ¾ ¼ Ó×¾ ¾ ¼ Ó×¾   Ð Ó× Ó׿ × Ò´ ¼   ½µ · ½¾Ð Ó× Ó× × Ò´ ¼   ½µ ¾ ¾ ¬¾ ½ Ð · ¾×Ò ¼ ½ Ð ½ Ó׿ ¼ ½ Ó×¾   Ð Ó× ½ Ó×¾ Ó×¾ ¼ × Ò ¼ Ó×¾ ¾ Ó×¾ ½ ½ ܽ ܾ Ü¿ Ü Ü Ù½ Ù¾ ܾ Ù½ Ü¿ Ù½ Ü Ù½ ½ ½ This change of co-ordinates and preliminary feedback.16) is not in chained form. Simulations The kinematic equations of motion for the vehicle can be described by 67 Ü Ý Ú Ó× ¼ Ú× Ò ¼ to the rear wheels of the car.5.1) via a change of co-ordinates and preliminary state-feedback (Murray and Sastry 1993): ܽ ܾ Ü¿ Ü Ü Ú where Ü ØÒ × Ò ½ × Ò¾´ ¼   ½µ ·   ¾× Ò´ ¼  Ó×½ µ ¾ Ó×¾ ¼ Ó׿ ½ Ð ½ Ó׿ ¼ Ó× ½ ½ ½ Ó× ¼ ¿ ½ × Ò´ ¼   ½ µ ½ Ó× ¼ Ó×¾ ½ ØÒ½ ½·×Ò ½ Ý   ½ ÐÓ Ó× ½ ¬½ ´ (5. results into the chained-form system .16) Ú ØÒ Ð Ú × Ò´ ¼   ½µ ½ ½ where Ð is the wheelbase of the tow car and ½ is the distance from the wheels of the trailer ¼ Although the system (5. it can locally be transformed into the chained form (5. (5.6.

18) In chained-form co-ordinates the reference (5. We used pressed as a standard linear time-invariant system of the form Ü optimal control to arrive at the control law Ù  ÃÜ for which the costs · ½ ¼ ܴص ¾ · Ö Ù´Øµ ¾ dØ . ¾ behaved as shown in Figure 5.8).18) can be expressed as ¢ ܽ Ö Ü¾ Ö Ü¿ Ö Ü Ö Ü Ö Ù½ Ö Ù¾ Ö Ü £ ¼ ¼ ¼ ¼ ½ ¼£ ¼ £Ì Starting from the initial condition ´¼µ ¢ ½ ¼ ¼ ¼ it was shown by Jiang and Nijmeijer (1999b) that their backstepping based controller Ù½ Ö ·   ܽ   Ù¾ Ö Ü¾ Ö Ü½   Ù¾ Ö Ü Ö Ü½ · Ù¾ Ö Ü¿ · Ù¾ Ö Ü   ¾ ܾ Ö Ü¿ Ö Ü¾   ¾ Ü¿ Ö Ü Ö Ü¾   ¿Ù½ Ö Ü¾ Ö Ü¾   Ù½ Ö Ü¾ Ö Ü Ö Ü¾ ½ ½ ¾ ½ ½   Ù½ Ö Ü¾ Ö Ü¾   Ù¾ Ö Ü½ ܾ · ܾ Ö Ü½ ܾ · Ü Ö Ü½ ܾ ½ · ¾ Ü¿ Ö Ü½ Ü¿ · Ù½ Ö Ü¾ Ö Ü½ Ü¿ · Ù½ Ö Ü Ö Ü½ Ü¿   Ù¾ Ö Ü½ Ü · ¾ ܾ Ö Ü½ Ü · ¾ Ü Ö Ü½ Ü · ¾ Ü¿ Ö Ü½ Ü · Ù½ Ö Ü¾ Ö Ü½ Ü · ¾Ù½ Ö Ü Ö Ü½ Ü   ܾ Ü¿   ܾ Ü   ¿Ù½ Öܾ   ¾ Ü¿ Ü ¿   Ù½ Ö Ü¿ Ü   ¾ Ü Ü   Ù½ Ö Ü¾   ¾ Ü¿ Ö Ü¾ ܾ ½   Ù½ Ö Ü¾ Ö Ü¾ ܾ   Ù½ Ö Ü Ö Ü¾ ܾ   ¾ ܾ Ö Ü¾ Ü¿ ½ ½ ½   ¾ Ü Ö Ü¾ Ü¿   ¾ Ü¿ Ö Ü½ ¾ Ü   Ù½ Ö Ü¾ Ö Ü¾ Ü ½ ½   ¾Ù½ Ö Ü Ö Ü¾ Ü · ܽ ܾ · Ù½ Ö Ü½ ܾ Ü¿ · ¿ ܽ ܾ Ü ½ ¾ · Ù½ Ö Ü½ ܾ Ü · ¾ ܽ ܾ · Ù½ Öܽ Ü¿ Ü · ¾ ܽ Ü¿ Ü ¿ ¾ · ¾Ù½ Ö Ü½ Ü Ü   ¿Ù½ Ö Ü¾ ܾ   ¾ ܾ ܾ Ü¿ · ¾ ܽ Ü ½ ¾ ½ ¾ ܾ Ü   ¾ ܾ Ü¿ Ü   Ù½ Ö Ü¾ ܾ ¡ Æ     Ù½ Ö Ü½ ½ ½ · ܾ Ö Ü¿ Ö Ü½ · ¾Ü¿ Ö Ü Ö Ü½   ¾Ü¾ Ö Ü¾   Ü¿ Ö Ü¿   ܾ Ö Ü   ¾Ü¿ Ö Ü · Ü¿ Ö Ü½ ܾ · ܾ Ö Ü½ Ü¿ · ¾Ü Ö Ü½ Ü¿ · ¾Ü¿ Ö Ü½ Ü ¡  ¾Ü¾   ܾ Ü   ܾ   ¾Ü¿ Ü · ܽ ܾ Ü¿ ¾ ¿ · ¾Ü½ Ü¿ Ü Ù¾ Ù¾ Ö · ¾ Ü¿ Ö Ü½ · ¿Ù½ Ö Ü¾ Ö Ü½ · Ù½ Ö Ü Ö Ü½   ܾ   ¿Ù½ Ö Ü¿   ¾ Ü   Ù½ Ö Ü · ¿Ù½ Ö Ü½ ܾ · ¾ ܽ Ü¿ · Ù½ Ö Ü½ Ü Ù½ where .68 Chapter 5. For tuning the gains we considered the two linear subsystems that result from the cascaded analysis. Both can be exÜ Ù. This behavior was compared to that of the control law (5. Tracking of non-holonomic systems in chained form (Jiang and Nijmeijer 1999b) studied for this system the problem of tracking the straight line ¢ Ü Ö Ý Ö Ö ¼Ö ½ Ö ÚÖ Ö £ ¢ Ø ¢ ¼ ¼ ¼ ¼ ½ ¼£ Ø (5.   (5.19a) (5.19b) ¾.2.

5 0 5 10 15 20 time [s] 25 30 35 40 2 controls Ù½ Ù¾ 0 -2 0 5 10 15 20 time [s] 25 30 35 40 Figure 5. .5 angular error [rad] 5 10 15 20 time [s] 25 30 35 40   Ö 0 -0.5.2: Tracking errors and inputs for backstepping controller (5.6.19). Simulations 69 2 position error [rad] Ü   ÜÖ Ý   ÝÖ 1 0 -1 0 0.

5 angular error [rad] 5 10 15 20 time [s] 25 30 35 40   Ö 0 -0. 5.5 0 5 10 15 20 time [s] 25 30 35 40 2 controls Ù½ Ù¾ 0 -2 0 5 10 15 20 time [s] 25 30 35 40 Figure 5.20a) (5.70 Chapter 5.19) are only global respectively semi-global in the chained-form co-ordinates. The control law we used was Ù½ Ù¾ Ù½ Ö ´Øµ   ¿ ½ Ù¾ Ö ´Øµ   ¾ ¼ ¾¿Ü½ ¼Ü¾   ¾ ½¼ ¼Ü¿   ½ ½ Ü   ¼ ¿½ ¾Ü (5.20) and (5.17) the control laws (5.20) not only looks more appealing than (5.19) but also obtains better convergence and is much easier to tune. Tracking of non-holonomic systems in chained form are minimized. but local in the original co-ordinates.7 Concluding remarks In this chapter we studied the tracking problem for a special class of non-holonomic systems. This class contains several interesting examples of mechanical systems with non-holonomic constraints. 2 position error [rad] Ü   ÜÖ Ý   ÝÖ 1 0 -1 0 0.3: Tracking errors and inputs for cascade controller (5. where we took the constant Ö such that similar control effort is needed as by Jiang and Nijmeijer (1999b).20).3. .20b) The resulting behavior is presented in Figure 5. Notice that due to the local nature of the change of co-ordinates and preliminary feedback (5. We see from this simulation that the cascaded control law (5. namely the class of chained-form systems.

(1999b) used the weaker notion of ı uÆ -PE to study the stabilization problem for chained-form systems of order . Fortunately. Since for the mobile robot this enabled us to weaken the assumption on the reference trajectory.6) in particular) would extend the class of reference trajectories we can track under input saturation. provided that Ù½ Ö Ø is persistently exciting.5.and output-feedback) for only a special class of reference trajectories. Theorem 2. Next. a uniform global stabilization result for linear time-varying systems under input saturation (and for the system (5. It is worth investigating whether weakening the PE condition on Ù½ Ö to a uÆ PE condition can be done for general chained-form systems of order Ò too. Using the boils down to substituting ܽ remaining input. For the mobile robot the cascaded design was an eye-opener to recognize a simpler structure for backstepping. we can start from the nonlinear tracking error dynamics (5. Then assume that this stabilization has worked out (which and Ù½ Ù½ Ö Ø in the remaining dynamics).7. That is. namely those with constant Ù½ Ö Ø . this system can be stabilized.3) with the input Ù¾ as in Proposition 5. Then the cascaded approach used in this chapter can easily be applied to the system in original coordinates. All results assume a persistence of excitation condition on Ù½ Ö Ø . first stabilize the tracking error dynamics that corresponds to ܽ using the input that corresponds to Ù½ .and output-feedback tracking problems globally for this class of systems. On the contrary: a cascaded approach to the model of the mobile robot in the original coordinates leads to less complicated expressions for the controller. it is worth trying the same for chained-form systems. the change of coordinates that brings the system in chained form is not global in general. This is also what should be kept in mind when designing controllers for general systems that can be transformed into chained form: apply the approach in the original co-ordinates! Besides.1 and “step back” the virtual control to obtain an expression and the virtual control ܽ for the input Ù½ . Concluding remarks 71 Using a cascaded systems approach we solved both the state. ´µ ´µ We like to emphasize that the fact that the kinematic model of a mobile car is contained in the class of chained-form systems does not make the results in the previous chapter redundant. ¼ . ´µ ¿ Furthermore.4. We also solved the saturated tracking problem (both under state. ¼ ´µ ´µ As mentioned in the previous chapter a persistence of excitation (PE) condition on the reference input Ù½ Ö Ø might not be required.3. Lor´a et al.3 guarantees global uniform asymptotic stability of the cascaded system under some additional boundedness assumptions on the reference trajectory. very often ܽ and Ù½ are natural co-ordinates of the system (as for the mobile robot). That is.

72 Chapter 5. Tracking of non-holonomic systems in chained form .

1 Introduction In this chapter we study the tracking problem for an under-actuated ship. sway (sideways) and yaw (heading). moment available for steering the ship.1: Definition of state variables in surge. As a result. this assumption is not valid for all ships. but are without any bow or side thrusters. heave. roll. see Figure 6. It is very well possible that ships are either equipped with two independent aft thrusters or with one main aft thruster and a rudder. However. pitch and yaw for a marine vessel. For a conventional ship it is common to consider the motion in surge (forward). Often. we have surge and sway control forces and yaw control Ý Ú (sway) Õ (pitch) Ô (roll) Ù (surge) Ü Ö (yaw) Û (heave) Þ Figure 6. sway. we 73 . we have no sway control force.Chapter 6 Tracking control of an under-actuated ship 6. like for instance many supply vessels. Therefore.1.

Ý . Tracking control of an under-actuated ship assume to have only surge control force and yaw control moment available. and a controller providing global exponential stability (GES) of the desired trajectory was developed. Furthermore we assume that the inertia. inspired by the results of Jiang and Nijmeijer (1999b). typically the ship may turn around and the desired position trajectory is followed backwards. The experimental results are presented in Section 6. (Godhavn 1996) investigated output-tracking control based on a nonlinear model of the ship. The controller derived in Section 6. and wave drift forces. currents. denote the position and orientation of the ship in the earth-fixed frame. giving local results. Some conclusions are drawn in Section 6. sway and yaw respectively and Ü. Reyhanoglu 1996. (Fossen 1994)): Ù Ú Ö Ü Ý Ñ¾¾ ½ Ù½ ÚÖ   ½½ Ù · ѽ½ ѽ½ ѽ½ ѽ½ ¾¾ Ú   Ñ ÙÖ   Ñ ¾¾ ¾¾ ѽ½   Ѿ¾ ½ Ù¾ ÙÚ   ¿¿ Ö · Ñ¿¿ Ñ¿¿ Ñ¿¿ Ù Ó×   Ú × Ò Ù × Ò · Ú Ó× Ö (6.3 is implemented for tracking control of a scale model of an offshore supply vessel.2 we derive the tracking error dynamics considered in (Pettersen and Nijmeijer 2000) and also more natural error dynamics.1c) (6. For modeling the ship we follow Fossen (1994). Pettersen and Nijmeijer 1998). In Section 6. The parameters Ñ are given by the ship inertia and added mass effects. In that case the ship dynamics can be described by (see e. added mass and damping matrices are diagonal.5. Pettersen and Nijmeijer (2000) achieved semi-global asymptotic stability by means of backstepping. We are not aware of any global tracking results for the tracking control of an under-actuated ship in literature.1e) (6. The parameters are given by the hydrodynamic damping.74 Chapter 6. Pettersen and Egeland 1996.1) is neither static feedback linearizable. . The available controls are the surge force Ù½ . and steering only two degrees of freedom. That is why we focus on state-tracking instead of output-tracking. Tracking control of ships has mainly been based on linear models.1d) (6. The first complete state-tracking controller based on a nonlinear model was developed by Pettersen and Nijmeijer (1998) and yields global practical stability. It was shown by Pettersen and Egeland (1996) that no continuous or discontinuous static state-feedback law exists which makes the origin asymptotically stable. and the yaw moment Ù¾ .4. The ship model (6.1b) (6. Ú and Ö are the velocities in surge.3 we solve the statefeedback tracking control problem. In Section 6. ¼ ¼ The stabilization problem for an under-actuated ship has been studied in (Wichlund. roll and pitch and a slowly-varying bias term due to wind.g.1a) (6.1f) where Ù.. In this chapter we present a global solution to the tracking problem for an under-actuated ship. Since we need to control three degrees of freedom and have only two inputs available we are dealing with an under-actuated problem. Sørdalen and Egeland 1995. As only the position variables are controlled. We neglect the dynamics associated with the motion in heave. nor can it be transformed into chained form.

4b) (6.5a) (6.e.4e) (6.5e) (6.5f) ..4d) (6. i. Next.4c) (6.4f) In this way.4a) (6. This problem is solved by defining the change of co-ordinates as proposed by Pettersen and Egeland (1996) which boils down to considering the dynamics in a frame with an earth-fixed origin having the Ü.2) ÙÖ ÚÖ   ¿¿ ÖÖ · ÖÖ Ñ¿¿ Ñ¿¿ Ñ¿¿ ÜÖ ÙÖ Ó× Ö   ÚÖ × Ò Ö ÝÖ ÙÖ × Ò Ö · ÚÖ Ó× Ö Ö ÖÖ Notice that a drawback exists in considering the error co-ordinates Ü   ÜÖ and Ý   ÝÖ .5d) (6. The tracking error dynamics 75 6.5c) (6.5b) (6.2 The tracking error dynamics Assume that a feasible reference trajectory trajectory satisfying ´ÙÖ ÚÖ ÖÖ ÜÖ ÝÖ Ö Ù½ Ö Ù¾ Ö µÌ is given.2. a these position errors depend on the choice of the inertial frame. since ÙÖ Þ½ Þ¾ Þ¿ The reference variables Þ½ Ö .and sway-axis: Ѿ¾ ½ Ù½ Ö ÚÖ ÖÖ   ½½ ÙÖ · ѽ½ ѽ½ ѽ½ ѽ½ ¾¾ Ú ÙÖ   ÚÖ   Ѿ¾ Ö Ö Ñ¾¾ Ö Ñ½½   Ѿ¾ ½ Ù¾ Ö (6.6. tracking errors Ó× · Ý × Ò  Ü × Ò · Ý Ó× Ü (6.3) Þ¾ Ö and Þ¿ Ö are defined correspondingly.and Ý-axis always oriented along the ship surge. we obtain the tracking error dynamics Ù Ú Ö Þ½ Þ¾ Þ¿ Ѿ¾ ´Ú Ö · Ú ÖÖ ´Øµ · ÚÖ Ö µ   ѽ½ Ù · ѽ½½ ´Ù½   Ù½ Ö µ ѽ½ ½½ ѽ½   Ñ ´Ù Ö · Ù ÖÖ ´Øµ · ÙÖ Ö µ   Ѿ¾ Ú ¾¾ ¾¾ ѽ½   Ѿ¾ ´Ù Ú · Ù ÚÖ · ÙÖ Ú µ   Ñ¿¿ Ö · ѽ ´Ù¾   Ù¾ Ö µ Ñ¿¿ ¿¿ ¿¿ Ù · Þ¾ Ö · Þ¾ ÖÖ ´Øµ · Þ¾ Ö Ö Ú   Þ½ Ö   Þ½ ÖÖ ´Øµ   Þ½ Ö Ö Ö (6. we define the Ù Ú Ö Þ½ Þ¾ Þ¿ Ù   ÙÖ Ú   ÚÖ Ö   ÖÖ Þ½   Þ½ Ö Þ¾   Þ¾ Ö Þ¿   Þ¿ Ö (6.

Tracking control of an under-actuated ship Like Pettersen and Nijmeijer (2000) we study the problem of stabilizing the tracking error dynamics (6. for comparison reasons we focus in Section 6.5f) reduces to the linear system Ö (6.4d.4f) is less clear.3) boils down to considering the dynamics in a frame with an earth fixed origin having the Ü.4c. a physical interpretation of the error co-ordinates (6.6) the tracking error dynamics becomes Ù Ú Ö Ü Ý Ñ¾¾ ´Ú Ö · Ú ÖÖ ´Øµ · ÚÖ Ö µ   ѽ½ Ù · ѽ ´Ù½   Ù½ Ö µ ѽ½ ½½ ½½ ѽ½ ¾¾ Ú   Ñ ´Ù Ö · Ù ÖÖ ´Øµ · ÙÖ Ö µ   Ñ ¾¾ ¾¾ ѽ½   Ѿ¾ ´Ù Ú · Ù ÚÖ · ÙÖ Ú µ   Ñ¿¿ Ö · ѽ¿¿ ´Ù¾   Ù¾ Ö µ Ñ¿¿ ¿¿ Ù   ÙÖ Ó×   ÚÖ × Ò · Ö Ý · ÖÖ ´ØµÝ Ú   ÚÖ Ó× · ÙÖ × Ò   Ö Ü   ÖÖ ´ØµÜ Ö (6. 6.and Ý -axis always oriented along the ship surge. 6.4b. This leads to the error co-ordinates ¾ Ü Ý ¿ ¾ Ó×  × Ò ¼ ×Ò Ó× ¼ ¼¿ ¾ Ü   ÜÖ ¿ ¼ Ý   ÝÖ ½   Ö (6. For that.8) is a new input.7) can be dealt with in a similar way. By defining the preliminary feedback Ù¾ where Ù¾ Ö   ´Ñ½½   Ѿ¾ µ´ÙÚ   ÙÖ ÚÖ µ · ¿¿ Ö Ö Þ¿ · Ñ¿¿ (6. Therefore. the change of co-ordinates (6. However.3 on stabilizing (6.76 Chapter 6.3 State-feedback: a cascaded approach We want to use the cascaded design approach for solving the state-feedback tracking problem. we follow the ideas presented in Section 3. for example   ½ Ö   ¾ Þ¿ ½ ¾ ¼ (6. This is done for both the ship and the reference. As mentioned.7) Therefore.2 and look for a control law for one of the two inputs. 6.6) When using the error co-ordinates (6.5c. Stabilizing (6.10) .and sway-axis.7).9) which can easily be stabilized by choosing a suitable control law for . the subsystem (6.4a.5). 6. 6. 6. we could as well study the problem of stabilizing the tracking error dynamics (6. 6. which is such that in closed loop a subsystem is asymptotically stabilized.4e.5). More natural error coordinates would be to consider the position errors Ü   ÜÖ and Ý   ÝÖ in a frame attached to the body of the ship (as for the mobile robot).

3.9) has been established. See Appendix A. as mentioned in Remark 5. we substitute Ö and Þ¿ in (6.13b) (6. that is.3. we can use any of the control laws available in literature for stabilizing linear time-varying systems.11) All that remains to be done. State-feedback: a cascaded approach 77 Next.5a. It follows from Corollary 2. 6. 6.2.1.10) and (6. Consider the system (6. 6.13d) If ÖÖ Ø is persistently exciting (PE) then the closed-loop system (6.8.12) ½ ¾ ¼ ´µ ¿ ´ · ½ · ½½   ¾¾µ ´ ¾¾ · ѽ½ ¿µ ´ ½ · ½½   ¾¾µ Ѿ¾ ½½ ¼ (6. 6.12) we are now able to formulate the cascaded systems based solution to the tracking control problem: . Notice that this condition is similar to that of the mobile robot studied in Chapter 4.11).5b.6. where also the reference angular velocity had to be persistently exciting. if the reference yaw velocity ÖÖ Ø is persistently exciting.12) is globally uniformly exponentially stable (GUES).5d.5e).11) in closed loop with the control law where ( ½ Ù½ µ satisfy Ù½ Ö   ½ Ù · ¾ ÖÖ ´ØµÚ   ¿ Þ½ ¾¾   ½½ ѽ½ Ѿ¾ · ÖÖ ´ØµÞ¾ (6.13c) (6.13a) (6. ´µ In addition to these results we arrived at the following.11) is uniformly completely controllable (UCC) if the reference yaw velocity ÖÖ Ø is persistently exciting.4 that the system (6. Combining the controllers (6. This results in ¼ ¼ Ù Ú Þ½ Þ¾ ¾ Ѿ¾ ½ ´Ù   Ù µ Ú Ö ´Øµ   ½½ Ù · ѽ½ Ö Ñ½½ ѽ½ ½ ½ Ö   ѽ½ Ù ÖÖ ´Øµ   Ѿ¾ Ú Ñ¾¾ ¾¾ Ù · Þ¾ ÖÖ ´Øµ Ú   Þ½ ÖÖ ´Øµ Ѿ¾ Ö Ø Ñ½½ Ö   Ѿ¾ ¾¾ which is just a linear time-varying system: Ù Ú Þ½ Þ¾ ¿ ¾ Ñ   Ñ ÖÖ ´Øµ ½½ ¾¾  Ñ ½½ ½½ ´µ ½ ¼ ¼ ½ ¼ ¿ ¾ Ù ¿ ¾ ѽ Ú · ¼ ¼ ¼ ÖÖ ´Øµ Þ½ ¼  ÖÖ ´Øµ ¼ Þ¾ ¼ ¼ ¼ ´µ ¿ ½½ Ù½   Ù½ Ö (6. Proof. 6.11. is to find a feedback controller for Ù½ that stabilizes the system (6. we assume that the stabilization of (6.3. Proposition 6.3. As a result.

10).8.14) has a cascaded structure: ¾ ´µ Ù ¿ Ú Þ½ Þ¾ ¾ ·   ½Ñ½½½½ ѽ½   Ѿ¾ ÖÖ ´Øµ ½ ¼ ¾ ·Ñ¾¾ Ö Ñ½½ ¾¾ Ѿ¾   Ö ´Øµ ¿   ѽ½ ¼ ½ Ö Þ¿   Ñ· ¿¿ ½ ¿¿ ¾ ´Ø Þ µ ¾ ¼ ßÞ  Ñ ¿¿ ´Ø Þ µ Ö Þ¿ ½ ½ ßÞ  ÖÖ ´Øµ ¼ ¼ ѽ½ Ö ¿ Ö ´Øµ ¾ Ù ¿ ¼ Ú Þ½ ÖÖ ´Øµ Þ¾ ¼ ¾ · ´Ú ·ÚÖ µ ¼ ¿   ´Ù ·ÙÖ µ ¼ Þ ·Þ Ö ¼  ´Þ ßÞÞ Ö µ ¼ · ´Ø Þ Þ µ Ѿ¾ ѽ½ ѽ½ Ѿ¾ ¾ ½ ¾ ½ ½ ¾ Ö Þ¿ From Proposition 6.3. Consider the ship tracking error dynamics (6. Applying Corollary 2.3.15) results in the globally uniformly exponentially stable closed-loop (sub)system   Ñ· ¿¿ ½ ¿¿  Ñ ¼ Ö .14b) ½ ¾ ¼ ¿ ´ · ½ · ½½   ¾¾µ ´ ¾¾ · ѽ½ ¿µ ´ ½ · ½½   ¾¾µ Ѿ¾ ½½ ¼ ѽ½ Ѿ¾ ¾¾   ½½   ¿¿ ¼ If ÙÖ . Proof. 6.e. Tracking control of an under-actuated ship Proposition 6. Þ½ Ö . 6. Notice that the stabilization of the “more natural” tracking error dynamics (6. Ø Þ½ Þ¾ satisfies (2.25).3. 6. ÚÖ .5) in closed loop with the control law Ù½ Ù¾ where Ù½ Ö   ½ Ù · ¾ ÖÖ ´ØµÚ   ¿ Þ½ Ù¾ Ö   ´Ñ½½   Ѿ¾ µ´Ù Ú · ÚÖ Ù · ÖÖ ´ØµÞ¾ · ÙÖ Ú µ   Ö   Þ¿ (6.5. Ù¾ Ù¾ Ö   ´Ñ½½   Ѿ¾ µ´ÙÚ   ÙÖ ÚÖ µ   Ö Ö     ¿¿ ¿¿ ¼ (6.7) follows along the same lines.3. due to the fact that ÙÖ .2 the control law (6. ´ µ ´ ´ µ µ Remark 6.14a) (6. ÚÖ .78 Chapter 6.11) globally uniformly exponentially stable works also.6 provides the desired result. then the closedloop system (6. Due to the design.1 we know that the system Þ½ ½ Ø Þ½ is globally uniformly exponentially stable (GUES) and from standard linear control that the system Þ¾ ¾ Ø Þ¾ is GUES. As pointed out in Remark 5.14) is globally Ã-exponentially stable. Þ½ Ö and Þ¾ Ö are bounded and ÖÖ Ø is persistently exciting (PE).3.2. Any control law for Ù½ that renders the system (6. i. Using the control law (6.. and Þ¾ Ö are bounded.14) is not the only control law that results in global Ã-exponential stability. Furthermore.4. the closed-loop system (6.5.

7) yields the linear time-varying system: Ѿ¾ Ö Ø Ñ½½ Ö   Ѿ¾ ¾¾ Ù Ú Ü Ý ¿ ¾ Ñ   Ñ ÖÖ ´Øµ  Ñ ´µ ½ ¼ ¼ ½ ¼ ¿ ¾Ù ¿ ¾ ѽ ¼ Ú · ¼ ¼ ÖÖ ´Øµ Ü ¼  ÖÖ ´Øµ ¼ Ý ¼ ¼ ¼ · ¾ ÖÖ ´ØµÚ   ¿ Ü ¿ ½½ Ù½   Ù½ Ö which is identical to (6. Experimental results Substitution of Ö ¾ 79 ¼ and ½½ ¾¾ ½½ ½½ ¼ in (6. The DSP sends the thruster commands to the ship via a radio-transmitter. NTNU. The model ship has a length of is equipped with four azimuth-controlled thrusters (i. The estimates for position and velocities generated by this observer are used for feedback in the control law. The maximum force from one thruster is approximately N. For stabilizing this system we can use the control law Ù½ where ( Ù½ Ö   ½ Ù · ÖÖ ´ØµÝ (6.e.4. The control law and position estimates are implemented on a Pentium 166 MHz PC which R is connected with the DSP via a dSPACE bus. By using Simulink­ blocks. ¼ The reference trajectory to be tracked was similar to that in (Pettersen and Nijmeijer 2000). Corollary 6. Navigation and Control Laboratory located at the Department of Engineering Cybernetics. then we have that the control law (6. Norway.meter pool with a depth of about meters.6. Trondheim.11). The vessel moves in a -by. From each camera the positions of the spheres are transfered via a serial line to a dSPACE signal processor (DSP). ½½ ½ ¼ ½ ¼ ½¼ ¼¾ Three spheres are mounted on the model of the vessel that can be identified by infra red cameras (for the simulation of a global positioning system (GPS)).15.. In the experiments we used Cybership I.16) ½ ) satisfy (6. No theoretical guarantee for a stable controller observer combination can be given (yet). and a mass of kg and shore supply vessel. From these positions the ship position and orientation can be calculated. 6. thrusters where the direction of the propeller force can be controlled). which is a scale model of an offm. as for nonlinear systems no separation principle exists. However.4.13). Assume that ÖÖ Ø is persistently exciting and that ÙÖ and ÚÖ are bounded. ´µ 6. Three infra red cameras are mounted in such a way that (almost always) one or two cameras can see the boat. in the experiments it turned out to work satisfactory.4 Experimental results To support our claims we performed some experiments at the Guidance. From namely a circle with a radius of meter using a constant surge velocity of ½ ¼¼ . m/s. A nonlinear passive observer of Fossen and Strand (1999) is used to estimate the unmeasured states. The sampling frequency used in the experiments was Hz.3. the software is compiled and then downloaded into the DSP.16) renders the closed-loop tracking error dynamics (6.7) globally Ã-exponentially stable.

Tracking control of an under-actuated ship Infrared camera Markers Cybership I Wave generator Figure 6. .80 Chapter 6. Navigation and Control Laboratory.2: Guidance.

the ..4. since these are restoring terms.14) that we obtained by a cascaded design with the control of Pettersen and Nijmeijer (2000) that was derived by means of backstepping.2).17a) ÖÖ ÖÖ · ѽ½ «ÖÖ ¾Þ¾ · ѽ½ «ÖÖ ¾ Þ½ Þ¿   ѽ½ ½ ½Þ½   ѽ½ ½ ÖÖ Þ¿ Þ½ Ö · ѽ½ «ÖÖ Ú   ѽ½ ½ ÖÖ Þ½ Þ¿ · ѽ½ ½Ö ¾ Þ¾ Ö ¾ ѽ½ ½ ¾ Þ½ Þ¿ · ѽ½ ¾«ÖÖ ÙÖ Þ¿ · ѽ½ «ÖÖ ¾Þ¿ Þ½ Ö   ÖÖ Ñ½½ ¾ Þ½ ѽ½ ¾ Þ¾ Þ¿   ѽ½ ½ ÖÖ Þ¾   · · ѽ½ ¾ Þ¿ Þ¾ Ö ¾ ¾ ¾ ¾   ѽ½ ¾ Þ¾ Þ¿   ѽ½ ¾ Þ¿ Þ¾ Ö   ѽ½Ö ¾ Ú · ѽ½ ¾ Þ½ Ö Ù¾ ´Ù Ú Ö Þ½ Þ¾ Þ¿ ص (6. and due to possible parameter uncertainties cancellations could result in destabilizing terms. This backstepping-based controller is given by Ù½ · ѽ½ ¾Ö ÙÖ Þ¿ Ö · ѽ½ ¾ ¾ ¾ ¾   ѽ½ Ö ÙÖ Þ¿   ѽ½ Ö Ù Þ¿ · ѽ½ Ö¾ Ú · ѽ½ ¾Ö Ù Þ¿ Ö ¾ ѽ½ ½ ¾ Þ¿ Þ½ Ö ½ ½ Þ¿ Þ¾ Ö · ѽ½ ½ ½ Þ¾ Þ¿   ÖÖ Ñ½½ ½ ¾ Þ¿ Þ½ Ö Ñ½½ ½ ¾ Þ½ Þ¿ · ѽ½ ¾«ÖÖ Ù Þ¿ · · (6. Experimental results the initial reference state 81 ÙÖ ´¼µ ÚÖ ´¼µ ÖÖ ´¼µ ÜÖ ´¼µ ÝÖ ´¼µ Ö ¼ ¼ m/s ¼ m/s ¼ ¼ rad/s ¿ m m rad ´¼µ and the requirement ÙÖ ´Øµ ÖÖ ´Øµ the reference trajectory ¼¼ ¼¼ m/s rad/s Ø Ø ¼ ¼ ÙÖ ÚÖ ÖÖ ÜÖ ÝÖ Ö Ì can be generated.6. As did Pettersen and Nijmeijer (2000) we chose in the experiments not to cancel or compensate for the damping terms (i. since it has to satisfy (6.e. For sake of not being incomplete.17b) « ­ where « ­ ´Ù Ú Ö Þ½ Þ¾ Þ¿ ص is a complex expression of 2782 terms ½ ¾ ½ ¾ ¿ ½ ¾ ½ ¾ ¿ ¾   ѽ½ ½ Ù   Ѿ¾ ÚÖ ÖÖ   Ѿ¾ ÚÖ Ö   Ѿ¾ Ú ÖÖ   Ѿ¾ Ú Ö   ѽ½ ½Ö ¾ Þ¾ Ö · ѽ½­Þ¾ Ö Þ¿ · ѽ½­Þ¿ Ö Þ¾ Ö   ѽ½ ½ Ù Ö   ѽ½ ­Þ½ over a little less complex expression of 64 terms. assume ½½ ¿¿ ). ¼ In the experiments we compare the control law (6.

This results in ´µ ¼¼ Ú½ Ú¾ · ´¼ ½ · ¾ ¿½Þ½ · ¿Þ¾ µÞ¿   ¼ ¼¼Ö   ¾ ¼Þ¿ · ¼ ¼½Ù · ½   ´¼ ¼¼ Þ½ · ¼ ¼¼¿Þ¾ µÞ¿  ½ ¼ Ù · ¿ ¼Ú   ¿Þ½ · ¾ ¿½Þ¾   ¼ ½ Ö Ú (6.18) where we take Ù½ and Ù¾ as in (6.17). ¼¾ ¼ ¾ ½ ¼ ¼ ¼¼ ¼ ¼ For tuning the gains of (6. The controller gains are given by ½ . for comparison we first look for a set of control parameters for (6.­ . We use optimal control to arrive at the control law Ù  ÃÜ for which the costs · ½ ¼ Ü´ØµÌ Éܴص · Ù´ØµÌ ÊٴصdØ .11) that resulted from the cascaded analysis. Therefore.14) we obtain   ½ Ù · ¾ ÖÖ ´ØµÚ   ¿ Þ½ · ÖÖ ´ØµÞ¾   ÖÖ ´Øµ   Þ¿ and Ú¾ where for the experiment ÖÖ Ø .14) more or less corresponding to the parameters of Pettersen and Nijmeijer (2000) for (6. ½ . Therefore.14) is to consider the two linear subsystems (6. This is a linear controller. ¾ .19b) In the cascaded analysis we first used Ú¾ for stabilizing the Ö Þ¿ Ì dynamics and while designing in the second stage Ú½ we assume that Ö Þ¿ . the first approach for tuning the controller (6. Tracking control of an under-actuated ship control law for Ù¾ is contained in Appendix B. Both can be expressed as a standard linear time-invariant system of the form Ü Ü Ù.20a) (6.20b) The second approach for tuning the controller (6. as well as the term with Ú .17).82 Chapter 6.9) and (6.14) we prefer a more systematic approach.19b).« . However.14) is to consider a first order Taylor approximation of the auxiliary signals (6.19a) and neglect the small terms in (6.18a) Ú¾ · ÚÖ Ù · ÙÖ Ú µ (6. . ¿ . using a computer model of the ship. and . First we define the auxiliary signals Ú½ and Ù½   Ù½ Ö Ù¾   Ù¾ Ö · ´Ñ½½   Ѿ¾ µ´Ù Ú Ú½ (6. This leads us to the control law Ù½ Ù¾ Ù½ Ö   ½ ¼ Ù · ¿ ¼Ú   ¿Þ½ · ¾ ¿½Þ¾ Ù¾ Ö   ´Ñ½½   Ѿ¾ µ´Ù Ú · ÚÖ Ù · ÙÖ Ú µ   ¼ ¼¼Ö   ¾ ¼Þ¿ (6.19a) · ¼ ¼¼¿Þ½   ¼ ¼¼ ´ ¼ µ Þ¾ (6.18b) and substituting for Ù½ and Ù¾ the control laws (6. we leave out the Ö and Þ¿ terms of (6. These gains were ¾ found by trial and error.

The third graph contains the error in orientation. For Ê we take the inverse of maximum allowed input. This results in the choice ¡ ½ ¾ É for the system (6. We did experiments with all three controllers.9). ¡Ü ( where Ü is the maximum error we can tolerate in Ü .6. is shown in Figure 6. As long as the camera status equals zero we have position measurements from the camera-system.17) with the cascaded controller (6. Experimental results 83 ½ are minimized. The bottom graph depicts the camera status. ´µ ´µ From the fact that the presented controllers can be applied successfully in experiments. trying to create as much waves as possible and disturbing the ship as much as he could.5. is presented in Figure 6.21b) for the system (6.21). The fourth and fifth graph depict the controls applied to the ship.4. The reason for showing this is that the infrared cameras from time to time loose track of the ship.21a) (6. The resulting performance of the cascaded controller (6. For É we choose a diagonal matrix with entries Õ ).21). To illustrate this robustness even more.17) as proposed in (Pettersen and Nijmeijer 2000). its cascaded-systems-based linearization (6.3 we can see that for instance after about seconds we had a temporary failure and in the of the camera-system.11) and ½¼ ¼ ¼ ¼ ¿ ¼ ¼ ¼ ¼ ¼ ¾¼ ¼ ¼ ¼ ¼ ½¼ ¼ ¼ ¼ ¼ Ê ½½ É Ù½ Ù¾ Ê ½½ (6.17) is presented in Figure 6.4. The results are .20) and the linear optimal control based cascaded controller (6. This explains the sudden change in the orientation error control Ù½ . When we compare the backstepping controller (6. we might conclude that they possess some robustness with respect to modeling errors and with respect to disturbances due to currents and wave drift forces. In this way we obtain for the control law: Ù½ Ö   ½¼ ¾ Ù · ¾Ú   Þ½ Ù¾ Ö   ´Ñ½½   Ѿ¾ µ´Ù Ú · ÚÖ Ù · ¾ Þ¾ · ÙÖ Ú µ   ¼¾Ö   Þ¿ To summarize: we have three different control laws available. of which the gains were chosen by means of optimal control theory. namely the backsteppingcontroller (6. In Figure 6.21) the tracking of the reference positions ÜÖ Ø and ÝÖ Ø is comparable.20). However. which was based on a linearization of the backstepping controller. In the first two graphs we compare the actual position of the ship with its desired position. we performed one experiment using the “optimal gains” in which the author was wearing boots and walking through the pool. ½¾¼ The resulting performance of the backstepping controller (6. but as soon as the camera status is non-zero we no longer get correct position measurements. the angular tracking error is considerably less for the cascaded controller. The resulting performance of the cascaded controller (6.3.

84 Chapter 6.3: Cascade controller (6.5 0 0 50 100 150 time [s] 200 250 300 Figure 6.20) with gains based on linearization of the backstepping controller. . Tracking control of an under-actuated ship 6 Ü-co-ordinate [m] Ü ÜÖ 5 4 3 0 5 50 100 150 time [s] 200 250 300 Ý-co-ordinate [m] 4 3 2 1 0 0 50 100 150 time [s] 200 250 Ý ÝÖ 300 20 angular error [deg] 10 0 -10 -20 0 1 surge control force [N] 50 100 150 time [s] 200 250 300 Ù½ 0 -1 0 1 50 100 150 time [s] 200 250 300 yaw control moment [Nm] Ù¾ 0 -1 0 50 100 150 time [s] 200 250 300 1 camera status 0.

. Experimental results 85 6 Ü-co-ordinate [m] Ü ÜÖ 5 4 3 0 5 50 100 150 time [s] 200 250 300 Ý-co-ordinate [m] 4 3 2 1 0 0 50 100 150 time [s] 200 250 Ý ÝÖ 300 20 angular error [deg] 10 0 -10 -20 0 1 surge control force [N] 50 100 150 time [s] 200 250 300 Ù½ 0 -1 0 1 50 100 150 time [s] 200 250 300 yaw control moment [Nm] Ù¾ 0 -1 0 50 100 150 time [s] 200 250 300 1 camera status 0.17).4.6.4: Backstepping controller (6.5 0 0 50 100 150 time [s] 200 250 300 Figure 6.

5: Cascade controller (6.5 0 0 50 100 150 time [s] 200 250 300 Figure 6.21) with gains based on optimal control. Tracking control of an under-actuated ship 6 Ü-co-ordinate [m] Ü ÜÖ 5 4 3 0 5 50 100 150 time [s] 200 250 300 Ý-co-ordinate [m] 4 3 2 1 0 0 50 100 150 time [s] 200 250 Ý ÝÖ 300 20 angular error [deg] 10 0 -10 -20 0 1 surge control force [N] 50 100 150 time [s] 200 250 300 Ù½ 0 -1 0 1 50 100 150 time [s] 200 250 300 yaw control moment [Nm] Ù¾ 0 -1 0 50 100 150 time [s] 200 250 300 1 camera status 0. .86 Chapter 6.

Experimental results 87 6 Ü-co-ordinate [m] Ü ÜÖ 5 4 3 0 5 50 100 150 time [s] 200 250 300 Ý-co-ordinate [m] 4 3 2 1 0 0 50 100 150 time [s] 200 250 Ý ÝÖ 300 20 angular error [deg] 10 0 -10 -20 0 1 surge control force [N] 50 100 150 time [s] 200 250 300 Ù½ 0 -1 0 1 50 100 150 time [s] 200 250 300 yaw control moment [Nm] Ù¾ 0 -1 0 50 100 150 time [s] 200 250 300 1 camera status 0.6: Cascade controller (6.21) under disturbance of author walking through the pool. .5 0 0 50 100 150 time [s] 200 250 300 Figure 6.6.4.

which is a common situation for many supply vessels. . Nevertheless a reasonable tracking performance was achieved. In an attempt to get better robustness results. By means of a cascaded approach we developed a global tracking controller for this tracking problem. adding integral control to the cascaded controller adds robustness against constant disturbances. Therefore.5 Concluding remarks In this chapter we studied the tracking problem for an under-actuated ship that has only surge control force and yaw control moment. 6. the cascaded approach might be helpful.6. This insight simplified the gaintuning a lot. since optimal control could be used to arrive at suitable gains. The controllers presented in this chapter also proved to work reasonably well in experiments. The cascaded approach reduced the problem of stabilizing the nonlinear tracking error dynamics to two separated problems of stabilizing linear systems.2.88 Chapter 6. Tracking control of an under-actuated ship depicted in Figure 6. The resulting control law has a much simpler structure than the backstepping based controller that was available in literature and which guarantees semi-global tracking. Obtaining asymptotic stability of a linear system under a constant disturbance is a well-known problem that can be solved by using integral control. Disturbances due to currents and wave drift forces are in general modeled by a constant force acting on the ship. As in the previous chapters. This implies a certain robustness against modeling errors and disturbances due to currents and wave drift forces. we could as well consider two linear systems. Starting from this we might be able to weaken this condition that the angular velocity of the reference should not tend to zero. The cascaded approach learned us that instead of looking at the nonlinear tracking error dynamics. Another possibility might be to consider the idea of uÆ -PE. Remark 4.6). the cascaded approach leads us to a simpler structure for backstepping (cf. A disadvantage of both the backstepping and the cascade controller is the demand that the reference angular velocity does not tend to zero. the camera system had much more difficulties in keeping track of the ship. It can be noticed that due to the heavy waves.

Part II Robot manipulators and adaptive control 89 .

.

One of the drawbacks of this approach is that the vector of gravitational forces is assumed to be known accurately. In what follows we briefly describe the subject of the three papers and finally give a short discussion on a number of relevant similarities in these papers.1 Paper I: Global asymptotic stability of robot manipulators with linear PID and PI¾D control This paper deals with the control of a special nonlinear mechanical system. but each focuses on different aspects than those discussed so far in the first part of the thesis. Whenever the gravitational vector is not known exactly and an estimate is used. which results in asymptotic stability. namely a fullyactuated rigid robot manipulator. The tracking problem considered is to move the manipulator to a desired fixed point. First we apply the global PD controller. Takegaki and Arimoto (1981) showed that a PD plus gravity compensation controller can globally asymptotically stabilize a rigid robot manipulator to any desired fixed point. Common practice is to use PID control to overcome this problem.g. Then we switch to the PID controller by activating the integral action. whereas Chapter 10 treats a new adaptive control problem for an under-actuated system. All three deal with the tracking control of nonlinear mechanical systems. Based on the ideas presented by Lefeber (1996) and applied to the bounded tracking control of chaotic systems in (Lefeber and Nijmeijer 1996. This particular tracking problem also goes under the name set-point control or is called the regulation problem. Lefeber and Nijmeijer 1997b) we combine a PD controller with a PID controller. Using this hybrid linear controller we are able to show global asymptotic stability. 91 . only local (or at best semi-global) asymptotic stability of this scheme has been proven so far (see e. the position error converges to a bounded steady state error. Lefeber and Nijmeijer 1997a.Chapter 7 Introduction to Part II The second part of this thesis consists of three papers. (Kelly 1995)).. 7. Chapter 8 and Chapter 9 deal with tracking of fully-actuated rigid robots. Specifically. However. which leads to a steady state error.

In the first part of this thesis we studied the tracking problem for nonlinear mechanical systems with non-holonomic constraints.3 Paper III: Adaptive tracking control of non-holonomic systems: an example As in the previous paper. In this approach. for all these problems feasibility of the reference trajectory .92 Chapter 7.2 Paper II: Adaptive and filtered visual servoing of planar robots This paper also deals with the regulation problem for rigid robot manipulators. but now with time visual servoing (under a fixed camera configuration). Furthermore. Kelly. In fixed-camera robotic systems. The objective is to move the robot in such a way that its end-effector reaches a desired target. since this problem is almost unexplored in adaptive control. the vision system provides an image position error. measured directly on the image plane as the visual distance between the target and end-effector positions. this third paper also studies a control problem where certain parameters are unknown. This error is used to drive the controller. Introduction to Part II 7. we extend in this paper the results of (Kelly and Marquez 1995. Kelly 1996) to the cases where velocity measurements are not available and the camera orientation parameter is unknown. which enables us to deal with constraints on the inputs. Kelly (1996) showed robustness of this controller in face of unknown radial lens distortions and uncertainty in the camera orientation. Among the existing approaches to solve this problem. Kelly and Marquez (1995) introduced a model of the vision system incorporating a perspective projection based on thin lens geometric optics and they derived a control law to solve the problem. The example illustrates that the formulation of the adaptive state-tracking problem is far from trivial. In case some system parameters are not known exactly. Hager. This paper deals with the adaptive state-tracking problem for the kinematic model of a four-wheel mobile robot with unknown length. Coste-Mani` re. However. The latter problem involves a nonlinear parameterized adaptive system for which special analysis and synthesis tools have to be developed. it has been recognized that the “imagebased” scheme possesses some degree of robustness against camera miscalibrations (Miyazaki and Masutani 1990. We assumed all system parameters to be known exactly and no disturbances to be present. e Kelly 1996. Asymptotic stability of the resulting closed-loop system was shown under the assumption that the orientation of the camera is known. 7. Lei and Ghosh 1993. Kelly 1996) to a class of visual servoing controllers which also includes the controllers reported in (Miyazaki and Masutani 1990. a vision system fixed in the world co-ordinate frame captures images of the robot and its environment. Chang and Morse 1995). At first glance this might look surprising since adaptive tracking problems have been studied throughout in literature. Shirkey and Spong 1996). Lei and Ghosh 1993. In this paper we extend the controllers proposed in (Kelly and Marquez 1995. Couvignou and Khosla 1995. No intrinsic camera parameters were assumed to be known and neither the robot’s inverse kinematics nor the inverse Jacobian was used. things change considerably. Our class of controllers also contains saturated controllers.

In that case the integrated error can be seen as an estimate for this unknown parameter and the additional integral action can be seen as an update law for this estimate. That is why in this second part we shift attention to some of the uncertainties that are of interest when studying tracking of nonlinear mechanical systems. Most results in adaptive control deal either with the adaptive stabilization problem or the adaptive output-tracking problem.1. The question then arises how to formulate the adaptive state-tracking problem. Discussion 93 is not an issue. we insist on a state-tracking problem instead of an outputtracking problem. all analysis is done on the known model. An other approach of the problem would be to view the unknown constant gravitational vector as an unknown parameter. However.4 Discussion This thesis is concerned with the tracking of nonlinear mechanical systems. As mentioned in Section 1. like in this example the length of the vehicle. By means of the kinematic vehicle model we illustrate in this paper the above mentioned conflict.4. one has as many inputs to the system as outputs. whereas in practice all kinds of uncertainties play a role. In the first part we focussed on tracking of under-actuated systems and developed a new approach which can lead to simple controllers. and specifying a feasible reference trajectory is in conflict with each other. no matter what value the unknown parameters have. To start with. since it turns out to be a priori guaranteed. Usually it is undesirable in the tracking problem for a mobile robot or ship that the system turns around and follows the reference trajectory backwards. . The result we obtain is just a switch between a PD and a PID control. In Chapter 8 we assume that the vector of gravitational forces is not known exactly. We illustrate in this paper that not knowing certain parameters. In case of the adaptive output tracking problem. The question remains if using integral action from the beginning can also result in global asymptotic stability.1). but this time in the presence of parametric uncertainties. In Chapter 10 we return to the under-actuated problem studied in Part I. The adaptive stabilization problem usually is studied for systems without drift which can be stabilized to an arbitrary point. Switching controllers is closely related to the results of Teel and Kapoor (1997) and Prieur and Praly (1999). Therefore.7. output-tracking does not suffice and state-tracking is what we are looking for. which is assumed to be accurate. The fully-actuated system studied in Chapter 8 and Chapter 9 is a rigid robot manipulator. This output can be tracked for any parameter. We propose a natural formulation for the adaptive state-tracking problem and present a general methodology for solving this problem. We show that the common practice of using PID control yields global asymptotic stability when the integral action is activated after some time. We like to extend the results of the first part of this thesis to the case where certain system parameters are unknown. in Chapter 8 and Chapter 9 we deal with uncertainties for fully-actuated systems. As a result an arbitrary signal can be specified for the output. 7. One of the necessary conditions in formulating the adaptive state-tracking problem is that in case the parameters are known it reduces to the state-tracking problem (as formulated in Section 1.

94

Chapter 7. Introduction to Part II

In Chapter 9 we study a different adaptive control problem. Here we adapt for the unknown orientation of the camera that provides us with measurements. The main difficulty in solving this adaptive problem is given by the fact that the unknown parameter enters the dynamics in a nonlinear way. As a result none of the standard linear-in-the-parameters adaptive control techniques can be used. Our solution consists of switching between two controllers which results in a chattering behavior. Although not presented, also a smooth semi-global solution is available. In Chapter 4 we studied the output-feedback tracking problem where part of the position was unmeasured. However, if we consider the same problem where we assume that the position is measured but not the orientation, then we run into similar problems. In this case we also have to reconstruct an unknown angle that enters the dynamics nonlinearly. As illustrated by Jakubiak, Lefeber, Tch´ n and Nijmeijer (2000) the problems turn out to be the o closely related. Actually, similar semi-global solutions can be derived. Results on adaptive control of nonlinearly parameterized systems (in a more general framework) can for instance be found in the work of Loh, Annaswamy and Skantze (1999) and Kojic and Annaswamy (1999). Whereas in Chapter 8 and Chapter 9 we studied the adaptive regulation problem for fullyactuated systems, in Chapter 10 we study the adaptive state-tracking problem for underactuated systems. In light of Chapter 8 and Chapter 9 the extension of the state-tracking problem as considered in Part I to an adaptive state-tracking problem seems to be a natural next step. As it turns out the first step to be made is arriving at a suitable problem formulation. A precise statement of the general adaptive state-tracking control problem is not so simple. An attempt is made in Chapter 10 by considering the example of a four-wheel mobile robot with unknown length. The reason for considering this example is not given from a practical point of view (since in practice the length of a mobile car can easily be measured), but mainly to illustrate the difficulties one runs into when considering the (also from a practical point of view interesting) adaptive state-tracking problem. For this specific example a problem formulation is presented which could form a basis to arrive at a general problem formulation of the adaptive tracking control problem as for instance presented by Lefeber and Nijmeijer (1998).

Chapter 8

Global asymptotic stability of robot manipulators with linear PID and PI¾D control
This chapter consists of the following paper: A. Lor´a, E. Lefeber, and H. Nijmeijer, “Global asymptotic stability of robot manipulators ı with linear PID and PI¾ D control,” 1999a, Submitted to Stability and Control: Theory and Applications. The current affiliation of the first author is: Antonio Lor´a ı Charge de Recherches au CNRS. Laboratoire d’Automatique de Grenoble B.P. 46, ENSIEG 38402, St. Martin d’H` res e FRANCE Tel: +33 4.76.82.62.30 Fax: +33 4.76.82.63.88 e-mail: aloria@inpg.fr URL: http:/ /www-lag.ensieg.inpg.fr/recherche/cser/people/aloria

95

96

Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control

ÐÓ Ð ×ÝÑÔØÓØ ËØ Ð ØÝ Ó ÊÓ ÓØ Å Ò ÔÙÐ ØÓÖ× Û Ø Ä Ò Ö ÈÁ Ò ÈÁ¾ ÓÒØÖÓÐ
ÒØÓÒ Ó ÄÓÖ
Ý
ÆÓÖÛ

Ý

Ö ÒÄ
Ò Ò Ë ÇºËº Æ Ö ¼¿ ×Ø

Ö ÀÒ ÆÑ Ö
Þ
Ö Ò Ò × ÔÐ ×× Ñ¸ Ý Ò ¸ ÖÒ Ø Ì × ÒÓÐÓ Ý

Þ

Ô ÖØÑ ÒØ Ó

Ò ÍÒ Ú Ö× ØÝ Ó

ÌÖÓÒ

ÆÇÊÏ ÐÓÖ Ø ºÒØÒÙºÒÓ ÔÔÐ Å Ø ÌÛ ÒØ Ò× ÖÐ Ò × ¸ Ñ Ø ×

Þ

ÙÐØÝ Ó

ÍÒ Ú Ö× ØÝ Ó ÈºÇº ÓÜ ¾½ ¸ Ì º ºÂºÄ ÙÐØÝ Ó Ò ÈÇ Å ¼¼

Æ Ø Ö

Ñ Ø ºÙØÛ ÒØ ºÒÐ Ò Ð Ò Ì Ò Ò Ö Ò

ÓÚ Ò ÍÒ Ú Ö× ØÝ Ó ÓÜ ½¿¸ Ì ¼¼ Å Æ Ø

ÒÓÐÓ Ý ÓÚ Ò

ÖÐ Ò ×

ÖÙ ÖÝ ¾ ¸ ¾¼¼¼

×ØÖ Ø
ÁÒ Ø × Ô Ô Ö Û Ö ×× Ø ÔÖÓ Ð Ñ Ó × Ø¹ÔÓ ÒØ ÓÒØÖÓÐ Ó ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ× Û Ø ÙÒ ÖØ Ò Ö Ú ØÝ ÒÓÛÐ Ý ÓÑ Ò Ò × Ú Ö Ð ÔÖ Ú ÓÙ× ÓÒØÖ ÙØ ÓÒ× ØÓ ÈÁ ÓÒØÖÓк Ì Ñ Ò ÓÒØÖ ÙØ ÓÒ × Ð Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö Û Ò×ÙÖ × ÐÓ Ð ×ÝÑÔ¹ ØÓØ ×Ø Ð ØÝ Ó Ø ÐÓ× ÐÓÓÔº Ì Ý ØÙÖ Ó Ø ÓÒØÖÓÐÐ Ö¸ Û ÐÐÓÛ× ØÓ ÔÖÓÚ ÐÓ Ð ØÝ × Ø Ø Ø ÒØ Ö Ø ÓÒ × ×Ø ÖØ Ø Ö × ÓÖØ ØÖ Ò× Òغ ÁÒ Ø × Ó ÙÒÑ ×ÙÖ Ð Ú ÐÓ Ø ×¸ × Ñ Ð Ö Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Ö × × ÓÛÒ ØÓ ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Þ Ø Ñ Ò ÔÙÐ ØÓÖº

½

1.8. Introduction 97 ½ ÁÒØÖÓ Ù Ø ÓÒ ½º½ Ä Ø Ö ØÙÖ Ö Ú Û ÖÓÑ Ø × Ñ Ò Ð Ô Ô Ö ½ Ø × Û ÐÐ ÒÓÛÒ ÒÓÛ Ø Ø È ÔÐÙ× Ö Ú ØÝ ÓÑÔ Ò× Ø ÓÒ ÓÒØÖÓÐÐ Ö Ò ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Þ Ö ¹ Ó ÒØ× Ñ Ò ÔÙÐ ØÓÖº ÀÓÛ Ú Ö¸ Ø × ÔÔÖÓ × ØÛÓ Ö Û × ´Û Ú Ò ÐÖ Ý ÜØ Ò× Ú ÐÝ ×ØÙ µ ½µ Ø Ú ØÓÖ Ó Ö Ú Ø Ø ÓÒ Ð ÓÖ × × ××ÙÑ ØÓ ÒÓÛÒ ÙÖ Ø ÐÝ Ò ¾µ Ú ÐÓ Ø × Ö Ò ØÓ Ò Ø Ø Ò ×× ÖÝ ÑÔ Ò º Ò Ó ×ÓÐÙØ ÓÒ ØÓ Ø ¬Ö×Ø ÔÖÓ Ð Ñ × ØÓ ÓÑÔ Ò× Ø ÓÖ Ø Ö Ú Ø Ø ÓÒ Ð Ú ØÓÖ Û Ø Ø ×Ø ×Ø Ñ Ø Ú Ð Ð º ÁØ × Û ÐÐ ÒÓÛÒ Ø Ø Ò ×Ù × ¸ Ø Ñ Ò ÔÙÐ ØÓÖ Ò ÐÓ× ÐÓÓÔ Û Ø × ÑÔÐ È ÓÒØÖÓÐÐ Ö Û ÐÐ Ü Ø ×ÓÑ ÖÓ Ù×ØÒ ×× ÔÖÓÔ ÖØ × ÑÓÖ ÔÖ × Ðݸ Ø ÔÓ× Ø ÓÒ ÖÖÓÖ Û ÐÐ ÓÒÚ Ö ØÓ ÓÙÒ ×Ø Ý ×Ø Ø ÖÖÓÖº Ì × ÔÖÓ Ð Ñ × ÒÓØ Ü ÐÙ× Ú ØÓ ÖÓ ÓØ ÓÒØÖÓÐ ÙØ Ø × Ó Ø Ò Ò ÓÙÒØ Ö Ò « Ö ÒØ Ò Ù×ØÖ Ð ÔÖÓ ×× ×º ØÝÔ Ð Ò Æ ÒØ Ö Ñ Ý × ØÓ Ù× ÈÁ ÓÒØÖÓи ÓÖ Ò ÐÐÝ ÔÖÓÔÓ× Ý Æ ÓÐ × Å ÒÓÖ× Ý Ò ½ ¾¾º ÁÒ Ø Û ×Ø ÖÒ Ð Ø Ö ØÙÖ Ø ¬Ö×Ø ×Ø Ð ØÝ ÔÖÓÓ Ó ÈÁ ÓÒØÖÓÐÐ Ö Ò ÐÓ× ÐÓÓÔ Û Ø Ö ¹ Ó ÒØ× Ñ Ò ÔÙÐ ØÓÖ × ØØÖ ÙØ ØÓ ¿ º ÍÒ ÓÖØÙÒ Ø Ðݸ Ù ØÓ ×ÓÑ Ñ Ø Ñ Ø Ð Ø Ò Ð Ø × Ó Ø ÑÓ Ð ÓÒÐÝ ÐÓ Ð ´ÓÖ Ø ×Ø × Ñ ÐÓ Ðµ ×ÝÑÔØÓØ ×Ø Ð ØÝ Ò ÔÖÓÚ Òº ´Ë ÓÖ Ò×Ø Ò ½½ µº ÓÒ ÖÒ Ò Ø ÔÖÓ Ð Ñ Ó ÙÒÑ ×ÙÖ Ð Ú ÐÓ Ø ×¸ Û ÒÓÛ Ø Ð ×Ø Ø ÓÐÐÓÛ Ò Ð Ò Ö ÝÒ Ñ ÔÓ× Ø ÓÒ ÓÒØÖÓÐÐ Ö× Û ÔÔ Ö Ò Ô Ò ÒØÐÝ ½¼¸ ½¸ ¸ ¸ × Ð×Ó ½ Û Ö Ø ÓÒ ÔØ Ó Ä ÓÒØÖÓÐÐ Ö× Û × ÒØÖÓ Ù Ò Û Ò Ö Ð Þ × Ø Ö ×ÙÐØ× Ó Ø ÔÖ Ú ÓÙ× Ö Ö Ò ×º × Ò Ø × ÑÔÐ È ÓÒØÖÓÐ × ÓÒ Ò ÜÔ Ø Ø Ø ÓÒ ÓÑÔ Ò× Ø × Û Ø ÓÒ×Ø ÒØ Ú ØÓÖ ×Ø Ñ Ø Ó Ø Ö Ú Ø Ø ÓÒ Ð ÓÖ × Ò×Ø Ó Ø ØÖÙ ÓÒ ¸ Ø Ñ Ò ÔÙÐ ØÓÖ ÖÖÓÖ ØÖ ØÓÖ × Û ÐÐ ÓÒÚ Ö ØÓ ÓÙÒ ÓÑ Òº ÙÖØ ÖÑÓÖ ¸ × Ò Ø × Ó Ñ ×ÙÖ Ð Ú ÐÓ Ø ×¸ ÓÒ Ò ÔÖÓÚ Ø Ø Ø ×Ø Ý ×Ø Ø ÖÖÓÖ Ò ÐÑÒ Ø Ý Ò Ò ÒØ Ö ØÓÖº ÅÓÖ ÔÖ × Ðݸ Ø ÈÁ¾ ÓÒØÖÓÐÐ Ö ÓÖ Ò ÐÐÝ ÒØÖÓ Ù Ò ½ ¸ × × ÙÔÓÒ Ø È ×ØÖÙ ØÙÖ Ò Ø ÔÔÖÓÜ Ñ Ø « Ö ÒØ Ø ÓÒ ¬ÐØ Ö × ÔÖÓÔÓ× Ò ½¼ ¸ Ò ÓÑ Ò Ø ÓÒ Û Ø ÓÙ Ð ÒØ Ö ØÓÖ Ó Ø ÔÓ× Ø ÓÒ ÖÖÓÖ Ò Ó Ø ¬ÐØ Ö ÓÙØÔÙغ ÍÒ ÓÖØÙÒ Ø Ðݸ Ù ØÓ ×ÓÑ Ø Ò Ð Æ ÙÐØ ×¸ ÓÒ Ò ÔÖÓÚ ÓÒÐÝ × Ñ ¹ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð Øݸ × Ð×Ó º ÓÖ Ø × Ó Ñ ×ÙÖ Ð Ú ÐÓ Ø × ÓÒ Ò × Ò¸ Û Ø ×ÓÑ ×Ñ ÖØ ÑÓ ¬ Ø ÓÒ׸ ÒÓÒÐ Ò Ö ÈÁ ³× Û Ù Ö ÒØ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð Øݺ × Ö × Û ÒÓÛ¸ Ø ¬Ö×Ø ÒÓÒÐ Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö × Ù ØÓ ½ Û Û × Ò×Ô Ö ÙÔÓÒ Ø Ö ×ÙÐØ× Ó ÌÓÑ ¾½ º ÌÓÑ ÔÖÓÔÓ× È ÔÐÙ× ÔØ Ú Ö Ú ØÝ Ò ÐÐ Ø ÓÒ Ò Ù× ÒÓÖÑ Ð Þ Ø ÓÒ ´¬Ö×ØÐÝ ÒØÖÓ Ù Ý ÃÓ Ø× Ò ½¾ µ ØÓ ÔÖÓÚ ÐÓ Ð ×ÝÑÔØÓØ ÓÒÚ Ö Ò º Í× Ò Ø × Ñ ÒÓÖÑ Ð Þ Ø ÓÒ ¸ à ÐÐÝ × ÓÛ Ò Ø Ø ÐÓ Ð ×ÝÑÔØÓØ ÓÒÚ Ö Ò × ×Ø ÐÐ ÔÓ×× Ð ÒØ × Û Ò ÓÒ ÓÑÔ Ò× Ø × ÓÖ Ø Ö Ú ØÝ ÓÖ × Ú ÐÙ Ø ØØ × Ö ÔÓ× Ø ÓÒº Ì Ð ØØ Ö ÐÐÓÛ× ØÓ Ö ÓÖÑÙÐ Ø Ø ÓÒØÖÓÐÐ Ö Ó × ÒÓÖÑ Ð Þ ÈÁ º Ä Ø Ö¸ Ö ÑÓØÓ ¾ ÔÖÓÔÓ× ØÓ Ù× × ØÙÖ Ø ÔÖÓÔÓÖØ ÓÒ Ð Ø ÖѺ Ì × ÐÔ× Ò Ø × Ñ Û Ý × Ø ÒÓÖÑ Ð Þ Ø ÓÒ ØÓ ÓÔ Û Ø Ø Ø Ö ÓÖ Ö Ø ÖÑ× Û ÔÔ Ö Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ö Ú Ø Ú Ò ÑÔ Ð Ñ Ò ÐÓ Ð Øݺ ½ Ú Ò Ø ÓÙ Û Ý Û Ù× Ø Ã ÐÐÝ ÔÖ × ÒØ ÈÁ ÕÙ Ð ¬ Öº × Ö ×ÙÐØ × Ò ÔØ Ú ÓÒØÖÓÐÐ Ö¸ Ò × Ø ÓÒ Ø Û ÐÐ ÓÑ Ð Ö ¾ .

98 Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control × Ö × Û ÒÓÛ¸ Ø Ö Ü ×Ø ÒÓ ÔÖÓÓ Ó ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ð Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö Ò ÐÓ× ÐÓÓÔ Û Ø ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖº ÁÒ Ø × Ô Ô Ö Û Ù× ×ÓÑ Û ÐÐ ÒÓÛÒ Ö ×ÙÐØ× ØÓ ÔÖÓÚ Ø Ø Ø × Ø¹ÔÓ ÒØ Ö ÙÐ Ø ÓÒ Ò ×Ø Ð × Ý Ñ Ò× Ó Ð Ò Ö Ð Ý ÈÁ ÓÒØÖÓÐÐ Ö¸ Ø Ø × × ÑÔÐ È ÓÒØÖÓÐÐ Ö ØÓ Û Ò ÒØ Ö Ð Ø ÓÒ × Ø Ö ×ÓÑ ØÖ Ò× ÒØ Ó Ø Ñ º ÁÒ Ø × ÕÙ Ð Û Û ÐÐ Ö Ö ØÓ Ø × ÓÒØÖÓÐÐ Ö × Ð Ý ÈÁ ¸ Ò × ÓÖØ ÈÁ º Ð×Ó¸ Ò × Û Ò ÒÓ Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ× Ö Ú Ð Ð Û × ÓÛ Ò × Ñ Ð Ö Û Ý Ø Ø Ø ÒØ Ö Ð Ø ÓÒ Ó ÈÁ¾ ÓÒØÖÓÐÐ Ö Ò Ð Ý × ØÓ Ù Ö ÒØ Ø ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÐÓ× ÐÓÓÔº Ì × Ð Ò Ö ÓÒØÖÓÐÐ Ö Û ÐÐ Ö ÖÖ ØÓ Ò Ø × ÕÙ Ð × Ð Ý ÈÁ¾ ÓÖ Ò × ÓÖØ ÈÁ¾ º ÇÙÖ ÔÔÖÓ × Ò×Ô Ö ÓÒ Ø × Ó ÓÑÔÓ× Ø ÓÒØÖÓÐ Ú ÐÓÔ Ò¾ ½ º Ì Ó Ø ÓÑÔÓ× Ø ÓÒØÖÓÐ ÔÔÖÓ × × ÑÔÐ Ò ÔÖ Ø ÐÐÝ ÔÔ Ð Ò ØÓ ÔÔÐÝ Ò ¬Ö×Ø Ô × ¸ ÐÓ Ð ÓÒØÖÓÐ Ð Û¸ Û Ö Ú × Ø ÐÓ× ÐÓÓÔ ØÖ ØÓÖ × Ò× ×ÓÑ ÔÖ ¹×Ô ¬ ÓÙÒ × Øº ÁÒ Ø × ÓÒ Ô × ¸ ÑÓÖ ÔÖ × ÐÝ Ø Ø Ñ Ò×Ø ÒØ Ø× Û Ò Ø ØÖ ØÓÖ × Ö ÓÒØ Ò Ò Ø ÓÙÒ ÓÑ Ò¸ ÓÒ ×Û Ø × ØÓ ÐÓ ÐÐÝ ×Ø Ð Þ Ò ÓÒØÖÓÐ Ð Û Û Ö Ú × Ø ØÖ Ò ÖÖÓÖ ØÓ Þ ÖÓº ×Ù ×× ÙÐ Ù× Ó Ø × ÔÔÖÓ Ò × ÙÔÓÒ Ø Ð ØÝ Ó × Ò Ò ÓØ ÓÒØÖÓÐÐ Ö× Ò Û Ý ×Ù Ø Ø Ø ÓÙÒ × Ø Ó Ø ¬Ö×Ø Ô × ¸ × ÓÒØ Ò ÛØ ÒØ ÓÑ Ò Ó ØØÖ Ø ÓÒ × Ò ÓÖ Ø ÐÓ× ÐÓÓÔ Ò Ø × ÓÒ Ô × º Ò ÐÐÝ Ø × Ð×Ó ÛÓÖØ Ñ ÒØ ÓÒ Ò Ø Ö Ð Ø ß ÙØ « Ö ÒØß ÛÓÖ ¾¼ Û Ö Ø ÙØ ÓÖ× ÔÖÓÔÓ× Ò Ð ÓÖ Ø Ñ ØÓ ÓÑ Ò ÐÓ Ð Û Ø ÐÓ Ð ÓÒØÖÓÐÐ Ö× Û Ø Ø Ñ Ø ÑÔÖÓÚ Ò ÓØ ÖÓ Ù×ØÒ ×× Ò Ô Ö ÓÖÑ Ò º ½º¾ ÅÓ Ð Ò ÔÖÓ Ð Ñ ÓÖÑÙÐ Ø ÓÒ Ì Ö ¹ Ó ÒØ× ÖÓ ÓØ Ò Ø Ò Ö Ý × Ú Ò Ý Ì ´Õ Õµ ½ Õ ´Õ µÕ¸ Û Ö Õ ¾ ÁÊÒ Ö ÔÖ × ÒØ× ¾ Ø Ð Ò ÔÓ× Ø ÓÒ׸ ´Õ µ ´Õ µ ¼ × Ø ÖÓ ÓØ Ò ÖØ Ñ ØÖ Ü¸ Ò Ø ÔÓØ ÒØ Ð Ò Ö Ý Ò Ö Ø Ò Ö Ú ØÝ ÓÖ × × ÒÓØ Ý Í ´Õ µº ÔÔÐÝ Ò Ø ÙÐ Ö¹Ä Ö Ò ÕÙ Ø ÓÒ× Û Ó Ø Ò Ø Û ÐÐ ÒÓÛÒ ÑÓ Ð Û Ö ´Õ µ ´Õ µ¸ ´Õ ÕµÕ Ö ÔÖ × ÒØ× Ø ÓÖ ÓÐ × Ò ÒØÖ Ù Ð ÓÖ ×¸ Ò Ù ¾ ÁÊÒ Ö Ø ÔÔÐ ØÓÖÕ٠׺ ÁØ × Ð×Ó Û ÐÐ ÒÓÛÒ ÒÓÛ ´× ÓÖ Ò×Ø Ò ½ µ Ø Ø Ø ÓÐÐÓÛ Ò ÔÖÓÔ ÖØ × ÓÐ º Í Õ ´Õ µÕ · ´Õ ÕµÕ · ´Õ µ Ù ´½µ È ½ ÓÖ ÐÐ Õ ¾ ÁÊÒ Ø Ñ ØÖ Ü ´Õ µ × ÔÓ× Ø Ú ¬Ò Ø Ò ¸ Û Ø ×Ù Ø Ð ØÓÖ Þ Ø ÓÒ ´ÑÓÖ ÔÖ × ÐÝ Ù× Ò Ø ×Ó¹ ÐÐ Ö ×ØÓ« Ð ×ÝÑ ÓÐ× Ó Ø ¬Ö×Ø Ò µ Ø Ñ ØÖ Ü Æ ´Õ Õ µ ´Õ µ   ¾ ´Õ Õµ × × Û¹×ÝÑÑ ØÖ º ÅÓÖ ÓÚ Ö¸ Ø Ö Ü ×Ø ×ÓÑ ÔÓ× Ø Ú ÓÒ×Ø ÒØ× Ñ Ò Å ×Ù Ø Ø ÑÁ ´Õ µ ­ ­ ×ÙÔ ­ Ò­ Õ¾ÁÊ ÅÁ ´¾µ Ⱦ Ì Ö Ü ×Ø× ×ÓÑ ÔÓ× Ø Ú ÓÒ×Ø ÒØ× Ò Ø Ø ÓÖ ÐÐ Õ ¾ ÁÊÒ ­ ¾ Í ´Õ µ ­ ­ ­ ¾ Õ Ú ×Ù ´¿µ ¾ ÁØ × ÛÓÖØ Ñ ÒØ ÓÒ Ò Ø Ø Ø Ò Ñ ÓÑÔÓ× Ø ÓÒØÖÓÐ ÔØ Þ « Ö ÒØ ÔÔÖÓ × Ø Ò Ø ÓÒ Ù× Ò Ø × Ô Ô Ö¸ × × ÐÖ Ý Ò Ù× ÓÖ Ò×Ø Ò ½ º Ý ÓØ Ö ÙØ ÓÖ× ØÓ ¿ .

Introduction 99 Ú ­ ­ ­ Í ´Õ µ ­ ­ ×ÙÔ ­ ­ Õ ­ Õ¾ÁÊÒ ´ µ È¿ Ì Ñ ØÖ Ü ´Ü Ý µ × ÓÙÒ Ò Ü Ò Ð Ò Ö Ò Ý ¸ Ø Ø ×¸ ÓÖ ÐÐ Þ ´Ü Ý µÞ ´Ü Ý µ ´Ü Þ µÝ Ý ¾ ÁÊÒ ´ µ ´ µ ¼ ÁÒ Ø × Ô Ô Ö Û Ö ÒØ Ö ×Ø Ò Ø ×ÓÐÙØ ÓÒ ØÓ Ø ×Ø Ø Ò ÓÙØÔÙØ ´ÔÓ× Ø ÓÒµ × Ø¹ÔÓ ÒØ ÓÒØÖÓÐ ÔÖÓ Ð Ñ Ó ´½µ ××ÙÑ Ò Ø Ø Ø ÔÓØ ÒØ Ð Ò Ö Ý Í ´Õ µ¸ × ÒÓØ Ü ØÐÝ ÒÓÛÒº ÅÓÖ ÔÖ × Ðݸ ÓÒ× Ö Ø ÓÐÐÓÛ Ò ÔÖÓ Ð Ñ׺ Ë Ø¹ÔÓ ÒØ ÓÒØÖÓÐ ÔÖÓ Ð Ñ Û Ø ÙÒ ÖØ Ò Ö Ú ØÝ ÒÓÛÐ º ××ÙÑ Ø Ø Ø Ö Ú Ø Ø ÓÒ Ð Ò Ö Ý ÙÒ Ø ÓÒ Í ´Õ µ × ÒÓØ Ü ØÐÝ ÒÓÛÒ ÙØ ÓÒÐÝ Ø× ×¹ Ø Ñ Ø Í ´Õ µ × Ú Ð Ð º ÅÓÖ ÓÚ Ö¸ ××ÙÑ Ø Ø Ø ×Ø Ñ Ø Ó Ø Ö Ú Ø Ø ÓÒ Ð ÓÖ × Í Ú ØÓÖ¸ ´Õ µ ´Õ µ × Ø ×¬ × Õ Ú Õ ÁÊÒ ×ÙÔ ¾ ´Õ µ Õ ¾ ÁÊÒ ´ µ Û Ö Ú × ¬Ò ´×Ø Ø ´ÓÙØÔÙØ Ò ´ µº ÍÒ Ö Ø × ÓÒ Ø ÓÒ× × Ò ÓÒØ ÒÙÓÙ× ÓÒØÖÓÐ Ð Û× µ Ù Ù´Ø Õ Õ µ µ Ù Ù´Ø Õ Õ µ¸ Õ ´Õ Õ µ ÖÖÓÖ Õ ¼ Ú Ò Ø × Ó Ð¸ ÑÓÖ ÔÖ ¼ ¼ Ø Õ ×Ù Ø Ø¸ Ú Ò ÒÝ × Ö ÓÒ×Ø ÒØ ÔÓ× Ø ÓÒ Ø Ú Ö Òظ Ø Ø ×¸ ÓÖ ÒÝ Ò Ø Ð ÓÒ Ø ÓÒ× Ø  Õ ×ÝÑÔØÓØ ÐÐÝ ÓÒ¹ ´ µ × ÐÝ ´ µ Ð Ñ Õ ´Øµ ½ ÁÒ Ô ÖØ ÙÐ Ö Û Ö ÒØ Ö ×Ø Ò ÈÁ ¹Ð Û × ÓÖ ÓÒØÖÓÐÐ Ö× Ó Ø ÓÖÑ Ù ÓÒØÖÓÐ Ð Û× Ô   ÔÕ ´Øµ ´Õ µ   Õ · ´Øµ Ø Ø× Ø×   ÊØ Ø× Õ ´×µ × · ´Õ µ ¼ ´½¼µ ÓÖ Ø × Û Ò Ó ÒØ Ú ÐÓ Ø × Ö Ñ ×ÙÖ º ÁÒ Ø × ÕÙ Ð ÓÒØÖÓÐÐ Ö× Ð ´ µ¸ ´½¼µ Û ÐÐ Ö ÖÖ ØÓ × ÈÁ º ÁÒ Ø × Û Ò Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ× Ö ÒÓØ Ú Ð Ð Û × ÓÖ ÔÓ× Ø ÓÒ ÈÁ¾ ÓÒØÖÓÐÐ Ö¸ Ø Ø × Ù   ÔÕ   · ´Øµ Ô Õ ´½½µ ¼ Ô Ø Ô· ´½¾µ ¼ Ø Ø× Ø× ´Õ µ   ÊØ Ø× ´Õ ´×µ   ´×µµ × · ´Õ µ Ø ´½¿µ .8.1.

100 Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control Ì Ô Ô Ö × ÓÖ Ò Þ × ÓÐÐÓÛ׺ ÁÒ × Ø ÓÒ ¾ Û Ò ÐÝÞ Ò ÖØ Ò Ø Ð ×ÓÑ Ö ×ÙÐØ× Û ÔÔ Ö ÙÒ Ñ ÒØ Ð ØÓ ÓÙÖ Ñ Ò ÓÒØÖ ÙØ ÓÒ׺ Ë Ø ÓÒ ¿ ÓÒØ Ò× ÓÙÖ Ñ Ò Ö ×ÙÐØ׸ Ø Ø ×¸ Û × ÓÛ Ø Ø Ø ÈÁ Ò ÈÁ¾ ÓÒØÖÓÐÐ Ö× ×ÓÐÚ Ø ÐÓ Ð × Ø¹ÔÓ ÒØ ÓÒØÖÓÐ ÔÖÓ Ð Ñ Û Ø ÙÒ ÖØ Ò Ö Ú ØÝ ÒÓÛÐ º ÁÒ × Ø ÓÒ Û × Ù×× Ø Ú ÒØ × Ó ÓÙÖ Ö ×ÙÐØ× ÓÚ Ö Ø Ñ ÒØ ÓÒ ÒÓÒÐ Ò Ö ÈÁ ³×º Ë Ø ÓÒ ÔÖ × ÒØ× ÓÑÔ Ö Ø Ú × ÑÙÐ Ø ÓÒ ×ØÙ Ý Ó ÓÙÖ ÓÒØÖ ÙØ ÓÒ× Ò×Ø ÔÖ Ú ÓÙ× Ö ×ÙÐØ׺ Ï ¬Ò × Ø Ô Ô Ö Û Ø ×ÓÑ ÓÒ ÐÙ Ò Ö Ñ Ö × Ò × Ø ÓÒ º ÆÓØ Ø ÓÒº ÁÒ Ø × Ô Ô Ö Û Ù× ¡ ÓÖ Ø Ù Ð Ò ÒÓÖÑ Ó Ú ØÓÖ× Ò Ñ ØÖ ×º Ï ÒÓØ Ý ÔÑ Ò ÔÅ Ø ×Ñ ÐÐ ×Ø Ò Ð Ö ×Ø ÒÚ ÐÙ × Ó Ñ ØÖ Ü ÃÔ Ò × Ñ Ð ÖÐÝ ÓÖ ÒÝ Ñ ØÖ Ü Å ¾ ÁÊÒ¢Ò º ¾ ÈÖ Ð Ñ Ò ÖÝ Ö ×ÙÐØ× ÁÒ ÓÖ Ö ØÓ ÔÙØ ÓÙÖ ÓÒØÖ ÙØ ÓÒ× Ò Ô Ö×Ô Ø Ú Ò ØÓ ÒØÖÓ Ù ×ÓÑ ÒÓØ Ø ÓÒ Ù× Ò Ø × Õ٠и Û ¬Ò Ø ÓÒÚ Ò ÒØ ØÓ × Ö Ò ÑÓÖ Ø Ð ×ÓÑ Ó Ø ÓÚ ¹Ñ ÒØ ÓÒ ÈÁ Ò È ÓÒØÖÓÐÐ Ö׺ Ú Ò Ø ÓÙ ×ÓÑ Ó Ø × Ö ×ÙÐØ× Ö Û ÐÐ ÒÓÛÒ ÓÖ Ò × ÐÝ Ö Ú Ø Ý Ö ÙÒ Ñ ÒØ Ð ØÓ ÓÙÖ Ñ Ò ÓÒØÖ ÙØ ÓÒ׺ ¾º½ Ö×Ø × Ñ ×ÙÖ Ð Ú ÐÓ Ø × Ø × ÓÒ Ø Ö ×ÙÐØ× Ó ½ ÙÒ ÖØ ÒØÝ Ó ´Õ µº ÓÒ× Ò ÖØ Ù ¾¾ Û ÔÖ × ÒØ ÐÓÛ × ÑÔÐ ÖÓ Ù×ØÒ ×× Ö ×ÙÐØ Ú ×¹ ¹Ú × ¾º½ ÈÖÓÔÓ× Ø ÓÒº ÓÒØÖÓÐ Ð Û ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø È ´½ µ   ÃÔ Õ   Ã Õ · ´Õ µ Ä Ø ÔÑ ¸ Ø Ò Ø Ö Ü ×Ø× ÙÒ ÕÙ ÕÙ Ð Ö ÙÑ ÔÓ ÒØ ´Õ Õ µ ´¼ Õ×µ ÓÖ Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñº Ì ÔÓ ÒØ ´Õ Õ µ ´¼ Õ× µ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð ÓÖ ´½µ¸ ´½ µ Ò Ø ×Ø Ý ×Ø Ø ÖÖÓÖ Õ× Õ×   Õ × Ø ×¬ × ¾ Ú Õ× ´½ µ Ô Ñ £ Õ ÈÖÓÓ º Ì ÐÓ× ÐÓÓÔ ÕÙ Ø ÓÒ ´½µ¸ ´½ µ × ´Õ µÕ · ´Õ ÕµÕ · ´Õ µ   ´Õ µ · ÃÔ Õ · Ã Ò ×Ý×Ø Ñ Û Ø ÔÓØ ÒØ Ð Ò Ö Ý ´ µ Í ÚÒ Ý ¼ ´½ µ ËÝ×Ø Ñ ´½ µ × Ä Ö Ò ½ Õ ÃÔ Õ ¾ ÁØ × Û ÐÐ ÒÓÛÒ Ø Ø ´½ µ × Ø× ÕÙ Ð Ö Ø Ø Ñ Ò Ñ Ó Í½ ´Õ µº ÌÓ Ú ÐÙ Ø Ø × ÕÙ Ð Ö Û Ð ÙÐ Ø Ø Ö Ø Ð ÔÓ ÒØ× Ó Í½ ´Õ µ × Ý¸ ÐÐ ÔÓ ÒØ× Õ Õ× × Ø × Ý Ò Í½ Õ ´Õ µ   Í ´Õ µ   Õ ´Õ µ · ͽ Õ ´Õ× µ ¼ ¸ ÃÔ Õ× ´   Õ µ · ´Õ×µ   ´Õ µ ¼ ´½ µ .

8. Preliminary results 101 ÑÓÖ ÓÚ Ö Ø ÕÙ Ð Ö ÙÑ Õ Õ× × ÐÓ Ð Ò ÙÒ ÕÙ Û Ö × Ø ×¬ × ´¿µº Ì ÔÑ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ö ×ÙÐØ × Ø Ò ×Ø Ð × ÛØ ÐÔ Ó Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ò Ø ½ ν ´Õ Õ µ Õ ´Õ µÕ · ͽ ´Õ µ ´½ µ ¾ Û ÓÖÖ ×ÔÓÒ × ØÓ Ø ØÓØ Ð Ò Ö Ý Ó Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ ´½ µ¸ Ò Ø × ÔÓ× Ø Ú ¬Ò Ø Ò ÑÓÖ ÓÚ Ö × ÐÓ Ð Ò ÙÒ ÕÙ Ñ Ò ÑÙÑ Ø ´Õ Õ µ ´¼ Õ×µ º Ì ÔÑ ØÑ Ö Ú Ø Ú Ó Î½ ÐÓÒ Ø ØÖ ØÓÖ × Ó ´½ µ × Î½ ´Õ µ  Õ Ã Õ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÕÙ Ð Ö ÙÑ ´Õ Õ µ ´¼ Õ×µ ÑÑ Ø ÐÝ ÓÐÐÓÛ× Ù× Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò ÔÖ Ò ÔÐ º Ò ÐÐÝ Ø ÓÙÒ ÓÖ Ø ×Ø Ý ×Ø Ø ÖÖÓÖ ¬Ò Ò ´½ µ × × ÐÝ Ö Ú ÖÓÑ ´½ µ Ù× Ò Ø ØÖ Ò Ð Ò ÕÙ Ð ØÝ Ò Ø ÓÒ Ø ÓÒ× ÓÒ Ú Ú Ò Ý ´ µ¸ ´ µº ¤ ÀÓÛ Ú Ö¸ × Ø × Û ÐÐ ÒÓÛÒ Ø ×Ø Ý ×Ø Ø ÖÖÓÖ Õ× Ò ÐÑÒ Ø Ý Ø Ù× Ó Ò ÒØ Ö ØÓÖ¸ Ø × Ö ×ÙÐØ Û × ¬Ö×ØÐÝ ÔÖÓÚ Ò ¿ º Ê ÓÖÑÙÐ Ø Ò ´ ÓÖ ÙÖØ Ö Ò ÐÝ× ×µ Ø ÓÖ Ò Ð ÓÒØÖ ÙØ ÓÒ Ó ¿ Û Ú Ø ÓÐÐÓÛ Ò ¾º¾ ÈÖÓÔÓ× Ø ÓÒº ÓÒ× ÖØ Ù ÝÒ Ñ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø ÈÁ ÓÒØÖÓÐ Ð Û ´½ µ ´¾¼µ  ÃÔÕ   Ã Õ ·  Ã Õ ´¼µ Ò ¼ ¾ ÁÊ Û Ö ÃÔ ¸ à ¸ Ò Ã Ö ÓÒ Ð ÔÓ× Ø Ú ¬Ò Ø Ñ ØÖ ×º Á ÃÔ × ×ÙÆ Ø ÐÓ× ÐÓÓÔ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Ø Ø ÓÖ Ò Ü ÓÐ Õ Õ ÈÖÓÓ º ÒØÐÝ Ð Ö ¼º Ø Ò £ ÓÓ× ÒÝ ÔÓ× Ø Ú ¬Ò Ø ÃÔ ¼ ÓÒ Ð Ñ ØÖ Ü ÃÔ Ò Ð Ø ¼ ÃÔ · à ½ ´¾½µ Ò ´¾¾µ ´¾¿µ Û Ö ¼ × ´×Ñ Ðе ÓÒ×Ø ÒØ ØÓ Ø ÖÑ Ò ¸ Ð ÖÐÝ ÃÔ × Ð×Ó ÔÓ× Ø Ú ¬Ò Ø ÓÒ Ð ÓÖ ÒÝ ¼º Ì Ò Ø ÖÖÓÖ ÕÙ Ø ÓÒ ´½µ¸ ´½ µ¸ ´¾¼µ Ò ÛÖ ØØ Ò × ¼ ´Õ µÕ · ´Õ ÕµÕ · ´Õ µ   ´Õ µ · ÃÔ Õ · Ã Õ  Ã Õ  ½ Ã Õ · Û Ö Û Ú ¬Ò   ´Õ µ Ò ÓÖ Ö ØÓ ÓÑÔ Ø Ø ÒÓØ Ø ÓÒº × ÑÔÐ Ò×Ô Ø ÓÒ × ÓÛ× Ø Ø Ø ÙÒ ÕÙ ÕÙ Ð Ö ÙÑ Ó Ø ×Ý×Ø Ñ ´¾¾µ¸ ´¾¿µ × Õ ¼¸ ¼ Ò Õ ¼º ÆÓÛ Û ÔÖÓ ØÓ Ò ÐÝÞ Ø ×Ø Ð ØÝ Ó Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ¸ ÓÖ Ø × Û Ù× Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ò Ø ½ ½ ½ ½ ¼ ξ ´Õ Õ µ Õ Õ ·Í  Í  Õ · Õ ÃÔÕ · ´  Ã Õ · µ à  ½ ´  Ã Õ · µ · Õ Õ ¾ ¾ ¾ ´¾ µ Û Ö Û Ú ÖÓÔÔ Ø Ö ÙÑ ÒØ× Ò ¬Ò Í Í ´Õ µ¸ ´Õ µ ØÓ × ÑÔÐ Ý Ø ÒÓØ Ø ÓÒº ÁØ × ÛÓÖØ Ñ ÒØ ÓÒ Ò Ø Ø × ÔÓ ÒØ Ø Ø¸ ÄÝ ÔÙÒÓÚ Ò Ø ÙÒ Ø ÓÒ× Û Ø ÖÓ×× .2.

GAS of robot manipulators with linear PID and PI¾ D control Ø ÖÑ× × Î¾ Ú Ò Û ÐÝ Ù× Ò Ø Ð Ø Ö ØÙÖ ×Ø ÖØ Ò ÔÖÓ ÐÝ Û Ø ½¾ ´× Ð×Ó ¾¿¸ ½½¸ ¾¸ ½ Ò Ö Ö Ò × Ø Ö Òµº Ï ¬Ò Ø ÓÒÚ Ò ÒØ ØÓ Ø × ÔÓ ÒØ ØÓ ×ÔÐ Ø Ø Ò Ø ¸ Ò Ô ÖØ Ó Ø ÔÓØ ÒØ Ð Ò Ö Ý Ø ÖÑ× × ¼ Õ ÃÔ Õ Õ ´Õ µÕ ´ ´ · ½· ½ · ¾· ¾ ¿ µÕ ¿ µÕ ¼ ÃÔ Õ ´Õ µÕ ÛØ ½ ¼¸ ½ ¾ ¿º Ì Ò ÓÒ Ô ¼ Ò × ÓÛ Ø Ø Ñ Ü ¾ ½ Å Ñ ½ ¾ ´¾ µ Ø ÒØ ÙÒ Ø ÓÒ Î¾ ´Õ Õ µ × Ø ×¬ × Ø Î¾ ´Õ Õ ÐÓÛ Ö ÓÙÒ ¾ ¿ ¼ Õ ÃÔ Õ µ · ¾ · ¾ ¿ Õ Õ ´¾ µ Ò Ø × ÔÓ× Ø Ú ¬Ò Ø Ò Ö Ø Ò Ö Ý Ø ÖÑ× Û ÐÐ ÓÑ ÑÓÖ ´ µ Ò ÐÐÝ ÙÒ ÓÙÒ º Ì ÑÓØ Ú Ø ÓÒ ÓÖ Ø × Ô ÖØ Ø ÓÒ Ò Ó Ú ÒØ Ò Ø × Õ٠к Æ Üظ Ù× Ò Ø Û ÐÐ ÒÓÛÒ ÓÙÒ × ´Õ µ   ´Õ µ Õ ´¾ µ ØÓÖ × Ó ´¾¾µ¸ ´¾¿µ × ÓÙÒ Û Ó Ø ÒØ ØØ Ý Î¾ ´Õ Õ µ ØÑ ÖÚ ØÚ Ó Î¾ ´Õ Õ µ ÐÓÒ Ø ¾ ØÖ   Ñ ¾ Å   Ñ Ô ¼ Õ   Å Õ   µ Ô ¼ Ñ   ½ ¾ Å Õ ¾ ´¾ µ Û ×Ò ØÚ × Ñ ¬Ò Ø ÓÖ Ò×Ø Ò ´ Å ·¾ ½ · ¾ Å ¾ Å ´¾ µ ´¿¼µ ´¿½µ Ñ Õ Ñ ÄÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÓÖ Ò Ü ¼ ÓÐÐÓÛ× Ù× Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò ÔÖ Ò ÔÐ º ÙÖØ ÖÑÓÖ ÓÒ Ò ¬Ò ÓÑ Ò Ó ØØÖ Ø ÓÒ ÓÖ Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ ´¾¾µ¸ ´¾¿µ × ÓÐÐÓÛ׺ ¬Ò Ø Ð Ú Ð × Ø Æ ¨ Ü ¾ ÁÊ¿Ò Î¾ ´Üµ Æ © ´¿¾µ Û Ö Æ × Ø Ð Ö ×Ø ÔÓ× Ø Ú ÓÒ×Ø ÒØ ×Ù Ø Ø Î¾ ´Üµ ¼ ÓÖ ÐÐ Ü ¾ Æ º Ë Ò Î¾ × Ö ÐÐÝ ÙÒ ÓÙÒ Ò ÔÓ× Ø Ú ¬Ò Ø ¸ Ò Î¾ ´Üµ ¼ ÓÖ ÐÐ Ü ¾ Æ ¸ Ø × Ð Ú Ð × Ø × ÔÓ× Ø Ú ÒÚ Ö ÒØ ´ º º Ü´¼µ ¾ Æ Ø Ò Ü´Øµ ¾ Æ ÓÖ ÐÐ Ø ¼µ Ò ÕÙ Ð ¬ × × ÓÑ Ò Ó ØØÖ Ø ÓÒ ÓÖ Üº ¤ .102 Chapter 8.

2. Preliminary results 103 ¾º¾ Ë ÓÒ × ÙÒÑ ×ÙÖ Ð Ú ÐÓ Ø × ÁÒ Ø × × Ø ÓÒ Û Ö ­Ý ÔÖ × ÒØ ×ÓÑ × Ñ Ð Ö Ö ×ÙÐØ× ØÓ Ø Ó× ÓÒØ Ò Ò ÈÖÓÔÓ× Ø ÓÒ× ¾º½ Ò ¾º¾ ÓÖ Ø × Û Ò ÓÒÐÝ ÔÓ× Ø ÓÒ × Ú Ð Ð º Ö×ظ × ÓÒ Ø Ö ×ÙÐØ× Ó Ã ÐÐÝ ½½ ¸ ÓÒ× Ö Ø ÓÐÐÓÛ Ò ¾º¿ ÈÖÓÔÓ× Ø ÓÒº ÓÒ× ÖØ Ù Õ ÝÒ Ñ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø È ÓÒØÖÓÐ Ð Û ´¿¿µ ´¿ µ ´¿ µ  ÃÔ Õ   à ·   ´Õ · Õµ Õ · Õ ´Õ µ Û Ö ¸ ¸ Ã Ò ÃÔ Ö ÓÒ Ð ÔÓ× Ø Ú ¬Ò Ø Ñ ØÖ ×º Ì Ò¸ ÕÙ Ð Ö ÙÑ ÔÓ ÒØ ´Õ Õµ ´¼ ¼ Õ× µ Û Ö Õ× × Ø ×¬ × ´½ µ¸ Ó Ø ÐÓ× ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð º Ì Û ÛÖ Ø Ø ÈÖÓÓ º ¸ Ø ÔÑ ÐÓÓÔ ×Ý×Ø Ñ × £ ÔÖÓÓ Ò × ÐÝ Ú Ò ÐÓÒ Ø ÖÖÓÖ ÕÙ Ø ÓÒ ´½µ¸ ´¿¿µ¹´¿ µ × ÐÒ ×Ó Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º½º ¼ Ö×Ø ´¿ µ ´¿ µ ´Õ µÕ · ´Õ ÕµÕ · ´Õ µ   ´Õ µ · ÃÔÕ · à   · Õ Ì Ò¸ ÓÒ× ÖØ ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Î¿ ´Õ Õµ Ò Ø ½ ¾ à ν ´Õ Õ µ ·  ½ ´¿ µ Û × ÓÐ Õ × ÔÓ× Ø Ú ¬Ò Ø Ò Ö Õ ÓÐ ¼ ¼ Õ× ÔÑ ÐÐÝ ÙÒ ÓÙÒ ÛØ º ÁØ× Ø Ñ ÖÚ ØÚ Î¿ ´ µ ÐÓ Ð Ò ÙÒ ÕÙ Ñ Ò ÑÙÑ Ø ÐÓÒ Ø ØÖ ØÓÖ × Ó ´¿ µ¸ ´¿ µ   à  ½ ÔÖ Ò ÔÐ Ø Ò ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ ÓÐÐÓÛ× Ý ÒÚÓ Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò Ò Ù× Ò ×Ø Ò Ö Ö ÙÑ ÒØ׺ ¤ Ì ÔÖÓÔÓ× Ø ÓÒ ÓÚ Ò×ÙÖ × Ø ÐÓ Ð ×ÝÑÔØÓØ ÓÒÚ Ö Ò Ó Õ Õ× × Ø ½ Û Ö Õ× × Ø ×¬ × ´½ µº ÀÓÛ Ú Ö¸ × Ò Ø × Ó Ñ ×ÙÖ Ð Ú ÐÓ Ø ×¸ ÓÒ Ò ÐÑÒ Ø Ø ×Ø Ý ×Ø Ø ÖÖÓÖ Ý Ù× Ò ÈÁ ÓÒØÖÓк ÅÓÖ ÔÖ × Ðݸ Ò ½ Û ÒØÖÓ Ù Ø ÈÁ¾ ÓÒØÖÓÐÐ Ö Û ×Ø Ð × × × Ñ ¹ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Û Ø ÙÒ ÖØ Ò Ö Ú ØÝ ÒÓÛÐ º ÓÖ × ÑÔÐ ØÝ Ò ÓÖ Ø ÔÙÖÔÓ× × Ó Ø × Ô Ô Ö Û ÓÖÑÙÐ Ø ÐÓÛ ÔÖÓÔÓ× Ø ÓÒ Û ÓÐÐÓÛ× × ÓÖÓÐÐ ÖÝ Ó Ø Ñ Ò Ö ×ÙÐØ ÓÒØ Ò Ò ½ ´× Ð×Ó µº Ì Ö ×ÙÐØ ÐÓÛ Ù Ö ÒØ × ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð Øݺ ¾º ÈÖÓÔÓ× Ø ÓÒº ÓÒ× ÖØ Ù Õ ÖÓ ÓØ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø ¼ ÈÁ¾ ÓÒØÖÓÐ Ð Û ´¿ µ  ÃÔÕ   à ·  Ã ´Õ   µ   ´Õ · Õµ Õ · Õ ´¼µ ¾ ÁÊÒ .8.

GAS of robot manipulators with linear PID and PI¾ D control Ä Ø ÃÔ Ã Ã Ò ÔÓ× Ø Ú ¬Ò Ø ÓÒ Ð Ñ ØÖ × Û Ö × ×Ù Ø Ø ´Õ µ ´Õ µ ¼º ÍÒ Ö Ø × ÓÒ Ø ÓÒ׸ Û Ò ÐÛ Ý× ¬Ò ×ÙÆ ÒØÐÝ Ð Ö ÔÖÓÔÓÖ¹ Ø ÓÒ Ð Ò ÃÔ ´ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ Ã µ ×Ù Ø Ø Ø ÕÙ Ð Ö ÙÑ ÓÐ Õ Õ ¼ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð º £ ÐÓÛ Û Ú Ò ÓÙØÐ Ò Ó Ø ÔÖÓÓ ÔÖÓÔÓ× Ò ½ ¸Û Û Û ÐÐ Ù× × ÕÙ Ð ÓÖ ÓÙÖ Ñ Ò Ö ×ÙÐØ׺ Ö×ظ Ø ÖÖÓÖ ÕÙ Ø ÓÒ ´½µ¸ ´¿ µ Ò ÛÖ ØØ Ò × ÈÖÓÓ º ÒØ ´ ¼µ ´ ½µ ´ ¾µ ¼ ´Õ µÕ · ´Õ ÕµÕ · ´Õ µ   ´Õ µ · ÃÔÕ · à  Ã ´Õ   µ   · Õ   ½ÃÕ ¼ Û Ö ÃÔ × ¬Ò Ý ´¾½µº ÒÓÛ Ø Ø Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ò Ø ½ ¾ ÖÓÑ ½ Û Î ´Õ Õ µ ξ ´Õ Õ µ · à  ½   ´Õ µÕ × ÔÓ× Ø Ú ¬Ò Ø Ò Ö ÐÐÝ ÙÒ ÓÙÒ ÛØ ÐÓ Ð Ò ÙÒ ÕÙ Ñ Ò ÑÙÑ Ø Ø ÓÖ Ò × ×ÙÆ ÒØÐÝ ×Ñ Ðк ÓÖ Ø × Ó ÓÑÔÐ Ø Ò ×× Û Ö ÛÖ Ø Ø ÓÒ Ø ÓÒ× Ö Ú Ò ½ ÛØ ×Ð Ø ÑÓ ¬ Ø ÓÒ ÓÒÚ Ò ÒØ ÓÖ Ø ÔÙÖÔÓ× Ó Ø × Ô Ô Öº Ä Ø Ù× Ô ÖØ Ø ÓÒ Ø Ø ÖÑ Ã  ½ ´ ½ · ¾ µ à  ½ Û Ö ¼ ½¸ ¼ÛØ ½ ¾º Ï Ø ½· ¾ Ø × ¬Ò Ø ÓÒ׸ ÓÒ Ò ÔÖÓÚ Ø Ø ´¾ µ ÓÐ × Ò ¾ Ñ Ø Ò Î ´Õ Õ µ × Ø ×¬ × Ø ÓÙÒ ¾ ¿ ¾ ¾ ½ ¾ Å Å ´ ¿µ Î ´Õ Õ µ ÙÖØ ÖÑÓÖ ¸ Ø × Ø×Ý Ò × Ð×Ó Õ Õ· ¾ ¿ ¼ Õ ÃÔÕ · Ø Ñ Ñ ¾ ½ à  ½ ´ µ ¬ÐØ Ö ÓÙØÔÙØ ´ µ µ Ñ Ò × ÓÛÒ Ò ½ Ø Ø · Õ ÔÓ× Ø ÓÒ ÖÖÓÖ Õ Ò Ø ¾ ¼ × ×ÙÆ ÒØÐÝ ×Ñ ÐÐ ØÓ × Ø × Ý ¼ ´ ÔÑ   µ Ñ Ñ ÑÒ ¢ ¼ ¾ Å ÔÅ · Å · ´ £¾ ¾ Ñ Ñ Ñ ¾ Ñ Å ¾ Ñ Å Å ØÑ ´ µ ÖÚ ØÚ Ó Î ´ µ Ø Ò Ø Ö Ü ×Ø ×ØÖ ØÐÝ ÔÓ× Ø Ú ÓÒ×Ø ÒØ× ¬½ ¸ ¬¾ ¸ Ò ¬¿ ×Ù ÐÓÒ Ø ÐÓ× ÐÓÓÔ ØÖ ØÓÖ × ´ ¼µ¸ ´ ½µ × ÓÙÒ Ý Ø ØØ ¾ Î ´Õ Õ µ  ¬½ Õ ¾   ¬¾ Õ ¾   ¬¿ Ë Ò Î × ÔÓ× Ø Ú ¬Ò Ø Ò Î × ÐÓ ÐÐÝ Ò Ø Ú × Ñ ¬Ò Ø ¸ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó ¼ Ò ÔÖÓÚ Ò Ý ÒÚÓ Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò ÔÖ Ò ÔÐ Ò Ù× Ò ×Ø Ò Ö Ö ÙÑ ÒØ׺ ÙÖØ ÖÑÓÖ ÓÑ Ò Ó ØØÖ Ø ÓÒ ÓÖ ×Ý×Ø Ñ ´ ¼µ¸ ´ ¾µ Û Ø ×Ø Ø .104 Chapter 8.

8.3. Main results

105

×Ø

ÓÐ Õ Õ

Ò

¬Ò

× Ñ Ð ÖÐÝ × Ò Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾ × Ø ¨ © ¾ ÁÊ Ò Î ´ µ

ÐÚÐ ´ µ ÔÖÓÓ Ò

Û Ö × Ø Ð Ö ×Ø ÔÓ× Ø Ú ÓÒ×Ø ÒØ ×Ù Ø Ø Î ´ µ ¼ ÓÖ ÐÐ ¬Ò × × Ù× Ò × Ñ Ð Ö Ö ÙÑ ÒØ× × Ò Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾º ÈÁ¾ Ï Ö Ö Ý ÒÓÛ ØÓ ÔÖ × ÒØ ÓÙÖ Ñ Ò Ö ×ÙÐØ× ÓÒØÖÓк ÐÓ Ð ×ÝÑÔØÓØ ×Ø

¾

º Ì

¤

Ð ØÝ Û Ø ÈÁ

¿ Å Ò Ö ×ÙÐØ×
ÁÒ Ø × × Ø ÓÒ Û ÔÖ × ÒØ ÓÙÖ Ñ Ò Ö ×ÙÐØ׸ Û Ð Ò× ÓÒ Ø Ö ×ÙÐØ× Ö Ú Ò ÈÖÓÔÓ× Ø ÓÒ× ¾º½ ¹ ¾º Ò Ø ÓÑÔÓ× Ø ÓÒØÖÓÐ ÔÔÖÓ ÔÖÓÔÓ× Ò ½¿ º Ï × ÓÛ Ø Ø ÓÒ Ò Ú ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Û Ø ÈÁ Ò ÈÁ¾ ÓÒØÖÓÐ Ý × ÑÔÐÝ Ð Ý Ò Ø ÒØ Ö Ð Ø ÓÒº
¿º½ Ö×Ø × Ñ ×ÙÖ Ð Ú ÐÓ Ø ×

¿º½ ÈÖÓÔÓ× Ø ÓÒº

ÓÒØÖÓÐ Ð Û

ÓÒ×
Ù

ÖØ

ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ ÑÓ Ð ´½µ Ò ÐÓ×

ÐÓÓÔ Û Ø Ø

ÈÁ

 ÃÔÕ   Ã Õ ·  Ã
´Õ µ ÊØ

´Øµ

´Øµ

¼
Ø

´ µ
Ø Ø× Ø×

Õ ´×µ × · ´Õ µ Ø×

´ ¼µ

Û Ö ÃÔ¸ à ¸ Ò Ã Ö ÓÒ Ð ÔÓ× Ø Ú ¬Ò Ø Ñ ØÖ ×º Ì Ö ÐÛ Ý× Ü ×Ø ¬Ò Ø Ø Ñ Ò×Ø ÒØ Ø× ¼¸ ×ÙÆ ÒØÐÝ Ð Ö ÔÖÓÔÓÖØ ÓÒ Ð Ò ÃÔ Ò »ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ ÒØ Ö Ð Ò Ã ¸ ×Ù Ø Ø Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Ø Ø ÓÖ Ò Ü ÓÐ Õ Õ ¼º £

ÊÓÙ ÐÝ ×Ô Ò ¸ Ò Ø× ¬Ö×Ø Ô × ´Ø Ø × ¼ Ø Ø× µ¸ Ø Ð Ý ÈÁ Ó ÈÖÓÔÓ× Ø ÓÒ ¿º½ ÓÐÐ Ô× × ØÓ Ø ÖÓ Ù×Ø ÓÒØÖÓÐÐ Ö Ó ÈÖÓÔÓ× Ø ÓÒ ¾º½ Û Ù Ö ÒØ × ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó « Ö ÒØ ÕÙ Ð Ö ÙÑ Ø Ò × Ö ÙØ Ø Ð×Ó Ù Ö ÒØ × Ø Ø Ø ×Ø Ý ÔÓ× Ø ÓÒ ÖÖÓÖ × ÓÒ¬Ò ØÓ Ø ÐÓ× ÐÐ Ó Ö Ù× Ø ÖÑ Ò Ý ´½ µº ÁÒ Ø× × ÓÒ Ô × ´Ø Ø ×¸ ÓÖ ÐÐ Ø Ø× µ¸ Ø Ð Ý ÈÁ ÓÐÐ Ô× × ØÓ Ø ÓÒÚ ÒØ ÓÒ Ð ÈÁ ÓÒØÖÓÐÐ Ö Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾ Û Ø Ò Ø Ð ÓÒ Ø ÓÒ× Ü¼ Ü´Ø× µº ÖÓÑ Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾ Û ÒÓÛ Ø Ø Ø Ò Ø Ð ÓÒ Ø ÓÒ× Ü¼ Ö ×Ñ ÐÐ ÒÓÙ Ø Ò Ø ØÖ ØÓÖ × Ü´Øµ Ö ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð ´ Ò Õ ´Øµ Õ × Ø ½µº Ì Ù× Ø Ñ Ò Æ ÙÐØÝ Ò Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¿º½ × ØÓ × ÓÛ Ø Ø Ø Ö Ü ×Ø ×Ù Ø Ð Ò× ÃÔ Ò Ã ×Ù Ø Ø Ø ÓÙÒ × Ø Ó ÓÒÚ Ö Ò ¬Ò ÓÖ Ø ¬Ö×Ø Ô × × ÓÒØ Ò Ò Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ ¬Ò ÓÖ Ø × ÓÒ Ô × º Ì Ð ØØ Ö ÑÔÐ × Ø Ø¸ ÓÖ Ø ÒØ Ö Ð Ø ÓÒ × Ò ÓÖÔÓÖ Ø ´ º º ÓÖ ÐÐ Ø Ø× µ¸ Ø Ð Ý ÈÁ Ö Ú × Ø Ò Ö Ð Þ ÔÓ× Ø ÓÒ× Ò Ú ÐÓ Ø × ÒØÓ Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ ÖÓÑ Ø × Ø Ü ×Ø Ò Ó ¬Ò Ø ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× ØÓ Ù Ö ÒØ Æ Ò ¬Ò Ø Ø Ñ º ˸ ÓÐÐÓÛ׺ Ï ÔÖÓÚ ÐÓÛ Ø Ø Ø × × Ø ×º ½¼

106

Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control

ÁØ × ÑÔÓÖØ ÒØ ØÓ Ö Ñ Ö Ø Ø Ú Ò Ø ÓÙ Ø ´ Ð Ý µ ÈÁ ÓÒØÖÓÐÐ Ö ÔÖÓÔÓ× ÓÚ × ÒÓØ ×ÑÓÓØ ¸ Ø × ÓÒØ ÒÙÓÙ׺ Ì × Ô Ò × Ó ÓÙÖ× ÓÒ Ø ÓÖÖ Ø × ØØ Ò Ó Ø Ò Ø Ð ÓÒ Ø ÓÒ× Ó Ø ÒØ Ö ØÓÖ¸ Ø Ø × ´Ø× µ ´Õ µº
¿º¾ Ê Ñ Ö º ÈÖÓÓ

Ø Ð Ý ÈÁ ¸ ´Õ Õ µ ÙÔÔ Ö ÓÙÒ ´½ µº ¬Ò Ø  

Ó

ÈÖÓÔÓ× Ø ÓÒ ¿º½

º ÖÓÑ ÈÖÓÔÓ× Ø ÓÒ ¾º½ Ø ÓÐÐÓÛ× Ø Ø ÙÖ Ò Ø ¬Ö×Ø Ô × Ó ´Õ× ¼ ´Õ µ   ´Õ µµ × Ø ½º ÙÖØ ÖÑÓÖ ¸ Õ× × Ø ×¬ × Ø ×Ø

Ü

¾ ÁÊ¿Ò

Õ

¾
Ô

Ú

Ñ

Õ

¼

¾

Ú

ÖÓÑ Ø × Ù×× ÓÒ ÓÚ ¸ Û ÑÙ×Ø ¬Ò ÓÒ×Ø ÒØ Æ Ð Ö ÒÓÙ ×Ó Ø Ø   Ö Æ Û ¬Ò Ò ´¿¾µ¸ Ò ÓÖØ ×Ù Ø Ð Ø Ñ ÑÓÑ ÒØ Ø× ØÓ Ù Ö ÒØ Ë Ó Ø ÐÓ× Æ × ÐÓÓÔº ÆÓØ Ø Ø Ò ÓÖ Ö ØÓ Ú Ò ÜÔÐ Ø Ú ÐÙ ØÓ Æ Ò Ø ÖÑ× Ó Ø ÓÒØÖÓÐ Ò׸ ξ ´Üµ ×Ò ¸ ÓÛ Ú Ö Ø ÔÓØ ÒØ Ð Ò Ö Ý Ø ÖÑ Í ´Õ µ × ÒÓØ ÒÓÛÒ ÜÔÐ ØÐݺ Ì Ö ÓÖ ¸ Ð Ø Ù× ¬Ò ½ ½ ¾ ¾ ξŠ´Õ Õ µ Õ Õ · ´ ÔÅ · µ Õ · Õ · · Õ Õ ´ ½µ ¾ ¾ ¾ Ñ Ò Ø Ð ÚÐ× Ø
Å Æ

¨

Ü

¾ ÁÊ¿Ò

ξŠ´Üµ
Å Æ

Æ

©

ÁØ × ÒÓØ Æ ÙÐØ ØÓ × Ø Ø Î¾Å ´Üµ ξ ´Üµ Ò Å   Ø Ø ÓÙÖ ÓÖÒ Ö× Ó Ø Æ ¸ Ø ×ÙÆ × Ø Æ Ù× Ò ´ ½µ Ò ´½ µ Ø × ×ÙÆ ÒØ Ø Ø
Æ

ÔÐ Ò

Æ

º ÆÓÛ Û ÐÓÓ ÓÖ Æ ×Ù Å   ÓÒØ Ò Ò Æ

Ø Ø Ò ¸

¾ ½ ¾ Ú ¾ ¾ ¾ ´ ÔÅ · µ · Ú· Ú ´ ¾µ ¾ ÔÑ ÔÑ Ñ ÁÒ ÛÓÖ ×¸ Ø ÐÓÛ Ö¹ ÓÙÒ ÓÒ Æ Ú Ò ÓÚ ¸ Ò×ÙÖ × Ø Ø Ø Ð Ý ÈÁ ÓÒØÖÓÐÐ Ö Ò Ø× ¬Ö×Ø Ô × Û ÐÐ Ö Ú Ø ØÖ ØÓÖ × ÒØÓ Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ Æ Ò ¬Ò Ø Ø Ñ º Ì × ÓÒ Ö ÕÙ Ö Ñ ÒØ ÓÒ Æ × Ø Ø Î¾ Ò Ø Ú × Ñ ¹ ¬Ò Ø ÓÖ ÐÐ Ü ¾ Æ ¸ Ò Û ÔÖÓ ØÓ Ð ÙÐ Ø Ò ÙÔÔ Ö ÓÙÒ ÓÖ Æ ×Ó Ø Ø Î¾ ´ Æ µ ¼º ÖÓÑ Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾ ´× ´¾ µµ Û ÒÓÛ Ø Ø ´¾ µ ÑÔÐ × Ø Ø Î¾ ´Üµ Î¾Ñ ´Üµ Û Ö Û ¬Ò ¿ ¾ Î¾Ñ ´Üµ Õ ¾ ÔÑ ¨ © Ñ Ñ Ü ¾ ÁÊ¿Ò Î¾Ñ ´Üµ Æ º ÏØ Ø × ¬Ò Ø ÓÒ× Û Ú Ø Ø Æ ¬Ò Ø × Ø Æ Æ Ñ Ò Ø ×ÙÆ × ØÓ ÔÖÓÚ Ø Ø Î¾ ´ Æ µ ¼º ÆÓØ Ø Ø ÑÓÒ Ø Ø Ö ×ÙÆ ÒØ ÓÒ Ø ÓÒ× ´¾ µ¹´¿½µ ØÓ Ò×ÙÖ Î¾ ´Üµ ¼¸ Ø ÓÒÐÝ ÓÒ Û « Ø× Ø ¬Ò Ø ÓÒ Ó Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ ´ Ò Ó Æ µ × ´¿½µ Ø Ù׸ Ø × ÓÙÐ ÓÐ ØÖÙ Ø Ø
¼

¾Æ
¿ Ô
¼

¾

Ñ
¾

¾ ¾

Ñ

´ ¿µ

ÁÒ ×ÙÑÑ Öݸ Ö

ÐÐ Ò ´ ¾µ Ø × ×ÙÆ ÒØ Ø Ø Æ × Ø ×¬ × ½ ¾ Ú ´ · µ ¾ ÔÅ ÔÑ
¾

·

Ú Ô

Ñ

·

¾

Ú

¾

Ñ

Æ

¼ ¾ ¿ Ô ¾ ¾

Ñ Ñ

´ µ

½½

8.3. Main results

107

ØÓ Ò×ÙÖ Ø Ø Ø ØÖ ØÓÖ × Ü´Øµ ÓÒÚ Ö ØÓ Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ Æ Ò ¬Ò Ø Ø Ñ º Ò ÐÐݸ ØÓ Ò×ÙÖ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÓÖ Ò Ø ×ÙÆ × ØÓ ÓÓ× Ø Ø Ñ Ø× × Ø ¬Ö×Ø Ø Ñ ÑÓÑ ÒØ Û Ò Ø Ò Ø Ð ÓÒ Ø ÓÒ× Ü´Ø× µ ¾ Æ Ø Ø ×¸ Ø× Î¾ ´Ü´Ø× µµ Æ ÓÛ Ú Ö¸ × Ò Î¾ ´Üµ × ÒÓØ ÙÖ Ø ÐÝ ÒÓÛÒ ÓÒ× Ö Ø ÙÒ Ø ÓÒ
ξŠ´Üµ

½ Õ ¾

Õ

½ · ´ ÔÅ · ¾

µ

Õ

¾

·¾

Ú

Õ

·

Ú

Å

·

Õ

Õ

ÌÓ Ø × ÔÓ Òظ Û Ö ÐÐ Ø Ø ÓÖ ÐÐ Ø Ø× Ø ÈÁ ÓÒØÖÓÐÐ Ö × ÖÓ Ù×Ø È Û Ø ´Õ µ   ´Õ µ ÓÒ×Ø Òظ Ò ¾ Ú º ÖÓÑ Ø × Ø ÓÐÐÓÛ× Ø Ø Î¾Å ´Ü´Øµµ ξŠ´Ü´Øµµ ÓÖ ÐÐ Ø Ø× º Ì Ù׸ Ø ÔÖÓÓ × ÓÑÔÐ Ø Ý ¬Ò Ò Ø ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ ×
Ø× Î¾Å ´Ü´Ø× µµ Æ

´ µ Ò Ù ØÓ ´¾½µ ÓÖ ×ÙÆ ÒØÐÝ Ð Ö

Ò ÒÓØ Ò Ø Ø ´ µ ÓÐ × ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ ¸ ÒØÐÝ ×Ñ ÐÐ Å º ÔÑ Ò »ÓÖ ×ÙÆ

¤
ÔÖÓÓ × Ó ÈÖÓÔÓ¹ ÓÓ× Ø ÓÒØÖÓÐ ÓÒØÖÓÐ Ð Û ´ µ

ËÙÑÑ Ö Þ Ò ÐÐ ÓÒ Ø ÓÒ׸ Û Ö Û Ø ÓÐÐÓÛ Ò ÓÖÓÐÐ ÖÝ ÖÓÑ Ø × Ø ÓÒ× ¾º¾ Ò ¿º½º Ì × Ú × Ò Ò× Ø ØÓ Ø ÔÖ Ø Ø ÓÒ Ö ÓÒ ÓÛ ØÓ Ò× Ò Ø ×Û Ø Ò Ø Ñ Ø× ØÓ Ù Ö ÒØ Ë Ó Ø ÓÖ Òº
¿º¿

´ µ¸ ´ ¼µº Ä Ø

ÓÖÓÐÐ Öݺ

ÃÔ ¸ Ã

ÓÒ×

ÖØ ¸ Ò

Ã

ÝÒ Ñ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø ÈÁ ÓÒ Ð ÔÓ× Ø Ú ¬Ò Ø Ñ ØÖ ×¸ × Ø × Ý Ò ´ Å ·¾ Ñ Ü
Å

Ñ
Ô
¼

µ
Å

Ñ

¾ ½

¾ ½

·

¾

Å
ÓÒ Ø ÓÒ׸ Ø

´ µ ÐÓ×

Ò ´ µº ¬Ò Ø ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× × Ò ´ µº ÍÒ Ö Ø × ÐÓÓÔ ×Ý×Ø Ñ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Ø Ø ÓÖ Òº

Ì ¬Ö×Ø ØÛÓ Ô ÖØ× Ó ÓÒ Ø ÓÒ ´ µ Ò×ÙÖ Ø Ø Î¾ ´Üµ Î¾Ñ ´Üµ¸ ξ ´Üµ × ÔÓ× Ø Ú ¬Ò Ø Ò Ö ÐÐÝ ÙÒ ÓÙÒ º Ì Ò¸ ÓÒ Ø ÓÒ ´ µ Ò Ø Ø Ö Ô ÖØ Ó ÓÒ Ø ÓÒ ´ µ ÑÔÐÝ Ø Ø Î¾ ´Üµ × Ò Ø Ú × Ñ ¹ ¬Ò Ø º Ì Ù× ÐÐ ×ÙÆ ÒØ ÓÒ Ø ÓÒ× Ö Ú Ò Ø ÔÖ Ú ÓÙ× ÔÖÓÓ × Ú Ò ÓÐÐ Ø Ò Ø ÓÖÓÐÐ ÖÝ ÓÚ º ÁÒ ÓÖ Ö ØÓ × Ø × Ý Ø Ñ ÓÒ Ñ Ý ÔÖÓ × ÓÐÐÓÛ× ½º È Ò ÒÝ Ò ÖÚ ØÚ Ò Ø ÒØ ÖÚ Ð ´¼ ½µ × Ø × Ý Ò ½ · ¾ · Ò× ÃÔ Ò Ã × Ø × Ý Ò ´ µ Ò ´ µº Ò Ò
Ã
¿

½ Ò
Æ

ÒÝ ÔÖÓÔÓÖØ ÓÒ Ð × Ø × Ý Ò ´ µº Á ÓÖ Ò Û Ø ´¾½µº
Ø×

¾º È ÒÝ ×Ñ ÐÐ ÒØ Ö Ð ÒÓظ Ø Ò Ô ×Ñ ÐÐ Ö

Ò Û Ø ÖØ Ö × Ø Ö Ð Ö Ö ÔÑ ÓÖ ×Ñ ÐÐ Ö Å Ò ÐÐݸ Ò× Ò

¿º Ê Ô Ø ×Ø Ô× ½ Ò ¾ ÙÒØ Ð ÐÐ ÓÒ Ø ÓÒ× Ö × Ø ×¬ º Ì Ù× ÐÐ ÓÒ Ø ÓÒ× Ò Ò Ø Ð ÓÒ Ø ÓÒ× Ü´¼µº × ÐÝ Ú Ö ¬ Ò Ø ÓÒØÖÓÐÐ Ö

¬Ò

× Ò ´ µº ÓÖ

ÓÑÔÙØ

ÒÝ

½¾

GAS of robot manipulators with linear PID and PI¾ D control ¿º¾ ¿º Ë ÓÒ × ÙÒÑ ÖØ ×ÙÖ Ð Ú ÐÓ Ø × ÈÖÓÔÓ× Ø ÓÒº ÓÒ× ÖÓ ÓØ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø ÈÁ¾ ÓÒØÖÓÐ Ð Û Ù Õ  ÃÔÕ   à ·   ´Õ · Õµ Õ · Õ ´Õ µ ´Øµ ¼ Ø Ø× Ø× ´ µ ´ µ ´ ¼µ ´ ½µ  Ã Ä Ø ÃÔ Ã ÊØ Ø× ´Õ ´×µ   ´×µµ × · ´Õ µ Ø ÓÒ Ð Ñ ØÖ × Û Ö × ×Ù Ø Ø Ò ÐÛ Ý× ¬Ò ¬Ò Ø Ø Ñ Ò×Ø ÒØ Ø× ¼¸ ×ÙÆ ÒØÐÝ Ð Ö Ò× ÃÔ¸ Ò »ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ ÒØ Ö Ð Ò Ã ×Ù Ø Ø Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Ø Ø ÓÖ Ò ÓÐ Õ Õ ¼º £ ´Õ µ ´Õ µ à ¼ºÍÒ Ö Ø × ÓÒ Ø ÓÒ׸ Û Ò ÔÓ× Ø Ú ¬Ò Ø Ì ÔÖÓÓ ÓÐÐÓÛ× ÐÓÒ Ø Ð Ò × Ó Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¿º½¸ Ö ×ÙÐØ× Ó Ø Ò Ò ÈÖÓÔÓ× Ø ÓÒ× ¾º¿ Ò ¾º º Ï ×Ø ÖØ Ý ¬Ò Ò Ø × Ø ÈÖÓÓ º × ÓÒ Ø  ¼ Ò Û Ö ÒÓØ Ò Ø Ð ÚÐ× Ø ¾ ÁÊ Ò Å Õ ¾ Ô Ú Ñ Õ ¼ © ¾ Ú ¨ ¾ ÁÊ Ò Î Å ´µ ½ à  ½   Õ ¾ ÆÓØ ÖÓÑ Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º Ø Ø Î Å ´ µ Î ´ µ¸ Ò Å Ø Ø Î Å ´ ¼ µ ξŠ´ µ Ò  ¼ × Ø ×¬ × × Ñ Ð Ö ÓÙÒ ØÓ ¬Ò Ò ÙÔÔ Ö ÓÙÒ ÓÖ Û Ò×ÙÖ × Ø Ø Î ´ µ ¼º Ä Ø Î Å ´µ ξŠ´Üµ · Å º ÆÓØ Ð×Ó × ´ ¾µº Ï ÓÒÐÝ Ò Î Ñ´ µ Î¾Ñ ´Üµ · ¾ ½ Ñ Å ¾ ´ ¾µ Ø Ò ÖÓÑ ´ µ Û Ú Ø Ø Î ´ µ Î Ñ ´ µ ÓÒ Ø ÓÒ ´ ¿µ Ò ´¾ µ ÓÐ º ÓÒ× Ö Ò ÜØ ÓÒ Ø ÓÒ ×Ø Ð × Ý Ò ÕÙ Ð ØÝ ´ µ¸ Ø Ò Ò ÐÓ ÓÙ×ÐÝ ØÓ ´ ¿µ Û Ú Ø Ø Ñ Ü ¾ ¿ Ô ¼ ¾ Ñ ½ Å Ñ ½ ¾ ¾ Ñ Ñ ¾ Ò ´ µ ÑÔÐÝ Ø Ø Î ´ µ ¼ ÓÖ ÐÐ ×Ù ×ÙÑÑ Öݸ Ø × ×ÙÆ ÒØ Ø Ø × Ø ×¬ × Ø Ø Î Ñ´ µ ¸ Ò Ð×Ó ÓÖ ÐÐ ¾ º ÁÒ ¼ ¾ ¾ ¾ ½ ¾ Ú ¾ ¾ ¾ ¿ ÔÑ ½ Ñ Ñ Ñ ´ ÔÅ · µ · Ú· Ú ÑÒ ´ ¿µ ¾ ½ ¾ ¾ ¾Å ÔÑ ÔÑ Ñ ØÓ Ò×ÙÖ Ø Ø Ø Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Ö Ò Ø× ¬Ö×Ø Ô × Ö Ú × Ø ØÖ ØÓÖ × ´Øµ ÒØÓ Ø ÓÑ Ò Ó ØØÖ Ø ÓÒ ¬Ò ÓÖ Ø × ÓÒ Ô × º À Ò Ø Ö Ü ×Ø× ¬Ò Ø Ø× ¼ ½¿ .108 Chapter 8.

4.8. Discussion 109 Ò×ÙÖ Ò Ë Ó Ø ÓÖ Ò ¼º × Ò Ø ÐÐ Ø Ø× ¸ Ø Ö Ú ØÝ ÓÑÔ Ò× Ø ÓÒ ÖÖÓÖ Ò Ó× Ò × Ø× Î ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¿º½¸ ÓÒ× Ö Ò Ø Ø ÓÖ × ÓÒ×Ø ÒØ ÓÙÒ Ý ¾ Ú ¸ Ø Ò×Ø ÒØ Ø× ´ µ Å ´Ü´Ø×µµ Õ · Û Ö Î Å ´Üµ Ò ½ Õ ¾ Õ· ´ ½ ¾ ÔÅ · µ Õ ¾ ·¾ Ú Ñ Ú · Õ Õ· ½ ¾ à  ½   Õ Ì ÔÖÓÓ ¬Ò × × ÒÓØ Ò Ø Ø ´ ¿µ ÓÐ × ÓÖ ×ÙÆ ÒØÐÝ Ð Ö Ñ Ò ×ÙÆ Ù ØÓ ´¾½µ ÓÖ ×ÙÆ ÒØÐÝ Ð Ö ÔÑ Ò »ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ Å º Ê Ñ Ö º ÒØÐÝ ×Ñ ÐÐ ¸ ¤ ÆÓØ ÖÓÑ ´ µ Ø Ø Ø ×Û Ø Ò Ø Ñ Ø× Ó × Ô Ò Ò ÓÒ Ø ÙÒÑ ¹ Ú ÐÓ Ø × Õ´Ø× µº À Ò ¸ Ø ÔÖ × Ø ÓÖ Ø Ð Ö ×ÙÐØ Û × ÓÒØ Ò Ò ÈÖÓÔÓ× Ø ÓÒ ¿º × Ø Ø Ø Ö Ü ×Ø× ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× ×Ù Ø Ø Ø ÓÖ Ò ¼ × Ë º ÓÖ ÔÖ Ø Ð ÔÙÖÔÓ× × ÓÛ Ú Ö¸ Ó × ÖÚ Ø Ø Ø Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ× Ö ÒÓØ Ù× Ò Ø ÓÒØÖÓÐÐ Ö ÕÙ Ø ÓÒ× ´ µ ß ´ ½µº × Ø Ò × Ò ÖÓÑ Ø ÔÖÓÓ ÓÚ ¸ Ò ÔÖ Ø Ø ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× Ò ÓÑÔÙØ Û Ø ÒÓÛÐ Ó Ø ×Ø ×Ø Ñ Ø Ú Ð Ð Ó Ø Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ Ø ÔÖ × Ò×Ø Òغ ÓÖ Ò×Ø Ò ÒÝ Ø× ×Ù Ø Ø Î Å ´Ü´Ø× µµ Û Ö Û Ö ¬Ò ¿º ×ÙÖ Ð Î Å ´Ü´Ø×µµ ½ Õ ´Ø µ ¾ × Õ ´Ø× µ · ´ Ò Õ ´Ø× µ × Ø ×Ø Ñ Ø Ò ÑÓÑ ÒØ Ø× º ´ µ ×Ø ×Ø Ñ Ø Ú Ð Ð Ó Õ ´Ø× µ¸ Ø Ø ×¸ Ø Ø ÔÖ × Ò×Ø ÒØ Ø× º ËÙ ÓÑÔÙØ ÓÖ Ò×Ø Ò ÖÓÑ Ø Ð ×Ø ØÛÓ ÔÓ× Ø ÓÒ Ñ ×ÙÖ Ñ ÒØ× ÔÖ ÓÖ ØÓ Ø ÓÐÐÓÛ Ò ÓÖÓÐÐ ÖÝ ÖÓÑ ÈÖÓÔÓ× Ø ÓÒ× ¾º Ò ¿º º ½ ¾ ÔÅ · µ Õ ´Ø× µ ¾ ·¾ Ú Õ ´Ø× µ · Ñ Ú · Õ´Ø× µ Õ ´Ø× µ Ï ¬Ò ÐÐÝ Ö Û Ø ¿º ÓÖÓÐÐ Öݺ ÓÒ× Ö Ø ÝÒ Ñ ÑÓ Ð ´½µ Ò ÐÓ× ÐÓÓÔ Û Ø Ø ÈÁ¾ ÓÒØÖÓÐ Ð Û ´ µß´ ½µº Ä Ø ÃÔ ¸ à ¸ Ò Ã ÓÒ Ð ÔÓ× Ø Ú ¬Ò Ø Ñ ØÖ × Û Ø ÃÔ ¬Ò Ý ´¾½µ¸ × Ø × Ý Ò ´ ¿µ¸ ´ µ¸ Ò ´ ¿µº Ì Ö Ü ×Ø× Ø Ñ Ò×Ø ÒØ Ø× ´ ÓÖ Ò×Ø Ò ÚÒ Ý ´ µµ ×Ù Ø Ø Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð Ø Ø ÓÖ Ò ÓÐ Õ Õ ¼º ÓÖ ÔÖ Ø Ð ÔÔÐ Ø ÓÒ× ÓÛ Ú Ö ÓÒ Ñ Ý ÓÓ× Ø ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× ÓÖ Ò ØÓ ´ µº ÁØ × ÑÔÓÖØ ÒØ ØÓ Ö Ñ Ö Ø Ø Ø × Ñ ÐÓ Ð ×Ø Ð ØÝ Ö ×ÙÐØ× Ö ÔÓÖØ Ò ½ Ò Ú Ø × Ñ ÔÖ Ø Ð Ö Û Ø Ò Ø Ð ´ÙÒÑ ×ÙÖ Ð µ Ú ÐÓ ØÝ ÑÙ×Ø ÒÓÛÒº × Ù×× ÓÒ × Ø × Ð Ö ÒÓÛ ÖÓÑ Ø ÔÖÓÓ Ó ÈÖÓÔÓ× Ø ÓÒ ¾º¾¸ Û Ø ÑÔ × Ð Ñ Ò ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ ÓÖ ÈÁ ÓÒØÖÓÐÐ Ö × Ø ÔÖ × Ò Ó Ø Ù Ø ÖÑ Õ Õ ¾ Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ö Ú Ø Ú Î¾ º × Ñ ÒØ ÓÒ ÒØ ÒØÖÓ Ù Ø ÓÒ¸ Ø × Ø Ò Ð Æ ÙÐØÝ Ò ÓÚ Ö ÓÑ Ý Ñ Ò ×ÓÑ ×Ñ ÖØ ÑÓ ¬ Ø ÓÒ× ØÓ Ø ÈÁ ÓÒØÖÓÐ Ð Û¸ Ð Ò ÓÖ Ò×Ø Ò ¸ ØÓ Ø × Ò Ó ÒÓÒÐ Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö׺ ½ .

110 Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control ÌÓ Ø ×Ø Ó ÓÙÖ ÒÓÛÐ ¸ Ø ¬Ö×Ø ÒÓÒ¹Ð Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö× ÔÔ Ö Ò Ø Ð Ø Ö¹ ØÙÖ Ö Ò ¾ º ÁÒ Ø × × Ø ÓÒ Û × Ù×× Ø × ÓÒØÖÓÐÐ Ö× Ò × ÓÛ Ø Ø × Ñ Ò ÐÝ Ø × Ö ×ÙÐØ× ÒÒÓØ ×Ó × ÐÝ ÜØ Ò ØÓ Ø × Ó ÙÒÑ ×ÙÖ Ð Ú ÐÓ Ø ×º º½ Ì ÒÓÖÑ Ð Þ ÈÁ Ó Ã ÐÐÝ ÁÒ ÓÖ Ö ØÓ ÓÔ Û Ø Ø Ø Ú È ÓÒØÖÓÐÐ Ö Ù Ø ÖÑ Ù Õ Õ ¾ Ò ´¾ µ¸ à ÐÐÝ × ÔÖÓÔÓ× Ø Ô¹ ´ µ ØÓ Ø Ö Û Ø Ø ÙÔ Ø Ð Û  ÃÔ¼ Õ   Ã Õ · ¨´Õ µ   ½ ¨´Õ µ ­ Õ· ÆÓØ Ø Ø Ø Ñ ÒØ Ø ÓÒ ¼Õ ´ µ ½· Õ Û Ö ¼ ¼ × ×Ñ ÐÐ ÓÒ×Ø Òغ à ÐÐÝ ÔÖÓÚ Ø Ø Ø × ÔØ Ú ÓÒØÖÓÐÐ Ö Ò ÐÓ× ÐÓÓÔ Û Ø Ö ¹ Ó ÒØ ÖÓ ÓØ Ö ×ÙÐØ× Ò ÐÓ ÐÐÝ ÓÒÚ Ö ÒØ ×Ý×Ø Ñº ÀÓÛ Ú Ö¸ × Ò Ø Ö Ö ××ÓÖ Ú ØÓÖ ¨´Õ µ × ÓÒ×Ø ÒØ Ø ÙÔ Ø Ð Û ´ µ¸ ØÓ Ø Ö Û Ø Ø ÓÒØÖÓÐ ÒÔÙØ ´ µ Ò ÑÔÐ Ñ ÒØ × ÒÓÒÐ Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö Ý ÒØ Ö Ø Ò ÓÙØ Ø Ú ÐÓ Ø × Ú ØÓÖ ÖÓÑ ´ µ Ø ½   ­ ¨´Õ µ Õ · ½ ·¼ÕÕ · ´¼µ ´ µ Ó ÃÔ Ù ¼ ÃÔ · à ¸ ÛØ ¼ à ½ ­ ¨´Õ µ¨´Õ µ ¸ Ý Ð × Ø ´¼µ ÓÒØÖÓÐÐ Ö ÑÔÐ ¹ ´ µ ´ ¼µ  ÃÔÕ   Ã Õ ·   ¼ à ½ ·Õ Õ ¼ ¾ ÁÊÒ Ë Ò ÓÒØÖÓÐÐ Ö× ´ µ¸ ´ µ Ò ´ µ¸ ´ ¼µ Ö ÕÙ Ú Ð Òظ ÓÐÐÓÛ Ò Ø ×Ø Ô× Ó Ã ÐÐÝ ÓÒ Ò ÔÖÓÚ ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÐÓ× ÐÓÓÔ ×Ý×Ø Ñ ´½µ¸ ´ µß´ ¼µº Ú ÐÙ Ø Ò Ø Ø Ñ Ö Ú Ø Ú Ó Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ò Ø ½ ½ ¼ ½  ½ Î ´Õ Õ Þ µ ¾ Õ ´ÕµÕ · Í ´Õµ   Í ´Õ µ   Õ ´Õ µ· ¾ Õ ÃÔÕ · ¾ Þ Ã Þ · Õ ´ÕµÕ ´ ½µ Û Ö Û Ú ¬Ò Þ ½· Õ  Ã Õ · ¼ Å ¼ ´ ¾µ ´ ¿µ Õ ¾ Û Ó Ø ÒØ Ø Î ´Õ Õ Þ ´ µ ÓÛ Ú Ö¸ ÒÓØ Ø Ø Ø ÒÓÖÑ Ð Þ Ø ÖÑ Õ Õ ¾ Õ ¾ ¸ Ò Î ´Õ Õ Þ µ × Ò ¹ Ø Ú × Ñ ¹ ¬Ò Ø ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ ¼º ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ ÓÐÐÓÛ× Ý ÒÚÓ Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò ÔÖ Ò ÔÐ º ½ µ   Ñ   ½· Õ ¼ Õ     ¼¾Å   ½ ·¼ Õ ´ Ô ¼ Ñ    ¼¾Å µ Õ ¾ .

8.4. Discussion 111 º¾ Ì × ØÙÖ Ø ÈÁ ÓÒØÖÓÐÐ Ö Ó Ö ÑÓØÓ ¾ Ò ÐØ ÖÒ Ø Ú ØÖ ÈÁ ØÓ Ú Ù Ë ×Ø × Ñ Ó Ö ÑÓØÓ ¾ Û Ó ÔÖÓÔÓ× Ø ÒÓÒÐ Ò Ö ´ µ ´ µ Û Ö Ø × ØÙÖ Ø ÓÒ ¼ ÓÖ ÐÐ ¼¸ × Ø´¼µ ¼ Ò Ø × ÓÙÒ × × Ø´ µ ½º Ö ÑÓØÓ¿ ¾ ÔÖÓÚ Ø Ø ¸ Ò ÔÑ Ã × ×ÙÆ ÒØÐÝ ×Ñ Ðи Ø ÐÓ× ÐÓÓÔ × ÐÓ ÐÐÝ ×ÝÑÔØÓØ ÐÐÝ ×Ø Ð º Ì Ý Ù× Ò ¾ × ØÓ ÓÑ Ò Ø Ø Ù Ø ÖÑ× Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ö Ú Ø Ú Ý Ñ Ò× Ó Ø × ØÙÖ Ø ÔÖÓÔÓÖØ ÓÒ Ð Ò ´ µº ÅÓÖ ÔÖ × Ðݸ Ø Ò × ÐÝ ÔÖÓÚ Ò Ø Ø Ø Ø Ñ Ö Ú Ø Ú Ó Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ ½ ½ ¼ Î ´Õ Õ Þ µ Õ ´Õ µÕ ·Í ´Õ µ Í ´Õ µ Õ ´Õ µ· Õ ÃÔ Õ · Þ Ã  ½ Þ · × Ø´Õ µ ´Õ µÕ ´ µ ¾ ¾ ¾ Û Ö ¼ × ×Ñ ÐÐ ÓÒ×Ø ÒØ Ò Þ × ¬Ò Ý ½ Þ   ÃÕ ´ µ ´ µ¸ ÐÓÒ Ø ÐÓÓÔ ×Ý×Ø Ñ ´½µ¸ ´ µ¹´ µ × ÓÙÒ Ý ½ ¼ Î ´Õ Õ Þ µ  ´ Ñ   × Ø´Õ µ   Å   µ Õ ¾   ´ ÔÑ     µ × Ø´Õ µ¾ ´ µ ¾ Å ¾ Å ¾ ÓÛ Ú Ö¸ ÒÓØ Ø Ø Ø Ø ÖÑ × Ø´Õµ Õ ¾ Õ ¸ Ò Î ´Õ Õ Þ µ × Ò Ø Ú × Ñ ¹ ¬Ò Ø ÓÖ ×ÙÆ ÒØÐÝ ×Ñ ÐÐ Ò ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ ÓÐÐÓÛ× Ý Ó × ÖÚ Ò Ø Ø × Ò´ × Ø´Õ µµ × Ò´Õ µ Ò ÒÚÓ Ò ÃÖ ×ÓÚ× ¹Ä Ë ÐÐ ³× ÒÚ Ö Ò ÔÖ Ò ÔÐ º Ê Ñ Ö ×º ÙÒ Ø ÓÒ × Ø ÁÊÒ  ÃÔ¼ ×  Ã Õ ½ Ø´Õ µ   Ã Õ   Ã Õ · ´¼µ ÁÊÒ × Ø ×¬ × × Ø´ µ Ò ¼ ¾ ÁÊ ØÖ ØÓÖ × Ó Ø ÐÓ× ½º × × Ø Ö ÓÑÔÐ Ü Øݸ ÔÖ Ø Ò Û Ø Ö ×Ô Ø ØÓ Ø ÓÒØÖÓÐÐ ÜÔÖ ×× ÓÒ× Ó Ø ÄÝ ÔÙÒÓÚ Ö Ú Ð Ò Ö ÈÁ ÓÒØÖÓÐÐ Öº ÅÓÖ ÔÖ Ù× Ò Ø Ó× ÔÔÖÓ × Ð ÖÐÝ Ö ×Ô Ø ØÓ Õ º Ð Ö Û Ó Ø ÒÓÒÐ Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö× Ó ¾ Ö Ó ÈÖÓÔÓ× Ø ÓÒ ¿º½ × Ø Ø ÓÒ Ñ Ý ÜÔ Ø ÖÓÑ Ø Ø Ú × Î Ò Î ¸ Ø Ø Ø Ý ÓÒÚ Ö ×ÐÓÛ Ö Ø Ò ÓÙÖ × Ðݸ ÒÓØ Ø Ø Ø × ØÙÖ Ø ÓÒ Ò ÒÓÖÑ Ð Þ Ø ÓÒ ØØ ÒÙ Ø × Ø ÖÓÛØ Ö Ø Ó  Î Ò  Î Û Ø ¾º ÁÒ ÓÒØÖ ×Ø ØÓ Ø ×¸ ÓÙÖ ÔÔÖÓ Ù Ö ÒØ × ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Û Ø × ÑÔÐ Ð Ò Ö ÈÁ × Ø × Ù× Ò Ñ ÒÝ ÔÖ Ø Ð ÔÔÐ Ø ÓÒ׺ ÊÓÙ ÐÝ ×Ô Ò Ø Ù× Ö Ò ÔÔÐÝ × ÑÔÐ ÖÓ Ù×Ø È ÓÒØÖÓÐÐ Ö × Ø Ø Ó ÈÖÓÔÓ× Ø ÓÒ ¾º½ Ò ×Ø ÖØ Ø ÒØ Ö Ø ÓÒ « ØÛ ÒØ Ò Ö Ð Þ Ú ÐÓ Ø × Ò Ø ÔÓ× Ø ÓÒ× Ö ×Ñ Ðк ¿º ÖÓÑ Ø ÓÖ Ø Ð ÔÓ ÒØ Ó Ú Û¸ Ø ØÖ Ó ÒØÖÓ Ù Ò ÖÓ×× Ø ÖÑ× Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ × ÒÓØ Ò Û ½¾¸ ¾¿ ÓÛ Ú Ö¸ Ø Ó Ù× Ò × ØÙÖ Ø ÔÖÓÔÓÖØ ÓÒ Ð Ò ´ µ × Ù ¸ × Ö × Û ÒÓÛ¸ ØÓ ¾ º Ì × ØÖ Ò ÓÑ Ò Ø ÓÒ Û Ø Ø × ØÙÖ Ø ÖÓ×× Ø ÖÑ Ù× Ò Î Û Ö ÙÒ Ñ ÒØ Ð ØÓ ÔÖÓÚ Ëº Ì × Ñ Ó × ÖÚ Ø ÓÒ × Ú Ð ÓÖ Ø ÔÔÖÓ Ó Ã ÐÐÝ Û Ö Ø ÒÓÖÑ Ð Þ Ø ÓÒ ÔÐ Ý× ÖÙ Ð ÖÓÐ º ¿ ÁØ ÓÒ× × ÛÓÖØ Ö Ñ ÒØ ÓÒ Ò Ö ¸ ÓÛ Ú Ö Ø Ø Ø Ò ¾ ¸ Ö ÑÓØÓ Ù× ÙÒ Ñ ÒØ Ð × ØÙÖ Ø ÓÒ ÓÖ Ø Ú Ð ÙÒ Ø ÓÒ Û ØÝ Ó Ø × Ö ×ÙÐغ Ô ÖØ ÙÐ Ö × Ó × Ø × ÔÓ ÒØ × ÒÓØ ½ .

GAS of robot manipulators with linear PID and PI¾ D control º Ú Ò Ø ÓÙ Ø × ØÖ × Ò Æ ÒØÐÝ ÔÔÐ ØÓ ÓÙÒ Ø ÔÓ× Ø ÓÒ Ô Ò ÒØ ¾ ¾ Ù Ø ÖÑ× × Ø´Õ µÕ ¸ Ò ¼ ½· ÕÕ Õ ¸ Û Ò Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ× Ö ×ÙÔÔÓ× ÙÒ¹ Ú Ð Ð ¸ Ø × ÒÓØ ÔÓ×× Ð ØÓ ÔÔÐÝ Ø × Ñ ÔÔÖÓ ×º ÌÓ ÐÐÙ×ØÖ Ø Ø × ¸Ð Øß Û Ø ÓÙØ ÐÓ×× Ó Ò Ö Ð ØÝ ß × Ø´ µ Ø Ò ´ µ Ø Ò¸ ÓÒ Ù× × Ø × ØÙÖ Ø ÓÒ Ð Ö ¹ ÑÓØÓ Ò Ø ÈÁ¾ × Ñ ¸ × Ñ Ò ÐÝ Ø × ØÙÖ Ø ÖÓ×× Ø ÖÑ Ø Ò ´ µ ´Õ µÕ × ÓÙÐ Ù× Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ò Ø Î ¸ Ò×Ø Ó ´Õ µÕº ÀÓÛ Ú Ö¸ Ø × ­ ­ ¾ Ý Ð × Ø Ø ÖÑ   Ñ Ñ ­ × ¾ ´ µ­ Õ Ò Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ö Ú Ø Ú ¸ Ò×Ø Ó ­ ­ ¾ ­ × ¾ ´ µ­ Ú Ò × × ×   Ñ ÑÕ Ò × Ò ½ Ø ÄÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ Ö Ú Ø Ú Û ÐÐ ÐÓ ÐÐÝ Ò Ø Ú × Ñ ¬Ò Ø º Ë Ñ Ð Ö ÓÒ ÐÙ× ÓÒ× Ò Ö ÛÒ ÓÒ ØÖ × ØÓ Ù× Ø ÒÓÖÑ Ð Þ Ø ÓÒ Ù× Ò º ÓÖ Ø × Ö ×ÓÒ Ø Ö × ÒÓØ ÑÙ ÓÔ ØÓ ÜØ Ò Ø ÔÔÖÓ × Ó ¾ Ò ½½ ØÓ Ø ÓÙØÔÙØ × º ÓÖ Ø× × ÑÔÐ ØÝ Ò Ø Ö ÙÑ ÒØ× ÜÔÓ× Ò Ê Ñ Ö ¿º ¸ ÓÙÖ Ö ×ÙÐØ Ó ÓÖÓÐÐ ÖÝ ¿º × Ñ× ÑÓÖ ÔÖÓÑ × Ò ÓÖ ÔÖ Ø Ð ÔÔÐ Ø ÓÒ׺ Ë ÑÙÐ Ø ÓÒ Ö ×ÙÐØ× ÌÓ ÐÐÙ×ØÖ Ø Ø ÛÓÖ Ò Ó Ø ÓÒØÖÓÐÐ Ö× Ö Ú Ò Ø × Ô Ô Ö¸ × ÑÙÐ Ø ÓÒ× Ú Ò ÌÅ º Ï ÓÑÔ Ö Ø ÖÖ ÓÙØ Ù× Ò Ð Ý ÈÁ ÓÒØÖÓÐÐ Ö Ö Ú Ò × Ø ÓÒ ¿º½ Û Ø Ø ÒÓÖÑ Ð Þ ÈÁ Ó Ã ÐÐÝ Ò Ø × ØÙÖ Ø ÈÁ ÓÒØÖÓÐÐ Ö Ó Ö ÑÓØÓ ¾ º ÓÖ ÓÙÖ × ÑÙÐ Ø ÓÒ× Û Ù× Ø ÑÓ Ð ÔÖ × ÒØ Ò ¸ Û Ö Å ØÐ ´Õ µ ´Õ Õµ ´Õ µ · ½ ¼¾ Ó× Õ¾ ¼ · ¼ ½ Ó× Õ¾ · ¼ ½ Ó× Õ¾ ¼ ¾  Õ¾  ´Õ½ · Õ¾ µ ¼ ½ × Ò Õ¾ Õ½ ¼ × Ò Õ½ · ¼ ¿ × Ò´Õ½ · Õ¾ µ ½ ¼ ¿ × Ò´Õ½ · Õ¾ µ ¼ ¸ Å ¸ ½ ¿¸ Ú ¼ ¸ Ö Ú Ø Ø ÓÒ Ð ÓÖ × Ú ØÓÖ Ø Ò ´Õ µ ½ ¾º Ï ¼ ¼ º ¼ ÓÖ Ø × ×Ý×Ø Ñ Û Ú Ñ ¼ ØÓ Ú ÒÓ ØØ Ö ×Ø Ñ Ø Ó Ø ××ÙÑ Ï ÓÒ× Ö Ø ÔÖÓ Ð Ñ Ó ÓÒØÖÓÐÐ Ò Ø Ñ Ò ÔÙÐ ØÓÖ ÖÓÑ Ø ÔÓ× Ø ÓÒ ¾ ¼ ØÓ¹ Û Ö × ½ ½ º ÓÖ Ø × Û Ù× ÃÔ ¾ ¼Á ¸ à Á¸ à ½ ¼Á ¸ Û Ö ÃÔ ½¾¼Á º ÖÓÑ ´¾½µ Ø ÓÐÐÓÛ× Ø Ø ¯ ½ ¾ º ÖÓÑ ´¾ µ Û × Ø Ø Î¾ ¼ Ò Ý ÓÓ× Ò Æ ¾ ¼ Û Ñ Ø Å ´ ¾µ Ò Ö Ù Ö ÒØ ØÓ ÒØ Ö Ø × Ø Æ ¾ ÁÊ Ò Î¾Å ´Üµ Æ Ò Ø Ö ÓÖ Ø Ü ×Ø Ò Ó Ø× × ¬Ò Ò ´ µ × Ð×Ó Ù Ö ÒØ º Ý ÓÓ× Ò ½ ¼ ¸ ¾ ¼ ¾ Ò ¿ ¼ ½ Û × ÖÓÑ ´¾ µ Ø Ø Î¾ ´Õ Õ µ Ø Ø Õ Õ Õ ¾ ¸Û Õ ×Ó Ø Ø ÖÓÑ Î¾ ´Õ Õ µ Æ Û Ò ÓÒ Ð٠ξ Ö ×ÙÐØ× ÒØÓ  ¾ ¾ ¾ ½ ¾ ×Ø Ð ØÝ Ó Å Ú ×ÝÑÔØÓØ Ì Ö ÓÖ ¸ Û ×Ø ÖØ ÒØ Ö Ø Ò × ×ÓÓÒ × Û ÒØ Ö Æ Û Ø × ÓÒ Ô × Ò ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ø ÈÁ ÓÒØÖÓÐÐ Öº ½ .112 Chapter 8.

Simulation results 113 ÖÓÑ × ×Ø Ó Ð ØÝ Ó Ö ÑÓØÓ Ì Ø Ø ÓÒ Û ÒÓÛ Ø ÈÁ Ø ÓÙÖ × Ð Ó Ã ÐÐÝ Ø ÓÒ Ó ´Û Ù× Ò× Ð×Ó Ò Ù Ö ÒØ Ø × ÐÓ Ð ÈÁ ×ÝÑÔØÓØ ÓÒØÖÓÐÐ Ö ÒÓÖÑ Ð Þ ¾ º ¯¼ ½º ½µ × ØÙÖ Ø Ö ×ÙÐØ Ò Ô Ö ÓÖÑ Ò × Ô Ø Ò ÙÖ 1 PI D d Kelly Arimoto 0 PI D d Kelly Arimoto 0 2 4 t 40 60 40 0 −20 −40 0 2 4 t 0.5 1 0 −1 0 2 4 t 6 8 10 ~ q2 ~ q −0.8.2 0.1 0 0 1 2 3 4 5 t 6 7 8 PI D d Kelly Arimoto 9 10 6 ν2 ν 20 PIdD Kelly Arimoto 8 10 1 0.4 ise 0.5 6 8 10 30 20 10 0 0 2 4 t 6 PIdD Kelly Arimoto 8 10 ÙÖ ½ ÓÑÔ Ö Ø Ú ×ØÙ Ý Ï × Ø Ø Ø Ô ÖØ ÐÐÝ × ØÙÖ Ø Ö × Ø Ò ÔÖÓÔÓÖØ ÓÒ Ð Ø ÖÑ Ð × ØÙÖ Ø ÓÒ Ò Ã ÐÐÝ³× Ð Ý ÒØ × ØÓ ÓÒØÖÓÐÐ Ö Ð Ö Ð Ö ÓÚ Ö× × ØÓ Ø ÓÓØ ÓÖ Ö ¹ ÓÒ¹ ÑÓØÓ³× Ú Ö Ø Ò ÈÁ ÓÒØÖÓÐÐ Ö Ó ØÓ ¾ ¸ Û ×ÐÓÛ Ö ´Õ µº Ï Ð×Ó × Ø Ö Ø ÓÒ ´×Ø ÖØ Ò Ø× ¼ ¾ ¿¿µ Ó ÓÒØÖÓÐÐ Öº ÒÓØ ÓÒÐÝ ÐÓÓ ÕÙ Ð Ø Ø Ú Ø Ø ÙØ Ð×Ó ÕÙ ÒØ Ø Ø Ú ÓÑÔ Ö ×ÓÒ ØÛ Ò Ø Ø Ö ÌÓ Ñ ÓÒØÖÓÐÐ Ö׸ Û ÜÔÖ ×× ÓÒ Ø × ´Øµ ¼ ½ Õ ´×µ Õ ´×µ × .3 0.5.

5 q 0 5 10 t 15 20 25 0 ϑ2 0 5 10 t 15 20 25 1000 500 τ2 0 −500 0 5 10 t 15 20 25 −1000 0 5 10 t 15 20 25 −100 −200 0 5 10 t 15 20 25 ϑ τ 1 1 −100 −200 2000 0 −2000 −4000 ÙÖ ¾ Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Ö ´¬Ö×Ø ¾ × ÓÒ ×µ ÁÒ × Ó ÙÒÑ ×ÙÖ Ð Ú ÐÓ Ø × Û Ð×Ó ÓÒ× ÖÓÑ Ø ÔÓ× Ø ÓÒ ¾ ¼ ØÓÛ Ö × ½ ½ º ÓÖ Ø ½ Á¸ ¾¼¼Á º Ì ×Ñ ÐÐ Ö Ú ÐÙ ÓÖ Ã × Ù ØÓ Ø ÑÓÖ Ö ×ØÖ Ø Ú Ò ÕÙ Ð Ø ×º ¯ ¼ ¼½ ¿µ Û Ú Ø ØÎ ¼¸ Ò Ù× Ò ¼ ¼ ¸ ´ ¾µ ÓÑ × ¾ Î Ñ ´Õ Õ µ Õ ½ Ö Ø ÔÖÓ Ð Ñ Ó ÓÒØÖÓÐÐ Ò Ø Ñ Ò ÔÙÐ ØÓÖ × Û Ù× ÃÔ ¾ ¼Á ¸ à Á¸ à ¾Á ¸ Ò ÓÑÔ Ö ×ÓÒ Û Ø Ø ×Ø Ø ¹ × Ý ÓÓ× Ò ÃÔ ½¼¼Á ´Û Ö ×ÙÐØ× ÒØÓ ¼ ¾¸ ¾ ¼ ¼ ¸ ¿ ¼ ½¼¸ ½ ¼ ¸ ½ ¼ ¾ ·¼½ ½ ¾ ´ ¼µ . GAS of robot manipulators with linear PID and PI¾ D control Ì Ò Û × Ø Ø Ö ÑÓØÓ³× ÓÒØÖÓÐÐ Ö ¾ ÐÐ Ø Ø Ñ × Ø Ð Ö ×Ø × Ù ØÓ Ø Ô ÖØ ÐÐÝ × ØÙÖ Ø ÔÖÓÔÓÖØ ÓÒ Ð Ø ÖѺ Ï Ð×Ó × Ø Ø ÙÖ Ò Ø ¬Ö×Ø × ÓÒ ¸ Ø × Ó Ã ÐÐÝ³× ÓÒØÖÓÐÐ Ö × Ð ØØÐ Ø ÐÓÛ Ö Ø Ò Ø × Ó ÓÙÖ Ð Ý ÈÁ ÓÒØÖÓÐÐ Ö¸ ÓÛ Ú Ö¸ Ù ØÓ Ø × ØÙÖ Ø ÓÒ Ò Ø ÒØ Ö Ð Ô ÖØ Ø ¬Ò Ð ÓÒÚ Ö Ò Ó Ã ÐÐÝ³× ÓÒØÖÓÐÐ Ö × ×ÐÓÛ Ö¸ Ö ×ÙÐØ Ò ÒØÓ Ð Ö Ö × º 1 0 0 1 −1 −2 q2 0 5 10 t 0 15 20 25 −0.114 Chapter 8.5 −1 −1.

8.5 8 60 ν1 ν2 40 20 0 0 50 100 t 150 200 250 6 4 2 0 −2 0 50 100 t 150 200 250 0 50 100 t 150 200 250 q2 0 50 100 t 150 200 250 ÙÖ ¿ Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Ö ¾¼ .5.5 −1 −1. Simulation results 115 Ý Å ÓÓ× Ò ¾ Û Ñ ØØ Ð Ø Ò × Ò Ø Ö ÓÖ Ø Ü ×Ø Ò Ó Ø× × ¬Ò ÖÓÑ Î Ñ Û ÓÒ ÐÙ Ø Ø Õ · Ó ´ ¿µ Ò Ò ´ µº ¼ ¼ ¼¾  ¼ Ö Ù Ö ÒØ ØÓ ÒØ Ö Ø ×Ø ×Ó Ø Ø Î ¾   Õ Õ ¿ ¾ ¼¿ ¼  ½ ¼  ½ ¼  ¼ ¿¾ ¿ Õ Õ ¿ Ì Ö ÓÖ Û ×Ø ÖØ ÒØ Ö Ø Ò × ×ÓÓÒ × Ø × ÓÒ Ô × Ò ÐÓ Ð ×ÝÑÔØÓØ ×Ø × ÐÖ Ý ÔÓ ÒØ ÓÙØ Ò Ê Ñ Ö ¿º Ø ×Ò Û Ò Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ× ÓÖ Ø ÙÖ Ò Ø ¬Ö×Ø Ô × Û ÓÒÚ Ö ØÓ ¬Ü ÁÒ ÙÖ ¾ Û Ò× Ø Ú ÓÙÖ Ó ÙÐ Ø ÓÒº ÁØ Ò × ÒØ Ø ØØ ¾ Û Û ØÓ ×Ø ÖØ ÒØ Ö Ø Ò ÖÓÑ Ø ¾ Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Ö × Ô Ø Ò ÙÖ Ø Ø× ¾ Ö ×ÙÐØ× ÒØÓ Þ ÖÓ ÔÓ× Ø ÓÒ ÖÖÓÖº Û Ö Ò Å Û Ú ×ÝÑÔØÓØ ×Ø Ð ØÝ Ó Ð ØÝ Ó Ø Ð Ý ÈÁ¾ ÓÒØÖÓÐÐ Öº ÔÖÓ Ð Ñ × ØÓ Ø ÖÑ Ò Û Ò Û Ö Ò Å ¸ ÖÑ Ò Ò Ø ×º ÀÓÛ Ú Ö¸ Û Ö Ù Ö ÒØ Ø Ø ÔÓ ÒØ Ø Ø × ÓÒØ Ò Ò Å º Ø × Ò Ð× ÙÖ Ò Ø ¬Ö×Ø ¾ × ÓÒ × Ó × Ñ¹ Ú ÐÑÓ×Ø ÓÒÚ Ö ¬Ü ÔÓ Òغ Ì Ö ÓÖ ¸ ÓÒº Ì Ö ×ÙÐØ Ò ÓÚ Ö ÐÐ Ô Ö ÓÖÑ Ò Ó Ø ¿º Ï Ò × Ø Ø Ø ÒØ Ö Ø Ò Ø Ø ×Ø ÖØ 1 0 0 q1 −1 −2 −0.

116 Chapter 8. GAS of robot manipulators with linear PID and PI¾ D control ÓÒ ÐÙ Ò Ö Ñ Ö × Ï Ú Ö ×× Ø ÔÖ Ø ÐÐÝ ÑÔÓÖØ ÒØ ÔÖÓ Ð Ñ Ó ÐÓ Ð ×ÝÑÔØÓØ ×Ø Ð Þ Ø ÓÒ Ó ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ× Û Ø ÈÁ Ò ÈÁ¾ ÓÒØÖÓи º º Ø ÔÖÓ Ð Ñ Ó × Ø¹ÔÓ ÒØ ÓÒØÖÓÐ Û Ø ÙÒ ÖØ Ò Ö Ú ØÝ ÒÓÛÐ Ò ÓØ Ý ×Ø Ø Ò ÓÙØÔÙØ º ÇÙÖ Ñ Ò ÓÒØÖ ÙØ ÓÒ × Ø ÔÖÓÓ Ø Ø Ë × ÔÓ×× Ð Û Ø Ð Ò Ö ÈÁ Ò ÈÁ¾ ÓÒØÖÓÐÐ Ö× Ý × ÑÔÐÝ Ð Ý Ò Ø ÒØ Ö Ð Ø ÓÒº Ï Ú ÐÐ ÓÙÖ Ò Û ÓÒØÖÓÐÐ Ö× ÈÁ Ò ÈÁ¾ º ÖÓÑ Ø ÓÖ Ø Ð ÔÓ ÒØ Ó Ú Û Û Ú × ÓÛÒ ÓÖ ÓØ × × ×Ø Ø Ò ÔÓ× Ø ÓÒ ¹ ¸ Ø Ø Ø Ö Ü ×Ø× ×Ø Öع ÒØ Ö Ø ÓÒ Ø Ñ Ø× ×Ù Ø Ø Ë × Ù Ö ÒØ º ÖÓÑ ÔÖ Ø Ð ÔÓ ÒØ Ó Ú Û¸ Û Ú Ú Ò Ö Ø Ö ÓÒ ÓÛ ØÓ ÓÓ× Ø Ò×Ø ÒØ Ø× Ò Ø ÓÒØÖÓÐ Ò׺ ÍÒ ÓÖØÙÒ Ø Ðݸ Ò ÓØ × × Ø Ø Ñ Ø× Ô Ò ÓÒ Ø Û ÓÐ ×Ø Ø ÓÛ Ú Ö × Ò Ø ÈÁ¾ Ó × ÒÓØ Ù× Ú ÐÓ ØÝ ¸Ø × Ö Û Ò ÓÚ Ö ÓÑ Ò ÔÖ Ø Ý Ù× Ò Ò ×Ø Ñ Ø Ó Ø Ò Ö Ð Þ Ú ÐÓ Ø × Ø Ø ÔÖ × Ò×Ø ÒØ Ø× º Ò ÐÐݸ Û Ú × ÓÛÒ Ò × ÑÙÐ Ø ÓÒ× Ø ÔÓØ ÒØ Ð Ú ÒØ × Ó ÓÙÖ × Ñ × Ú ×¹ ¹Ú × Ü ×Ø Ò ÒÓÒÐ Ò Ö ÈÁ ÓÒØÖÓÐÐ Ö׺ ÖÓÑ Ø ÓÖ Ø Ð ÔÓ ÒØ Ó Ú Û¸ Ø Ø Ò ÕÙ Ó ×Û Ø Ò ÓÒØÖÓÐÐ Ö× × Ö ÒØÐÝ ÓÑ Ú ÖÝ ÔÓÔÙÐ Ö ´× º º ¾¼¸ ½ ¸ Ò Ö Ö Ò × Ø Ö Òµ Ò Ø ÒÓÒÐ Ò Ö ×Ý×Ø Ñ× Ð Ø Ö ØÙÖ º ÇÙÖ Ö ×ÙÐØ× ÐÐÙ×ØÖ Ø Ø ÑÔ Ø Ø Ø Ø × Ø ÓÖÝ × Ò ÔÖ Ø ¸ Ò Ò ÑÔÓÖØ ÒØ ××Ù Ó ÙØÙÖ Ö × Ö × ÓÛ ØÓ ÓÑ Ò ÐÓ Ð Ò ÐÓ Ð ÓÒØÖÓÐÐ Ö× ÚÓ Ò Ø ØØ Ø Ø ×Û Ø Ò Ø Ñ Ø× Ô Ò ÓÒ ÙÒÑ ×ÙÖ Ð ×Ø Ø Ú Ö Ð ×º ÒÓÛÐ Ñ ÒØ× È ÖØ Ó Ø × ÛÓÖ Û × ÔÔÐ Å Ø Ñ Ø × Ó ÈÖÓ º Æ Ñ Ö Ò Ý Ø ÔÖÓ Ø ÙÖÓÔ ÖÖ ÓÙØ Û Ð Ø ¬Ö×Ø ÙØ ÓÖ Û × Ê × Ö ÐÐÓÛ Ó Ø ÙÐØÝ Ó Ø Ø ÍÒ Ú Ö× ØÝ Ó ÌÛ ÒØ º À Ö Ø ÙÐÐÝ ÒÓÛÐ ×Ø Ó×Ô Ø Ð ØÝ Ø ËÝ×Ø Ñ× Ò ÓÒØÖÓÐ ÖÓÙÔº Ì × ÛÓÖ Û × Ô ÖØ ÐÐÝ ×ÙÔÔÓÖØ Ò Ô Ø Ð Ò ÀÙÑ Ò ÅÓ Ð Øݸ ÙÒ Ö Ö ÒØ ÒÓº Ê ¼ ¼ÈÄ ¿¼½¿ º Ê ½ Ö Ò × º ÐÓÒ Ò Êº ÇÖØ º Ò Ó × ÖÚ Ö¹ × × Ø¹ÔÓ ÒØ ÓÒØÖÓÐÐ Ö ÓÖ ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ× Û Ø ­ Ü Ð Ó ÒØ׺ ËÝ×غ ÓÒØÖº Ä ØØ Ö׸ ¾½ ¿¾ ß¿¿ ¸ ½ ¿º ¾ ˺ Ö ÑÓØÓº Ð ×× Ó ÕÙ × ¹Ò ØÙÖ Ð ÔÓØ ÒØ Ð× Ò ÝÔ Ö¹×Ø Ð ÈÁ × ÖÚÓ¹ÐÓÓÔ× ÓÖ ÒÓÒÐ Ò Ö ÖÓ ÓØ ×Ý×Ø Ñ׺ ÌÖ Ò׺ ËÓ º ÁÒ×ØÖÙÑ ÒØ ÓÒØÖº Ò º¸ ¿¼´ µ ½¼¼ ß½¼½¾¸ ½ º ¿ ˺ Ö ÑÓØÓ Ò º Å Ý Þ º ËØ Ð ØÝ Ò ÖÓ Ù×ØÒ ×× Ó È Ö Ú ØÝ ÓÑÔ Ò× Ø ÓÒ ÓÖ ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖº ÁÒ º Ϻ È ÙÐ Ò º ÊÓ ÓØ × Ì ÓÖÝ Ò ÔÔÐ Ø ÓÒ× Ë ¸ ÚÓÐÙÑ ¿¸ Ô × ß ¾¸ ½ Àº Ö Ù ×º ÅÓ Ð¹ × ÊÓ × ØÝ Ó ÌÛ ÒØ ¸ ÂÙÒ ½ ¿º ÓØ ÓÒØÖÓÐ ÖÓÑ Ì ÓÖÝ ØÓ ÈÖ Ø º ÓÒØÖÓÐ Û Ø ¹ÌÓÙÑ ¸ ØÓÖ׸ Ø × ×¸ ÍÒ Ú Ö¹ ºÈ Àº Ö Ù × Ò Àº Æ Ñ Öº ÐÓ Ð Ö ÙÐ Ø ÓÒ Ó ÖÓ ÓØ× Ù× Ò ÓÒÐÝ ÔÓ× Ø ÓÒ Ñ ×ÙÖ ¹ Ñ ÒØ׺ ËÝ×غ ÓÒØÖº Ä ØØ Ö׸ ¾½ ¾ ß¾ ¿¸ ½ ¿º ¾½ .

8. Concluding remarks 117 Áº κ ÙÖ ÓÚº ×ÝÑÔØÓØ ×Ø Ð Þ Ø ÓÒ Ó ÒÓÒÐ Ò Ö Ä Ö Ò Ò ×Ý×Ø Ñ× Û Ø ÓÙØ Ñ ¹ ×ÙÖ Ò Ú ÐÓ Ø ×º ÁÒ ÈÖÓ º ÁÒØ ÖÒ Øº ËÝѺ Ø Ú ÓÒØÖÓÐ Ò Å Ò Ð Ò Ò Ö Ò ¸ Ô × ¿ ß ½¸ ÄÝÓÒ¸ Ö Ò ¸ ½ º ʺ ÓÐ Ù Ò Ãº Ð ×׺ ÔØ Ú Ú ÐÓ ØÝ Ñ ×ÙÖ Ñ ÒØ׺ ÁÒ ÈÖÓ º Á ÔÖ Ð ½ º ź ÂÓ Ò××ÓÒ Ò ×Ý×Ø Ñ׺ ÁÒ ÈÖÓ º ʺ à ÐÐݺ ÊÓ ÓØ × ÙØÓÑ Øº ÓÑÔÐ ÒØ ÑÓØ ÓÒ ÓÒØÖÓÐ Ó Ñ Ò ÔÙÐ ØÓÖ× Û Ø ÓÙØ ÓÒ º ÊÓ ÓØ × ÙØÓÑ Øº¸ Å ÒÒ ÔÓР׸ Å Ò ×ÓØ ¸ º Ê ÒØÞ Öº Ø º ÙÖÓÔ Ò ÓÑÔÙØ Ø ÓÒ Ó È Û × ÐÝ ÔÙÒÓÚ ÙÒ Ø ÓÒ× ÓÖ Ý Ö ÓÒØÖº ÓÒ º¸ ÖÙ×× Ð׸ Ð ÙѸ ½ º È Ô Ö ÒÓº ½ º ÓÒØÖÓÐ Ó ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ׺ Á ÌÖ Ò׺ ÓÑÑ ÒØ× ÓÒ ÔØ Ú È ¸ ´½µ ½½ ß½½ ¸ ½ ¿º ½¼ ʺ à ÐÐݺ ÈÖÓ º ½¾Ø º Á × ÑÔÐ × ØßÔÓ ÒØ ÖÓ ÓØ ÓÒØÖÓÐÐ Ö Ý Ù× Ò ÓÒÐÝ ÔÓ× Ø ÓÒ Ñ ×ÙÖ Ñ ÒØ׺ ÁÒ ÏÓÖÐ ÓÒ Ö ×׸ ÚÓÐÙÑ ¸ Ô × ½ ¿ß½ ¸ ËÝ Ò Ý¸ Ù×ØÖ Ð ¸ ½ ¿º ÙÖ Ó ÈÁ ÓÒØÖÓÐ ÓÖ ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ׺ ÊÓ ÓØ ½½ ʺ à ÐÐݺ ØÙÒ Ò ÔÖÓ ½ ¸½ º ½¾ ¸ ½¿ ½ ½ß º º ÃÓ Ø× º ÊÓ ÓØ ÔÐ ÒÒ Ò Ö Âº º Ò ÄÓÞ ÒÓ¹È Ö Þ Ìº¸ ÅÁÌ ÈÖ ×׸ ½ º Ò ÓÒØÖÓÐ Ú ÔÓØ ÒØ Ð ÙÒ Ø ÓÒ׺ ÁÒ Ã Ø Çº¸ ØÓÖ׸ Ì ÊÓ ÓØ × Ê Ú Û ½¸ Ô × ¿ ß¿ º Ì Ò Ó Ö Ñ Ò ÔÙÐ ØÓÖ× Ù× Ò Å Ø Ñ Ø ×¸ ÍÒ Úº Ó ÌÛ ÒØ ¸ ½¿ º º º Ä Ö Ò Àº Æ Ñ Öº ÓÙÒ ÐÓ Ð ØÖ ÓÑÔÓ× Ø ÓÒØÖÓÐÐ Ö׺ Å ÑÓÖ Ò ÙÑ ÙÐØÝ Ó ÔÔÐ ½ º º º º Ä ×Ý×Ø Ñ׺ ÁÒ Ö Ò ÈÖÓ º Ø º ½ ½ ½ Àº Æ Ñ ÙÖÓÔ Ò Öº ÓÒØÖº ÐÓ × Ñ ÐÓ ÓÒ º ÐÐÝ ÓÙÒ ØÖ Ò ÓÒØÖÓÐÐ Ö× ÓÖ ÖÓ ÓØ ¸ ÖÙ×× Ð׸ Ð ÙѸ ½ º È Ô Ö ÒÓº º ÐÐÝ ×Ø ÓÒØÖº ʺ ÇÖØ ¸ º ÄÓÖ ¸ Ò Êº à ÐÐݺ ÓÖ ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ׺ Á ÌÖ Ò׺ ÙØÓÑ Øº Ð ÓÙØÔÙØ ÈÁ¾ ¸ ¼´ µ ½ ¿¾ß½ ¿ ¸ ½ Ö ÙÐ ØÓÖ º ʺ ÇÖØ ¸ º ÄÓÖ ¸ ʺ à ÐÐݸ Ò Äº ÈÖ Ðݺ ÇÒ Ô ×× Ú Øݹ × ÓÙØÔÙØ ÐÓ Ð ×Ø Ð Þ Ø ÓÒ Ó ÙÐ Ö¹Ä Ö Ò ×Ý×Ø Ñ׺ ÁÒغ º ÊÓ Ù×Ø Ò ÆÓÒÐ Ò Ö ÓÒØÖÓи ×Ô Ð ××Ù ÓÒ ÓÒØÖÓÐ Ó ÒÓÒÐ Ò Ö Ñ Ò Ð ×Ý×Ø Ñ׸ ´ µ ¿½¿ß¿¾ ¸ ½ º ´Àº Æ Ñ Ö Ò ºÂº Ú Ò Ö Ë Ø ×ºµº ʺ ÇÖØ Ò Åº ËÔÓÒ º Ñ Ø ¸¾ ¹ ß ¸½ º ÔØ Ú ÑÓØ ÓÒ ÓÒØÖÓÐ Ó Ö ÖÓ ÓØ× ØÙØÓÖ Ðº ØÓÖ¸ À Ò ÙØÓ¹ ½ ½ ½ ¾¼ ź ËÔÓÒ º ÅÓØ ÓÒ ÓÒØÖÓÐ Ó ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ׺ ÁÒ Ïº Ä Ú Ò ¸ ÓÒØÖÓи Ô × ½¿¿ ß½¿ ¼º Ê ÈÖ ×׸ ½ º ź Ì Ð ØÓÖ׺ ÓÒØÖº ÓÓ Ó Ò Ëº ËŠº Ö ÑÓØÓº Ò Û ×º ÓÒØÖº ÝÒº ËÝ×غ Å Ñ Ø Ó ÓÖ ÝÒ Ñ ¸ ½¼¿ ½½ ß½¾ ¸ ½ ½º ÓÒØÖÓÐ Ó Ñ Ò ÔÙ¹ ÈÖÓ º Ø º ÙÖÓÔ Ò º Ì Ð Ò Æº à ÔÓÓÖº ÍÒ Ø Ò ÐÓ Ð Ò ÐÓ Ð ÓÒØÖÓÐÐ Ö׺ ÁÒ ÓÒ º¸ ÖÙ×× Ð׸ Ð ÙѸ ½ º È Ô Ö ÒÓº º ÓÒØÖÓÐ ÓÖ ÖÓ ÓØ Ñ Ò ÔÙÐ ØÓÖ׸º Á ¾¾ ¾½ Ⱥ ÌÓÑ º ÔØ Ú È ´ µ ß ¼¸ ½ ½º ÌÖ Ò׺ ÊÓ ÓØ × ÙØÓÑ Øº ¸ .6.

GAS of robot manipulators with linear PID and PI¾ D control ¾¾ Ⱥ ÌÓÑ º ÓÒØÖº¸ × ÑÔÐ È ÓÒØÖÓÐÐ Ö ÓÖ ÖÓ ÓØ× Û Ø ½º Ò º º ÃÓ ÖÑ׺ ÁÒ Ø× ÈÖÓ º Á Ð ×Ø Ó ÒØ׺ Á ÌÖ Ò׺ ÙØÓÑ Øº ¿ ´½¼µ ½¾¼ ß½¾½¿¸ ½ Ø ÓÑ ¸ ÔØ Ú Ö Ñ ÒØÓ¸ º º Ê ÞÞ ¸ ¸ ½ ½º ¾¿ ĺĺ Ï Ò Û ¾ß ¸ Ë º ÓÑÔ Ö Ø Ú ÓÒ º ÊÓ ÓØ ÜÔ Ö Ñ ÒØ× Û Ø × ÙØÓÑ Øº¸ ÓÒØÖÓÐÐ Ö ÓÖ ÖÓ ÓØ Ô × ¾¿ .118 Chapter 8.

1998. Ortega. and H. pp. 119 . 2. Kelly. The Netherlands.Chapter 9 Adaptive and filtered visual servoing of planar robots This chapter consists of the paper: E.” in Proceedings of the Fourth IFAC Symposium on Nonlinear Control Systems c Design (NOLCOS’98). ­1998 IFAC. R. 563–568. Enschede. “Adaptive and filtered visual servoing of planar robots. Lefeber. vol. R. Nijmeijer.

This control philosophy is certainly effective for slow robot motion but its application is of a limited value when high speed motions are demanded.. Also. Casta˜ o and Hutchinn son. Most previous research has been started with the optics of the kinematic control where the robot velocity control (in joint or Cartesian space) is assumed to be computed in advance. A. however.Lefeber. The control goal is to place the robot end-effector over a desired static target by using a vision system equiped with a fixed camera to ‘see’ the robot end-effector and target. it is until recent years that its effectiveness has reached the real world applications thanks to the technological improvement in cameras and dedicated hardware for image processing (Hashimoto. To achieve this goal we introduce a class of visual servo state feedback controllers and output (position) feedback controllers provided the camera orientation is known. Mitsuda et al. 1996.O. Adaptive and filtered visual servoing of planar robots ADAPTIVE AND FILTERED VISUAL SERVOING OF PLANAR ROBOTS Erjen Lefeber £ Rafael Kelly ££ Romeo Ortega £££ Henk Nijmeijer £ ££££ £ Faculty of Mathematical Sciences. Espiau. Although this fact has been recognized decades ago. The Netherlands Abstract: In this paper we address the visual servoing of planar robot manipulators with a fixed–camera configuration..H. Hager et al.A. rkelly@cicese. Keywords: Visual servoing.fr ££££ Faculty of Mechanical Engineering Eindhoven University of Technology. For the case of unknown camera orientation a class of adaptive visual servo controllers is presented.. 1993. stability 1. Nelson et 563 al. INTRODUCTION External sensors such as visual systems enlarge the potential applications of actual robot manipulators evolving in unstructured environments. All three classes contain controllers that meet input constraints. inversion of the robot Jacobian and computation of the inverse kinematics. France. Chaumette et al. Postal 2615. 1994. and therefore the robot dynamics can be neglected (Allen et al. In physical systems. that is. actuators exert forces or torques.Nijmeijer @math. Department of Systems Signals and Control. Hutchinson et al. 1995.C. Mexico. 1991. 1993. and a rigorous stability analysis is given for an appropriate (adaptive) set point controller.nl ££ Divisi´ n de F´sica Aplicada. Apdo. Box 217. LSS/CNRS/Sup´ lec.. We focus the visual servoing problem from an automatic control point of view by considering the full robot nonlinear dynamics with the applied torques as the control actions. CICESE. Previous efforts in .. Feddema et al. This approach is an example of a mechanical control system in which a kinematic model is used for control design. PO Box 513. This paper deals with a fixed camera configuration for visual servoing of robot manipulators. The Netherlands...Ortega@lss. 91192 Gif Sur Yvette Cedex. 1991.J.mx £££ Laboratoire des Signaux et Syst` mes. Plateau e e de Moulon. 5600 MB Eindhoven. 1996). P. B.utwente. we are interested in simple control schemes avoiding the common procedures of camera calibration. robotics. the velocity of the system is assumed to be a direct input which can be manipulated. Adm. 1993. 1996). 7500 AE Enschede. Romeo.supelec.120 Chapter 9. 1. o ı Ensenada. 22800.

The image acquired by the camera supplies a two– dimensional array of brightness values from a three– dimensional scene. 1989)): Property 3. but the intrinsic camera parameters .2. Then the control problem can be stated as to design a control law for the actuator torques such that the robot tip reaches. 1995. In section 4 we derive a class of adaptive visual servo controllers in case the camera orientation is unknown. Spong and Vidyasagar. (Ortega and Spong.. where  ´Õµ ¾ Áʾ¢¾ is the analytic robot Jacobian.. we provide a simple common framework to design standard proportional or saturated controllers. 1989. In this paper we assume that the image features are the projection into the 2D image plane of 3D points in the scene space. The output variable Ý ¾ Á ¾ is defined as the position Ê (in pixels) of the robot tip in the image. (Spong and Vidyasagar. 1996.g. The Jacobian is bounded for all Õ i. where ´Õ µ stands for the elements of .g.3 Problem formulation Consider the robotic system (1) together with the output equation (3). 1996). Å ´Õ µ is the ¾ ¢ ¾ symmetric positive definite manipulator inertia matrix.g. is the ¾ ¢ ½ vector of applied joint torques.9. where the camera orientation is known. For convenience we define a Cartesian reference frame anywhere in the robot base.. 1988)).. 1996. PRELIMINARIES AND NOTATION 2. This Ê means there exist finite constants ¼ such that Ñ Ü Õ Á ¾ Ê ¾ ´Õ µ ½ ¾ ´Õ µ. 1989. An important property of this Jacobian is the following (see e. Kelly. 1995. Kelly et al. 1996) to the cases where velocity measurements are not available and the camera orientation parameter is unknown.e. center offset). 1990). The main contributions of our work are extensions of the results in (Kelly. The gravitational torque vector ´Õ µ is bounded for all Õ ¾ Á ¾ . ¾ denote intrinsic camera parameters (scale factors. focal length. and ´Õ µ is the ¾ ¢ ½ vector of gravitational torques. Ê and  ¼ ½  ½ ¼ Õ µÕ · ´Õ µ (1) where Õ is the ¾ ¢ ½ vector of joint displacements. Ê Let Ý ¾ Á ¾ be a desired constant position for the robot tip in the image plane. the dynamics of a serial ¾–link rigid robot manipulator can be written as (see e. (Craig. In section 5 we present a class of visual servo controllers in case we have no velocity measurements available. The first problem is solved invoking the (by now) standard ”dirty derivative” solution. preliminaries and notation this subject have been reported in (Coste-Mani` re et e al. 1989)) The time derivative of the inertia matrix. Section 6 contains our concluding remarks. a perspective projection and a linear transformation (Feddema et al. Hutchinson et al. PROBLEM FORMULATION. This image may undergo various types of computer processing to enhance image properties and extract image features. 1996) Ý  Â ´Õ µ   ½ · ¾ (3) where ¼ and ½ . 1996. the later problem involves a nonlinearly parametrized adaptive system. and the centripetal and Coriolis matrix satisfy: ÕÌ ½ ¾ The direct kinematics yields  ´ÕµÕ. 1991. (see e.. The mapping from the joint positions Õ to the output Ý involves a rigid body transformation. Furthermore. the target point 564 Å ´Õ µ   ´Õ Õµ Õ ¼ Õ Õ ¾ Áʾ (2) Property 2. Suppose that together with the position Ý of the robot tip in the image also measurements of the joint positions Õ and velocities Õ are available. there exists a finite constant ÂÅ such that ¾ Áʾ.g.  ´Õ µ ÂÅ Õ ¾ Áʾ 2. (see e. Section 2 contains the problem formulation. 1996). –a situation which is essentially unexplored in the field– hence special analysis and synthesis tools have to be developed for its solution. Á ¾ Ê Á ¾ stands for the robot direct kinematics. However. Spong and Vidyasagar. 1993.. This corresponds to the image of a point target which is assumed to be located strictly inside the robot workspace. ´Õ Õ µÕ is the ¾ ¢ ½ vector of centripetal and Coriolis torques.2 Output equation We consider a fixed CDD camera whose optical axis is perpendicular to the plane where the robot tip evolves. 2.1 Robot dynamics In the absence of friction or other disturbances. 1989)): Å ´Õ µÕ · ´Õ 121 For the purposes of this paper we consider a planar two degrees of freedom robot arm. 2. Problem formulation. In section 3 we introduce a class of visual servo controllers which includes the controllers reported in (Coste-Mani` re et al. Miyazaki and Masutani. e Kelly et al. Kelly. in the image supplied on the screen. Two important properties of the robot dynamic model are the following: Property 1. The orientation of the camera with respect to the robot frame is denoted by . preliminaries and notation. ½ and ¾ are unknown. The organisation of this paper is as follows. The corresponding output equation has the form (Kelly. (Ortega and Spong. Lei and Ghosh.

can be measured too. we define þ Ü Õ Ý Õ  Õ  Â and Ý ´ ´Õ µ Ý  Ý ´Õ µµ Since Ý is measurable and Ý is given. a Ê necessary condition for continuously differentiable is that its Jacobian Ü is symmetric. as well as the assumption that measurements of the joint velocities Õ are available. Also. Consider the system (1) in closedloop with the control law 2. 565 . ´Ü½ (5) ÜÒ µ ´Õ µ ÜÒ where ½ ¾¾   (6) ¾ ´Ýµ ¾.e the class of ¾ Ò for which there Ê exists a constant Å ¾ Á such that ´Üµ Å for all Ü ¾ Á Ò . PROOF. Ü ´Üµ × Ø´ ܵ or ´Üµ ¼) we ·Ü ( obtain elements of Ò . (5)). these controllers ensure local regulation.  Â ´Õ µ   ½ · ¾ We conclude this section by noticing that since Õ is  Â  ´ÕµÕ and therefore fixed. . 3. Ý ´Ý µ (4) ÕÌ Â ´ÕµÌ  ´Ý µ i.   Õ is not available for measurement. However. Furthermore. Under Assumptions 4–5 we have ÐÑ Ø ½ ´Õµ    ´ÕµÌ  and for which ÜÌ ´Üµ is a positive definite function. Kelly et al.122 Chapter 9. since Õ is unknown. Throughout we denote for ¾ Ò by ´Üµ the associated function of which is the gradient (cf. It is worth noticing that in case Ý corresponds to the image of a point target located strictly inside the robot workspace. . Definition 7. there exists a neighborhood around Õ for which Õ is the only solution of (4). if and only According to Assumptions 4–5 this equation has an isolated equilibrium at Õ Ì Õ Ì Ì Õ Ì ¼Ì Ì . e Assuming that the camera orientation is known.4 Notation Throughout we use the following notation. i. Adaptive and filtered visual servoing of planar robots In general it is not easy to verify whether a given Á Ò Ê Á Ò can be written as the gradient of Ê a radially unbounded Á Ò Ê Á . 1995. Kelly. then Assumptions 4 and 5 are trivially satisfied. 1996). By choosing ´Üµ Ø Ò ´ ܵ. It is easy to see that elements of Ò are the functions ´Üµ placed anywhere in the robot workspace. This is formally stated in the next Proposition 10. whereas ´Üµ þ Ü is an element of Ò but not of Ò . ÐÑ Ø Definition 8. and are continuous nondecreasing functions satisfying ´¼µ ¼ and ¼ ´¼µ ¼ ( ½ Ò).. However. and the full state (Õ Õ) is measured. 1996. In other words: ÐÑ Ø ½ ݴص Ý Later in this paper the assumption that the camera orientation is known will be relaxed. under Assumptions 4 and 5 we have that Õ Õ ¾ Á ¾ is an isolated solution of Ê Ý Õ Ì where ý ý is a Ò ¢ Ò diagonal positive definite Ì matrix. (Problem solvability) There exists a constant (unknown) vector Õ ¾ Á ¾ such that Ê Ý Ã½ ½ ´Ü½ µ ´Üµ Ò ´ÜÒ µ Ì and  Â ´Õ µ   ½ · ¾ Assumption 5. Ê An important property of Property 9. To be able to solve the problem formulated above we make the following assumptions: Assumption 4.e. A CLASS OF STABLE VISUAL SERVO CONTROLLERS In this section we introduce a class of visual servo controllers which includes those reported in (CosteMani` re et al. Let Ò denote the class of continuous functions Á Ò Ê Á Ò for which there exists a Ê positive definite Á Ò Ê Á such that Ê ¾ ´Üµ ´Ü½ ÜÒ µ ܽ ´Ü½ ÜÒ µ ¿ . Using the control law (6) results in the closed-loop dynamics Å ´Õ µÕ · ´Õ ¾ Then Ò is the following: ´Üµ ¼ Õ µÕ · ¾ ½ ´Õ µ ·  ´ÕµÌ  ¾ ´Ýµ ¼(7) Ò. Let if Ü ¼.. þ þ is a Ò ¢ Ò (not necessarily diagonal) positive definite matrix. Let Ò denote the class of ¾ Ò that are bounded. There exists a neighborhood around Õ for which Ø Â ´Õ µ ¼ (by smoothness of the Jacobian). ½ ݴص ½ մص ¼ provided the initial conditions Õ ´¼µ and Ý ´¼µ are sufficiently small. (Nonsingularity at the desired configuration) For the (unknown) vector Õ ¾ Á ¾ it Ê holds true that Ø Â ´Õ µ ¼ Corollary 6.

For the system (1) it is well known (Ortega and Spong. using Property 1: so we can conclude using Property 9: Î ´Õ Õ µ  ÕÌ ·Õ Ì ½ ´Õ µ ½ ݴص ¼ ¾ Consider the system (1) where we deal with the input constraints ´Øµ Ñ Ü ½ ¾ (8) Then we can derive the following . Using the control law (9) together with the parameter update law (10) results in the closed-loop dynamics Î ´Õ Õ µ ½ ½ Ì Õ Å ´Õ µÕ · ¾ ´Ý µ ¾ which is a (locally) positive definite function. in contrast with the previous section. Remark 15. for any initial condition in a small neighborhood of the equilibrium we have ÐÑ Ø Î ´Õ Õ µ ½ Ì ¾ Õ Å ´Õ µÕ · ½ ¾ ´Ý µ· ½ ­ ´½   Ó× µ (12) ½ մص ÐÑ Ø ÐÑ Ø ½ ¾ ´Ý ´Øµµ ¼ which is a (locally) positive definite function. Consider the Lyapunov function candidate ¾ ´Ý µ which is negative semidefinite in the state ´Õ Õ µ. From ¼ it follows that necessarily Õ ¼. The switching nature of the controller (9) leads to chattering. the closed-loop system tends to the largest invariant set of points ´Õ Õ µ for which Î Î  ÕÌ ½ ´Õµ ¼ ¼. Along the closed-loop dynamics (7) its time-derivative becomes. Along the closed-loop dynamics (11) its time-derivative becomes. If Ñ Ü defined in Property 2. If Ñ Ü defined in Property 2. Under Assumptions 4–5 where ­ we have. are . Consider the system (1) in closedloop with the control law Ì Â ½ ´Õ µ    ´Õ µ ¾ ´Ý µ if Õ Ì Â ´Õ µÌ  ¾ ´Ý µ ¼ Ì Â ´Õ µ   ½ ´Õ µ ·  ´Õ µ ¾ ´Ý µ Ì Ì Â if Õ Â ´Õ µ ¼ ¾ ´Ý µ ´Õ µ ½ ´Øµ ÐÑ Ø ½ մص ÐÑ Ø ½ ݴص ¼ ¾   Remark 13. Using LaSalle’s theorem and Corollary 6. According to LaSalle’s theorem. then there exist ½ ¾ ¾ ¾ such that the controller (6) meets (8) and in closedloop with the system (1) yields ÐÑ Ø Â ´ÕµÌ  ¾ ´Ýµ · Ì Ì Â · × Ò Õ Â ´Õ µ  ¾ ´Ý µ ¬ ¬ ¬ Ì Ì  Õ ½ ´Õµ   ¬Õ  ´ÕµÌ  ¾ ´Ýµ¬ · ¬ Ì Ì  Â  ·Õ  ´Õ µ ´ · × Ò Âµ ¾ ´Ý µ ¬ ¬ ¬ Ì ¬ Ì Ì Â  Õ ½ ´Õµ   ¬Õ  ´Õµ ¾ ´Ý µ¬ · Ì Ì Â · Ó× Õ Â ´Õ µ ¾ ´Ý µ Ì  Õ ½ ´Õµ which is negative semidefinite in the state ´Õ Õ µ. where has been Corollary 11. Using a suitably smoothed control law might be a way to overcome the chattering. Then from the closed-loop dynamics (11) we know ¼ and using Corollary 6 also ¾ ´Ý µ ¼. Still assuming that the full state ´Õ Õ µ is available for measurement we introduce a class of adaptive controllers that ensure local regulation: Proposition 12. Therefore LaSalle’s theorem gives us for any initial condition in a small neighborhood of the origin ÐÑ Ø   ¬ÕÌ Â ´ÕµÌ ¬ ¬  ¾ ´Ý µ¬ · ¬ ¬ ½ ݴص ¼ provided the initial conditions are sufficiently small. 4. which is undesirable.4. Ý ´¼µ and sufficiently small.9. Adaptive visual servoing Consider the Lyapunov function candidate 123 PROOF. also the camera orientation is unknown. We update the parameter as ­Õ Ì Â ´ÕµÌ  ÐÑ Ø Â ¾ ´Ý µ (10) ¼ is a constant. if we define   : ÐÑ Ø ½ ݴص ÐÑ Ø ½ մص ÐÑ Ø ½ ´Øµ ¼ provided the initial conditions are sufficiently small. ADAPTIVE VISUAL SERVOING In this section we consider the case in which. then there exist ½ ¾ ¾ ¾ such that the controller (9) meets (8) and in closedloop with the system (1) yields ÐÑ Ø (9) where ½ ¾ ¾ ¾. As in the previous section we can derive the following . using Property 1: Å ´Õ µÕ · ´Õ Õ µÕ · ½ ´Õ µ ­ Õ Ì Â ´Õ µÌ   ¦Â ´ÕµÌ ¾ ´Ý µ  ¾ ´Ý µ (11) ¼ Î ´Õ Õµ  ÕÌ ·Õ Ì  ÕÌ Â Õ ½ ´Õ µ Ì Â ´Õ µ ½ ´Õ µ   Ì Â ´ÕµÌ ¾ ´Ý µ ¼  ¾ ´Ý µ · where the ’¦’ reads as a ’·’ if Õ Ì Â ´Õ µÌ  and as a ’ ’ if Õ Ì Â ´Õ µÌ  ¾ ´Ý µ ¼. where has been Corollary 14. 1989) that there exist a reparametrization of all unknown system parameters into a 566 ½ ݴص ½ մص ÐÑ Ø ½ ´Øµ ¼ ´¼µ provided the initial conditions Õ ´¼µ.

To obtain other possible filters. as we have seen in deriving (17). The filter (17) is similar to the ones ´Õ ÕµÕ ·Â ´ÕµÌ  ½´Þ µ· ¾´Ýµ  Â  ´ÕµÕ   Þ Þ ¼ (15) As in the previous sections we can derive the following Corollary 18. we only add ½ ¢Ì   ½ ¢ to the Lyapunov function (12). for any initial condition in a small neighborhood of the equilibrium we can conclude Ø ¢    Ð Ñ Û´Øµ ½ Ø Ð Ñ Þ ´Øµ ½ Ø Ð Ñ Õ´Øµ ½ Ø Ð Ñ Ý´Øµ ½ ¼ Ì ´Õ Õ Õ Õ µÕ ¾ where    Ì ¼ is a positive definite matrix. be seen as a simple representative of a whole class of possible filters. As far as the proof is concerned. 1994. 5. CONCLUDING REMARKS In this paper we addressed the visual servoing of planar robot manipulators under a fixed camera config-  ¾ ´Ý µ· ¾ ´Ý µ· Þ ½ ´Þ µ· Õ Ì ½ ´Þ µ   ´ÕµÌ ½ ´Þ µ   567 . ¾ ). To prove asymptotic stability as in Proposition 12. 6. one possibly has to replace ½ ´Ý µ in (16) with a different . Adaptive and filtered visual servoing of planar robots parameter vector ¢ ¾ Á Ô that enters linearly in the Ê system dynamics (1). Also the more general class of linear filters presented in (Arimoto et al. then it can easily be seen that instead of (14) also the filter £½ Ý   £¾ Û (17) £¿ ´£¾ Ý   £¾ Ûµ can be used (replace in (16) ½ ´Ý µ with the ´Ý µ associated with £ ½ ½ ´Ý µ to obtain ½ ½ Ì Î   Þ £¿ £¾ £ ½ ½ ´Þ µ Þ Ì ½ ´Þ µ Þ Û ½ ´Õµ    ´ÕµÌ where ½ ¾  ½ ´Þ µ ¾    ´ÕµÌ  ¾ ´Ý µ (13) ¾. sometimes resulting in a different expression for ½ ´Þ µ in (13). This is formally stated in the next Proposition 16. Also a wide variety of nonlinear filters can be rewritten as (14). In general one can say that the filter (14) is a representative of a whole class of controllers that takes its simple form due to a well chosen change of coordinates. Using LaSalle’s theorem and Corollary 6. Consider the system (1) in closedloop with the control law Remark 17. Assuming that the camera orientation is known and only measurements of the joint positions Õ are available we introduce a class of controllers and filters that ensure local regulation. Along the closed-loop dynamics (15) its time-derivative becomes: Î ´Õ Õ Þ µ Ð Ñ Û´Øµ ½ Ø Ð Ñ Þ ´Øµ ½ Ø Ð Ñ Õ´Øµ ½ Ø Ð Ñ Ý´Øµ ½ ¼  ÕÌ ´ÕµÌ ·ÕÌ ´ÕµÌ  ÕÌ ´ÕµÌ    ½ ´Þ µ   ÕÌ ´ÕµÌ ½ Ì provided the initial conditions are sufficiently small. 1993. For instance if ½ ¾ ¾ satisfies the property that also £ ¾ ¾ . no measurements of the joint velocities Õ are available. just apply a suitable change of coordinates in Þ and Û (suitable in the sense that Î remains negative definite). Kelly and Santiba˜ ez.124 Chapter 9. where has been defined in Property 2. and Ý´¼µ are sufficiently small. Berghuis and Nijmeijer. 1996) can similarly be seen as a special n case of (14). Using the control law (13) together with the filter (14) results in the closed-loop dynamics Å ´Õ µÕ · Ð Ñ Û´Øµ ½ Ø Ð Ñ Þ ´Øµ ½ Ð Ñ Õ´Øµ ½ Ø Ð Ñ Ý´Øµ ½ ¼ with ½ ´Þ µ ¾ presented in (Ailon and Ortega. FILTERED VISUAL SERVOING In this section we consider the case in which. then there exist ½ ¾ ¾ ¾ such that the controller (13) meets (8) and in closedloop with the system (1) yields Ø Consider the Lyapunov function candidate Î ´Õ Õ Þ µ ½ ½ ½ Ì Õ Å ´Õ µÕ · ¾ ´Ý µ · ½ ´Þ µ ¾ (16) which is a (locally) positive definite function. If Ñ Ü . 1993). and Þ is generated from the filter Ý Ý Þ Û  Û  Û Ø (14) Under Assumptions 4–5 we have Ø provided the initial conditions Û´¼µ. Therefore the following holds:   ½ÞÌ   ½ ÞÌ ½ ´Þ µ ½ ´Þ µ ¢µÕ · ´Õ Õ ¢µÕ · ´Õ ¢µ ż ´Õ µÕ · ¼ ´Õ Õ µÕ · ¼ ´Õ µ · ´Õ Õ Õ Õµ¢ We can cope with those unknown system parameters in the ‘standard’ way by adding ´Õ Õ Õ Õ µ¢ to the control law (and replacing ´Õ µ with ¼ ´Õ µ). PROOF. similar to the one presented in (Lefeber and Nijmeijer. The filter (14) can. where ¢ is updated according to Å ´Õ ¼ which is negative semidefinite in the state ´Õ Õ Þ µ. ¾ where we defined ¢ ¢   ¢. Õ ´¼µ. 1997). in contrast with section 3. where £ is an arbitrary positive definite matrix..

USA. Craig. Casta˜ o. Papanikolopoulos and P. G. Global regulation of robots using only position measurement. 633–636. Naniwa (1994). 877–888. Vol. Shirkey and M. MA. Enschede. Corke (1996). and H. In: Proceedings of Experimental Robotics III: The 3rd International Symposium. Mitsuda. Hashimoto. Miyazaki. Addison-Wesley Pub. Robustness of sensory feedback control based on imperfect jacobian. IEEE Control Systems Magazine 15(1). Memorandum 1404. H. 2643–2649. 37–42. R. Robust asymptotically stable visual servoing of planar robots. IEEE Robotics and Automation Magazine 3(2). K. D. Hutchinson (1994). and R. Vol. Miyazaki (1996). 7. S. Visual servoing: Real-time control of robot manipulators based on visual sensory feedback. Espiau. MIT Press. 125 Feddema. Systems and Control Letters 21. and B. Nijmeijer (1993). K.. D. Rives and B. J. MN. Hager and P. In: Proceedings of the 1996 IEEE International Conference on Robotics and Automation. IEEE Transactions on Robotics and Automation 12(5). V. The results include controllers that satisfy input constraints. and H. Lee and O. In: Proceedings of the 1st Asian Control Conference. C. tracking it and estimating its velocity by visual servoing. Mitchell (1991). W.. Vidyasagar (1989). Nijmeijer (1997). Adaptive motion control of rigid robots: a tutorial. Kelly. N. Ghosh (1993). pp. Automated tracking and grasping of a moving object with a robotic hand-eye system. J. W. (1996). In case of unknown camera orientation a class of adaptive controllers has been presented. Effect of camera calibration errors on visual servoing in robotics. Robotic visual servoing and robotic assembly tasks. B. ACKNOWLEDGEMENTS This work was carried out while the first author was visiting R. we introduced a class of visual servo controllers for both the state feedback and output feedback case (position measurements). San Antonio. pp. Yoshimi and P. 23–31. T. Chang and A. B. 1–12. and V. 531–537. 289–293. Chaumette. A class of global n regulators by using output feedback for elastic joint robots.. Timcenko. 152–165. Japan. 31–47. Japan. M. A.. University of Twente.. B. pp. Nantes. T.. Spong (1996). K. K. REFERENCES Ailon. J. F. Minneapolis. Masutani (1990). TX. Bounded global tracking of rigid robot manipulators using composite controllers.. 568 .9. pp. and M. IEEE Transactions on Robotics and Automation 10(3). Kelly. and M. A. A. Ortega (1993). pp. Lefeber. Faculty of Applied Mathematics. G. pp. 2. Khosla (1996). Adaptive control of mechanical manipulators. Kyoto. Fixed camera visual servoing of planar robots. References uration. and Y. R. Maru. In: Proceedings of the 13th Triennial World Congress. Allen. P. Positioning of a robot with respect to an object. Espiau (1991). pp.. K.. 759–766.. Weighted selection of image features for resolved rate visual feedback control. (1993). 651– 670. A class of linear velocity observers for nonlinear mechanical systems. J. France. Ortega. In case the camera orientation is known. IEEE Transactions on Robotics and Automation 12(5). Robot Dynamics and Control. Parra-Vega and T. Khosla e (1995). CA. P. 334–342. 1104–1111. Visually guided robotic tracking and grasping of a moving object with a robotic hand-eye system. pp. Spong (1989). USA. Automatica 25(6). 2248–2253. California. Santiba˜ ez (1996). Kelly. Sacramenta. In: Robotics Research: The Fifth International Symposium. IEEE Transactions on Robotics and Automation 7(1). Adaptive calibration and control of 2D monocular visual servo system. S. R. C. F. A. Berghuis. Hager. Robot hand-eye coordination based on stereo vision. 329–335. 182–192. S. Cambridge. Hutchinson. World Scientific Publishing Co.7. Arimoto. R. In: Proceedings of the IFAC Symposium on Robot Control (SYROCO). Fujikawa and F. S. Bishop. An observer-based controller for robot manipulators with flexible joints. Visual servoing based on the use of binocular visual space. and M. Nelson. 30– 39. (1993).. USA. IEEE Transactions on Robotics and Automation 9(2). USA. Ortega in the frame of the Capital and Mobility European network on ”Nonlinear and Adaptive Control: Towards a design Methodology for Physical Systems”. Morse (1995). Systems and Control Letters 21. P. K. Spong. San Francisco. R. I. Vol. and S. pp. Visual servoing in the task-function framework: A contour following task. 1604–1609. Visual n compliance: Task-directed visual servo control. (1988). Journal of Intelligent and Robotic Systems 12. W. In: Proceedings of the 32nd Conference on Decision and Control. G. Lei. 201–208. Mitchelman (1993). Singapore. A tutorial on visual servo control. N. In: Proceedings IROS 96. Wiley. B. Couvignou and P. M. A. P. W. 3. The Netherlands. E. Spong (1997). P. Coste-Mani` re. USA. Tokyo. 7. E. In: Proceedings of the 1991 IEEE Conference on Robotics and Automation.

126 Chapter 9. Adaptive and filtered visual servoing of planar robots .

Phoenix. c Arizona. Nijmeijer.Chapter 10 Adaptive tracking control of non-holonomic systems: an example This chapter consists of the paper: E. ­1999 IEEE. Lefeber and H. pp.” in Proceedings of the 38th IEEE Conference on Decision and Control. 127 . “Adaptive tracking control of nonholonomic systems: an example. 1999. 2094–2099. USA.

Henk NijmeijerÝ Þ Faculty of Mathematical Sciences. Section 6 concludes the paper. The Netherlands Abstract car with rear wheel driving and front wheel steering: Ü Ý Ú Ó× Ú× Ò Ú Ø Ò Ä We study an example of an adaptive (state) tracking control problem for a four-wheel mobile robot. The proposed solutions to this problem follow mainly two routes.g. 4] and references therein. is the angle between front wheel ¼ is a constant that denotes the length of and car and Ä the car (see also Figure 1). Section 3 contains some definitions and preliminary results. it is possible that specifying a reference-state trajectory and not knowing certain parameters are in conflict with each other. see e. Section 2 contains the problem formulation of the tracking problem and illustrates the difficulties in arriving at the problem formulation for the adaptive tracking problem. Section 4 addresses the tracking problem and prepares for Section 5 in which the adaptive tracking problem is considered.utwente. For a good overview.00 ­ 1999 IEEE 2094 Ú   ¬ Ä Ý Ü Figure 1: The mobile car The organization of the paper is as follows. The question then arises how to formulate the adaptive tracking problem in such a way that it reduces to the state feedback tracking problem in case the parameters are known. 8].Nijmeijer]@math. since specifying the reference state-trajectory can be in conflict with not knowing certain parameters. Finally. see [5]. In this paper we consider a simple academic example that clearly illustrates the above mentioned conflict. 11.1 Tracking control problem Since we want the adaptive tracking control problem to reduce to the tracking problem for known Ä.Lefeber. It turns out that formulating the adaptive state-feedback tracking control problem is not straightforward. Eindhoven University of Technology. we first have to formulate the tracking problem for the case Ä is known. and is assumed to be unknown. A general methodology for solving the problem is derived. Adaptive tracking control of non-holonomic systems: an example Proceedings of the 38Ø Conference on Decision & Control Phoenix. 2 Problem formulation 2. namely discontinuous and/or time-varying control. as it is an illustrative example of the general adaptive state-feedback tracking control problem.J. . 5600 MB Eindhoven. or the state trajectory to be tracked is feasable for any possible parameter. We propose a formulation for the adaptive (state) tracking control problem and derive a general methodology for solving this problem. but all papers are either concerned with adaptive output tracking. see the survey paper [6].O. (1) The forward velocity of the rear wheel Ú and the angular velocity of the front wheel are considered as inputs. The Netherlands. The example we study is the kinematic model of a mobile c 0-7803-5250-5/99/$ 10. Results on adaptive stabilization can be found in [7. In [9. 7500 AE Enschede. 10. One of the reasons for the attention is the lack of a continuous static state feedback control since Brockett’s necessary condition for smooth stabilization is not met. Box 217. Airzona USA ¡ December 1999 WeM10 14:40 Adaptive tracking control of nonholonomic systems: an example Ý Erjen LefeberÝ . is the orientation of the body of the car. University of Twente. Less studied is the adaptive control of nonholonomic systems.nl Þ Faculty of Mechanical Engineering. P. 3. 1 Introduction In recent years a lot of interest has been devoted to (mainly) stabilization and tracking of nonholonomic dynamic systems. However. [1. [A. PO Box 513. 12] the adaptive tracking problem is studied.128 Chapter 10. ´Ü Ý µ is the center of the rear axis of the vehicle. 2. Our example illustrates this difficulty and we propose a problem formulation for the adaptive state-feedback tracking problem that meets the natural prerequisite that it reduces to the statefeedback tracking problem if the parameters are known.H.A.

3. since once ÚÖ ´Øµ. ÜÖ ´Øµ and ÝÖ ´Øµ.1 (Tracking control problem) Given a feasible reference trajectory ´ ÜÖ ÝÖ Ö Ö Ì ÚÖ Ö Ì µ. since these signals are just functions (depending on Ä) of ÚÖ ´Øµ. In order to be able to solve the (adaptive) tracking control problem. since it is possible to have ܴص ÜÖ ´Øµ. and then all other signals are derived from the equations (2). ÝÖ ´Øµ. So the question is how to formulate the adaptive tracking problem for the nonholonomic system (1) in such a way that it reduces to the state-feedback tracking control problem for the case Ä is known? Appearently we can not both specify ÚÖ . Ö ´Øµ. The tracking control problem then can be formulated as Problem 2. In specifying ÚÖ ´Øµ. First of all. but also of ÚÖ ´Øµ.10. ÝÖ ´Øµ. to determine Ö ´Øµ uniquely. we can assume that a feasible reference trajecÄ tory ´ ÜÖ ÝÖ Ö Ä Ì ÚÖ Ö Ì µ. Ö ´Øµ. satisfying (2). Ö ´Øµ and Ö ´Øµ we have to make sure that Ö ÚÖ Ø Ò Ö Ä When generating a feasible reference trajectory satisfying (2). and possibly their derivatives with respect to time. Find appropriate control laws Ú and of the form Ú Ú ´Ø Ü Ý   µ ´Ø Ü Ý µ (5) such that for the resulting closed-loop system Ø ÐÑ ½ ܴص · ´Øµ   ÜÖ ´Øµ · ¬Ý´Øµ   ÝÖ ´Øµ ·¡ ¬   Ö ´Øµ · ¬ ´Øµ   Ä ´Øµ¬ Ö ¼ (4) Remark 2. ÜÖ ´Øµ. Problem 2. by specifying ÜÖ ´Øµ and ÝÖ ´Øµ the reference trajectory can not be uniquely specified (e. Then also Ö ´Øµ and Ö ´Øµ can be used in the control laws again. ÜÖ ´Øµ. and ´Øµ   Ö ´Øµ converge to zero as Ø goes to infinity. Ö and Ö as functions of time. but Ä ´Øµ and Ö ´Øµ are unknown. But more important is the fact that tracking of ÜÖ ´Øµ and ÝÖ ´Øµ does not guarantee tracking of the corresponding Ö ´Øµ and Ö ´Øµ. due to the fact that Ä is unknown.5 Notice that the time-dependency in (5) allows for using ÚÖ ´Øµ. we need to make the following assumptions on the reference trajectory 2095 in order to obtain a feasible reference trajectory. as mentioned in Remark 2. Ö ´Øµ and ÚÖ ´Øµ are known timeÄ functions. ÜÖ ´Øµ. This explains the time-dependency in (3). 2. once ÚÖ ´Øµ. ÝÖ ´Øµ. Ö ´Øµ in the control laws (as well as their derivatives with respect to time). The only problem is that these signals are unknown. and . Ö ´Øµ and their derivatives with respect to time.2 Adaptive tracking control problem In case the parameter Ä is unknown. Once Ö ´Øµ is known. i. Notice that we can determine Ø Ò Ö ´Øµ. ÚÖ ´Øµ can be either positive or negative). However. Ö is not what we are interested in.g. That is why we insist on looking at the state tracking problem. Ý . e. This is someÄ thing we illustrate throughout by writing Ä ´Øµ and Ö ´Øµ. but ´Øµ not converge to Ö ´Øµ. Ö ´Øµ. Ö Therefore. ÜÖ ´Øµ. ÝÖ . one usually generates some sufficiently smooth reference signals. .e. but in this Ä case we can not use Ö ´Øµ or Ä ´Øµ. satisfying (2) is given Ö and study the problem of finding a state-feedback law that assures tracking of this reference state. due to the fact Ö that Ä is unknown). and Ö ´Øµ without assuming anything on Ä. and assume that Ä is unknown. Problem formulation 129 Consider the problem of tracking a feasible reference trajectory. Ö ´Øµ and Ö ´Øµ are specified it is possible to determine Ä from (4). Actually.6 It is clear that once Ä is known this problem formulation reduces to that of the tracking problem for known Ä. a trajectory ´ ÜÖ ÝÖ Ö Ö Ì ÚÖ Ö Ì µ satisfying ÜÖ ÝÖ Ö Ö ÚÖ Ó× Ö ÚÖ × Ò Ö ÚÖ Ø Ò Ö Ä Ö (2) This reference trajectory can be generated by any of the motion planning techniques available from literature. These signals mainly cover the behaviour of the mobile car. however. Ö Remark 2. find appropriate control laws Ú and of the form Ú Ú ´Ø Ü Ý   µ ´Ø Ü Ý µ (3) such that for the resulting closed-loop system (1. ´Øµ can even grow unbounded. Ý ´Øµ ÝÖ ´Øµ. tracking of the output ÜÖ . Ö ´Øµ and Ö ´Øµ are uniquely determined.e.2. ÝÖ ´Øµ. This is in conflict with the assumption that we do not know Ä.3 Notice that the tracking control problem we study here is not the same as an output tracking problem of the flat output ÜÖ ´Øµ ÝÖ ´Øµ Ì . Remark 2. the control laws (3) are not only a function of Ü. In case we know Ä it is possible. When Ä is unknown we still know that once ÚÖ ´Øµ.2 Notice that in general. ÝÖ ´Øµ and Ö ´Øµ are given. . Notice that it is possible to specify ÚÖ ´Øµ.4 (Adaptive tracking control problem) Let a Ä feasible reference trajectory ´ ÜÖ ÝÖ Ö Ä Ì ÚÖ Ö Ì µ Ö be given (i. we can not formulate the adaptive tracking problem in the same way. ÝÖ ´Øµ and Ö ´Øµ are given. This is due to the fact that for unknown Ä we can not specify a feasible reference trajectory ´ ÜÖ ÝÖ Ö Ö Ì ÚÖ Ö Ì µ. also Ö ´Øµ can be uniquely determined using (2). ÜÖ ´Øµ.3) Ø ÐÑ ½ ܴص · ´Øµ   ÜÖ ´Øµ · ݴص   ÝÖ ´Øµ · ¡   Ö ´Øµ · ´Øµ   Ö ´Øµ ¼ Remark 2. ÜÖ ´Øµ.g. from which Ö ´Øµ is uniquely determined (since Ö has to exist and therefore Ö ¾   ¾ ¾ ).

´µ ´µ Ð Ñ ´Øµ ¼ ½ ´µ Lemma 3. ¼ ½ ¼ ¼ ¼¼ 3 Preliminaries In this section we introduce the definitions and theorems used in the remainder of this paper. i. [16]): Û½ Ø ÛÒ Ø Ì persisDefinition 3. [13. If Ø converges to zero as Ø ½ and its derivative satisfies ´µ ´Øµ ¼ ´Øµ · ´Øµ ´µ ´µ Ø ¼ ´µ where ¼ is a uniformly continuous function and Ø tends ½. Using standard techniques it is easy to show that Lemma 3.   ½ .3 ([15]) Let Ê· Ê be any differentiable function.e. Û ¾ Ê .e. Ü but if we define their values for Ü as respectively . where ¡´Øµ is a bounded vanishing disturbance. . Remark 3. ܾ ܾ ܾ and × Ò´Üµ . the reference dynamics need to have a unique solution. Since ÚÖ Ø ÚÖ we know that if ¡ is a function that fulfills ¼ ´µ ´µ ¼ ´µ min ´Üµ  ÚÖ the control law Ü¾Ê Ø If in addition Û Ø and Û Ø are bounded for all Ø Ø¼ . To overcome the problem that the errors Ü   ÜÖ and Ý   ÝÖ depend on how we choose the inertial reference frame.e. ¾ ÊÑ . e. and Û Ø is persistently exciting then the system (6) is globally exponentially stable.130 Chapter 10.e. This is equivalent to assuming that ´µ where Ü ¾ ÊÒ is globally exponentially stable. then Ø and ¼ Ø tend to zero as Ø to zero as Ø ½. 4 A tracking controller First we consider the tracking problem for the case Ä is known. then Ø is bounded and ´µ ´ µ ´µ In order to be able to control the orientation of our mobile car by means of the input . Adaptive tracking control of non-holonomic systems: an example Assumption 2. we assume that ´Üµ is continuously differentiable and satisfies Ü ´Üµ ¼ Ü ¼ Examples of possible choices for ´Üµ are min ´Üµ ÚÖ ¡ Ø Ò ´Üµ Ü min ´Üµ ÚÖ ¡ ½· Ü Ú ÚÖ With the control law (8) the dynamics in the new coordinates (7) and become Ü Ý Ý  Ü ÚÖ · ´Ü µ ÚÖ Ä ÚÖ · ´Ü µ ÚÖ Ä ÚÖ · ´Ü µ ÚÖ Ö Ä Ä Ø Ò · ´ Ó×   ½µ   ´Ü µ Ø Ò · ×Ò ØÒ   ØÒ (9) Ü ´Ø ܵ ´Ø ¼µ ¼ Ø 2096 . we prefer to have Ú Ø for min all Ø . 14]) Consider the system  × Ò Ó× ¼ × Ò ¼ ¿¾ Ó× ¼ ¼ ½ ¼ ÜÖ   Ü ÝÖ   Ý Ö ¿   (7)  ­Û´Øµ Ì Ñ Ñ ¼ ´Øµ ¼ (6) where ¾ ÊÒ . These functions are discontinuous in Ü . which is why we need Ö Ø ¾   Å Å with Å ¾ . we assume that the reference is always moving in a forward direction with a bounded velocity. ¼ Ú min Ö ÚÖ Ø ´µ Ú max Ö Furthermore.5 Throughout this paper we use the expressions Ü Ó× Ü × Ò´Üµ Ü × Ò´Üµ Ü Ü . ½ Ü × Òܾ  Ó״ܵ . ÚÖ and Ö .g. . ­ . Assume that Ì ×Á    ½ Ñ is a strictly positive real transÅ × Ñ Ñ fer function. there exist min max such that constants ÚÖ and ÚÖ Ö is bounded. ×ÙÔ ¡´Øµ Å and ØÐ Ñ ¡´Øµ ¼ ½ Ø Ü ´Ø ܵ · ¡´Øµ is globally asymptotically stable. and it is easy to verify that all functions are ¾ continuous and bounded. Ó״ܵ ½ .4 Assume that origin of the system · ´Ü µ (8) automatically guarantees Ú ´Øµ ¼ for all Ø ¼. Furthermore.2 (cf. we define errors in a body reference frame.   ¾ . ½ . i. are bounded. i. we assume that the forward and angular acceleration. Then the disturbed system ÚÖ Second. in a coordinate-frame attached to the car (cf.7 First of all. Ó״ܵ ½ . i.1 We call Û Ø such that tently exciting if there exist constants Æ ¯½ ¯¾ for all Ø : ´µ Û ´µ ¼ ´µ ¼ ¯½ Á Ø·Æ Ø ´ µÛ´ µÌ ÑÛ Ì ¯¾ Á ¾ Ü Ý ¿ ¾ Lemma 3.

e.12) are globally uniformly bounded. but also the reference signals Ä ´Øµ and Ö Ä Ö ´Øµ (that appear in the expression ¬ ´Øµ) can not be used in the control law. Corollary 4. ¼   ¾ ÚÖ Ý and   ½   ¾ÚÖ Ó×  ½ Ü · ´ × Ò   ½µÝ · Þ gives also that Ý tends to zero as Ø goes to infinity. Fortunately. which is a direct result from the fact that Þ tends to zero (and Ü . Then all trajectories of (11. 5 An adaptive tracking controller From now on we assume that the parameter Ä is unknown.12) where the reference trajectory satisfies (2) and Assumption 2. Ý . provided that Assumption 2.10. we define ( ½ ¾ ¼): Þ ÚÖ Ø Ò Ö  ÚØ Ò · ½ · ¾ Ö Ú Ó×  ½ Ü · × Ò Ý (10) With this definition the error-dynamics (9) now become Ü Ý Ý Ú Ø Ò · ÚÖ ´ Ó× Ä Ú  Ü Ä Ø Ò · ÚÖ × Ò ½ ¾   Ä   Ä ÚÖ Ó× ·   ½µ   ´Ü µ   ½Ü Ó× · ×Ò Proof: Since Î is positive-definite and radially unbounded. Notice that in (10) we can replace the Ä When we choose the input Ä we obtain ½ «´Øµ × Ò Ó× · Ó×¾ Ú ´¬ ´Øµ · ¿ Þµ (12) Þ   ¿Þ 2097 . we can then claim ¾ ¾ that Ü . and Þ are uniformly bounded. we conclude from (14) that Ü . Also the derivatives of all these signals are bounded.7. Furthermore. It would be the ‘standard procedure’ to define the error variable We establish the following result Proposition 4. for simplicity of analysis we prefer to consider the error variable Þ ÄÞ .3 with . As mentioned in section 2 we have the difficulty that not only Ä is unknown. all closed-loop solutions converge to zero. For the resulting closed-loop system we have Ø (11a) (11b) Ý · (11c) Þ where Ó×¾ Ä Ú ½ Þ Ä ½ ÐÑ ½   «´Øµ × Ò   · ¬ ´Øµ (11d) · ´Øµ   ܴص   ÜÖ ´Øµ · ݴص   ÝÖ ´Øµ · ¡ Ö ´Øµ · ´Øµ   Ö ´Øµ ¼ «´Øµ ¬ ´Øµ Ý Ø Ò · Ò Ö ½   Úµ Ø Ò ·  ½ · ¾ ÚÖ Ý × Ò · · ¾ ´ÚÖ Ü   ÚÚÖ   Ú · ½ ÚÖ Ø Ò Ö · ¾ ÚÖ ½  × Ò ¾   Ó× Ü · Ä · Ó× ¾ × Ò Ý ÚÖ Ø Ò Ö ÚÖ Ø · Ó×¾ Ö Ö ÚÖ   ¾ ÚÖ ½  Ó× Ü ¾ · ´ÚÖ Ó× ¾ Ó× Öµ ·  ×¾Ò Ý Proof: Using (7) it follows from Ü and Ý tending to zero that also Ü   ÜÖ and Ý   ÝÖ converge to zero.2 Consider the system (1) in closed loop with the control laws (8. Ý and converge to zero. Ý and ). An adaptive controller 131 Differentiating the function ν lutions of (9) yields ½ ¾ ¾ Ü · ½ ¾ ¾ Ý along the so- Consider the Lyapunov function candidate ν  Ü ´Ü µ · ÚÖ Ó×   ½Ü   ½Ü · ×Ò Ý Î¿ ο ½ ¾ ½ ¾ Ä Ü · Ý · ¾ ¾ ¾ ¾ ¾ · ½ ¾ ½ ¾ ¿ Þ¾ Þ¾ (13) Differentiating (13) along solutions of (11. Ø Þ ÚÖ Ø Ä · Ò ¾ Ö Ú ÚÖ · ´Ü µ Ø Ò · Ö  Ä Ó×   ½ ×Ò Ü · Ý ÐÑ ½ · ½ Ü ´Øµ · Ý ´Øµ ·   ´Øµ · Þ ´Øµ ¡ ¼ However. Ý .5. Ø Ò and as a result also and   Ö are uniformly bounded. i.7 it follows that also Ú . It only remains to show that ´Øµ   Ö ´Øµ tends to zero as Ø tends to infinity.12) yields When we consider as a virtual control we could design an intermediate control law for that achieves ( ½ ¾ ¼):   ½   ¾ ÚÖ Ó× · ×Ò Ý  Ü ´Ü µ   ½ ¾ · ½¾ Þ   ¾  Ü ´Ü µ   ¾ ½¾ ¾   ¾ ½ ¾ Þ ¾ ½ ½ ¾ ½ ¼ (14) ½ ܾ · ½ Ý ¾ · Using the Lyapunov function candidate ξ ¾ ¾ ½ ¾ and similar reasoning as in [2].7 is satisfied. With Barbalat’s Lemma it foland Þ converge to zero as Ø goes to infinlows that Ü .1 Assume that Assumption 2. From (10) and Assumption 2. Using Lemma 3. we are not only allowed to use Ü . ity. but also Ö and Ö . This comes down to showing that Ø Ò ´Øµ   Ø Ò Ö ´Øµ tends to zero as Ø tends to infinity. i.7 is satisfied. and .e.

using the variable Þ as defined in (10) or (15) makes it hard to design a controller using conventional adaptive techniques because Þ includes the unknown parameter Ä. Also. As a result ÐÑ ¬ ´ µ Ø ½ ¬ Ø   Ä ¬ Ö´ µ Ø ¬ ¼ When we choose the input ±«´Øµ × Ò we obtain (± Ó× · Ó×¾ Assume that in addition Ö ´Øµ is persistently exciting. Adaptive tracking control of non-holonomic systems: an example occurence of Ä Ö by means of the signal · ½ · Ö: · So.2 we know that the LTV subsystem of (21) is globally exponentially stable and therefore.20) are globally uniformly bounded. we obtain that Ä Ö   Ú Ø Ò and therefore ÚÖ Ø Ò Ä   Ø Ò conÖ verges to zero. . if we define the parameter-update-laws Þ Ä Ö  ÚØ Ò ¾ ÚÖ Ó×  ½ Ü ×Ò Ý (15) we get Ä ± However.1 Assume that Assumption 2. · «´Øµ ¬ ´Øµ Ý Ä · · Ø Ò Ö· · ½   ¾ ÚÖ ½  Ó× Ü ¾  ×¾Ò Ý · Ä ¾ ´ÚÖ Ü ½ Ö· Ö · ´ Ö Ó× ¾   ÚÚÖ   ÚÖ µ Ö Ú · ¾ ÚÖ Ó× ¾  × Ò Ý   Úµ Ø Ò · Ó×  ½ · ¾ ÚÖ Ý × Ò ½  × Ò ¾   Ó× Ü · Proof: Similar to the proof of Proposition 4. the ´ ĵ dynamics can be seen as a cascade of a LTV subsystem with an additional disturbance that is bounded and goes to zero as Ø goes to infinity: ¾­½ ½ Ä   ­½ ¾ ¾  Ö ½ ¾ ½¾ Þ¾ · Ä Ò Ä· ± 2098 ­½   Ľ Ö´ µ Ø ½   Ä Ö ´Øµ ¼ Ä · ¿´Øµ ´Øµ ½ ¾ ¿ ­¾ ½ ±   ­¾ Þ«´ØµÚ Ø In the same way we can conclude that also Ä tends to zero as Ø tends to infinity. Notice that the ´Ü Ý µ-dynamics (17a. From Barbalat’s Lemma it follows that Ü . Then all trajectories of (17. From (16) we conclude that Ä   Ú Ø Ò · ¾ ÚÖ Ý converges to zero too. ½ Furthermore.132 Chapter 10. . we define ­½ Ö ­¾ Þ«´ØµÚ Ø ¾  (20a) Ò (20b) Î  Ü ´Ü µ   ¾ ½ ½ ¾ Þ Ä Ö  ÚØ Ò · ½ · ¾ ÚÖ Ó×  ½ Ü · ×Ò Ý (16) ¾ ½¾ Þ¾ ¼ and can establish the following result Proposition 5.7 is satisfied. ÐÑ Ø which can be seen as an estimate for Þ .1 we can show uniform boundedness of all signals and their derivatives with respect to time.18.4 that also Ý tends to zero as Ø tends to infinity. Combining these two results. Using Þ the tracking error dynamics ´ µ can be expressed as Ü Ý Ý Ú Ø Ò · ÚÖ ´ Ó× Ä Ú  Ü Ä Ø Ò · ÚÖ × Ò ½ ¾   Ä   Ä ÚÖ Ó× ·   ½µ   ´Ü µ   ½Ü   ± · ×Ò (17a) (17b) ½ Ü ´Øµ · ´ µ · Ø Þ´Øµ ¬ ¬ · ´ µ Ø   ¡ Ö´ µ Ø ¼ If in addition Ö ´ µ is persistently exciting.17b) can also be seen as a LTV subsystem with an additional disturbance that is bounded and goes to zero as Ø goes to infinity: (18) ±   ±): Þ Ú ¬ ´Øµ · ¿ Þ Ò Ü Ý     Ö ´Øµ ßÞ Ö´ µ ¼ Ø Ü Ý · ½ ´Øµ ¾ ´Øµ ßÞ   ¿ Þ   ±«´ØµÚ Ø · LTV subsystem disturbance (21) Consider the Lyapunov function candidate (­½ Î ½ ľ ܾ · ݾ · ¾ ¾ ¾ ¾ ½ ¾ Þ¾ ½¾¿ ľ · ¾ ¾ ­½ ­¾ · ¾ ¼) ±¾ ½ ¾ ¿ ­¾ (19) Differentiating (19) along solutions of (17. from Lemma 3. Furthermore. Therefore. and Þ converge to zero as Ø goes to infinity.3 we can conclude that also ¾ ÚÖ Ý · Ä Ö converges to zero. Using Lemma 3. we also have that Ø Ý · Þ Þ Ä Ö Ä Ä Ú ¾ ´±«´Øµ × Ò Ó× ½ ½ ÐÑ Ø ½ Ý ´Øµ   · ¬ ´ µ¬ · ¬ ¬ ÄØ ±´Øµ ¼ (17c) Ó× µ· ¬ ´Øµ (17d) where we introduced the parameter we defined Ä Ä   Ä and Ä.18) yields Î  Ü ´Ü µ   ¾ ½ · · From Lemma 3.

. Grenoble. Prentice-Hall. INRIA. Rep. June 1995. December 1997.-S. and M. Jiang and H. convergence and robustness.2 Consider the system (1) in closed loop with the control laws (8. no. “Adaptive control of nonholonomic mechanical systems. R. A. 2099 Þ ± ­¾ Ä Ö ´Øµ   ¿   ½Ä ¼ Ö ´Øµ Þ ± · ´Øµ ´Øµ Therefore. also ± tends to zero as Ø tends to infinity. Tomei. pp. [3] E. vol.18) where the parameter estimates Ä and ± are updated accoding to (20) and assume that the reference trajectory satisfies (2). California. This simple example clearly illustrates that for the general state tracking problem specifying the state trajectory to be tracked and not knowing certain parameters can be in conflict with each other. R. “Adaptive output linearizing control of wheeled mobile robots.” in Proceedings of the Third IFAC Symposium on Nonlinear Control Systems Design (NOLCOS’95).6. Springer-Verlag. Sastry and M. second edition. Samson. vol. [2] Z. [16] Y. Assumption 2. 1993. pp. which concludes the proof. Nijmeijer. USA.-P. USA.” European Journal of Control. Micaelli and C. pp.” in Proceedings IEEE International Conference on Robotics and Automation. Miyazaki. “Asymptotic stability and feedback stabilization. Nijmeijer. 4739–4744. Samson. Noguchi. no. and H.” in Stability and Stabilization of Nonlinear Systems. Englewood Cliffs. USA. [5] R. 1983. 1999. Corollary 5. number 246 in Lecture Notes in Control and Information Sciences. Y. Robertsson. Khalil. pp. References [1] A. pp. pp. For the resulting closed-loop system we have Ø ÐÑ ½   · ´Øµ ܴص   ÜÖ ´Øµ   Ö ´Øµ · · ¬ ´Øµ ¬ Ý ´Øµ   ÝÖ ´Øµ · ¬¡   Ä ´Øµ¬ Ö ¼ and convergence of the parameter-estimates to their true value. pp. Prentice-Hall. Lamnabhi-Lagarrigue.K. “Trajectory tracking for unicycle-type and two-steering-wheels mobile robots. Kimura. also the ´Þ ±µ dynamics can be seen as a cascade of a LTV subsystem with an additional disturbance that is bounded and goes to zero as Ø goes to infinity: Schaft. and H. Jiang and J. Campion and R. [15] A. pp.” in Proceedings of the 36th Conference on Decision and Control. Sampei. and A. Millman. Bodson. December 1996. Kolmanovsky and N. F. “Robust adaptive statefeedback tracking for nonlinear systems. Aeyels. “An adaptive control of a nonholonomic space robot. also a solution was derived.e. M. Chang and B. “Adaptive tracking control design of nonholonomic mechanical systems. Brockett. Nakagawa. Adaptive Control: stability. “Global stabilization of parametric chained-form systems by time-varying dynamic feedback.” in Proceedings of the European Control Conference. Marino and P.-C. F. vol. H. 84–89. 6. vol.-P. pp.S. 2097. Bastin and G.” IEEE Control Systems Magazine. Campion. Japan. pp. 1. pp. France. Astolfi. California. Pomet. 183–199. USA.J. “Exponential stabilization of nonholonomic systems via discontinuous control. Lefeber. “Application of backstepping techniques to the time-varying exponential stabilisation of chained form systems. 1996. 1393–1399. 47–59. i. 1997. pp. “Output feedback tracking of nonholonomic systems in chained form. Eds. Upper Saddle River. September 3–5 1997. San Diego. 1997. [14] S. [10] Y. We propose a formulation for the adaptive tracking problem that is such that it reduces to the tracking problem in case the parameters are known. 10. NJ. Nonlinear Systems. [4] P. Ø ÐÑ ½ ¬ ¬ ¬Ä´Øµ ½   Ĭ · ¬±´Øµ   Ä ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¼ 6 Concluding remarks In this paper we addressed the problem of adaptive state tracking control for a four wheel mobile robot with unknown length. 181–191. vol. pp.” Automatica. 1334–1338.-B. 741–746. [11] R. [9] G. Koga. United Kingdom. Nantes. New Jersey.” Tech. 1989. Concluding remarks 133 Since we have shown that Ý tends to zero. Birkh¨ user..J. 3.” IEEE Transactions on Automatic Control. a [6] I. USA.H. [8] Z. Morin and C.” International Journal of Adaptive Control and Signal Processing. 7. Hossa. MA. van der . Information and System Sciences Series.7.10. Boston. and that Ö is persistently exciting.W. Tahoe City. 15–36. Sussmann. 33. 20–36. July 1991.” in Symposium on Robot Control. Kanayama. 1996. “A stable tracking control scheme for an autonomous mobile robot. France. Kiyota. Brockett. Eds. no. “Tracking control of mobile robots: a case study in backstepping. 261–266. and T. Chen.W. January 1998. December 1995. Not only did we formulate the problem.” in Proceedings of the 35th Conference on Decision and Control. 3632–3633. London. McClamroch. “Developments in nonholonomic control problems. 43. Kobe. [13] H. 1990. 384–389. [12] T. [7] G. D. 16.” in Differential Geometric Control Theory.

134 Chapter 10. Adaptive tracking control of non-holonomic systems: an example .

and numerous other examples. mobile robots. We introduced the cascaded design approach as a new method for controller design. also a factory following a predetermined production schedule can be thought of as a tracking problem. Since the tracking control problem for arbitrary systems is too complex to be solved in general. the controllers derived by means of this method are less complex expressions. Under a persistence of excitation condition on one of the two reference inputs we arrived at globally Ã-exponentially stable tracking error dynamics. all we can do is restrict ourselves to classes of systems with a specific structure. As a result. Next. autopilots for airplanes. and an underactuated ship. One of the advantages of this method is that it is not necessary to transform the system into a certain form by means of a change of co-ordinates. including a rotating rigid body. we can say that backstepping assumes a triangular structure of the system and uses this structure to arrive 135 . This is a type of uniform stability which guarantees a certain robustness to disturbances. The cascaded design approach was applied successfully for several classes of mechanical systems.Chapter 11 Conclusions From a practical point of view. a spacecraft moving along a predetermined path. In case we compare the cascaded design to a backstepping design. mobile cars in the harbor driving a prescribed trajectory. part of the remaining dynamics can be ignored. One not necessarily has to think of robots. One can think of robots that have to perform a desired motion. Backstepping is a widely used controller design method for nonlinear systems with a triangular structure. the problem of making a system follow a certain trajectory is interesting. the second input can be used for stabilizing this simplified tracking error dynamics. chained-form systems. 11. This thesis is concerned with certain nonlinear mechanical systems described by continuous-time models. The key idea is to use one input for stabilizing a subsystem of the tracking error dynamics.1 Discussion In Part I of this thesis we studied the tracking control problem for several non-holonomic systems with two inputs. Assuming that this stabilization has been achieved.

136 Chapter 11. As a result. These controllers yield global uniform asymptotic stability and partially deal with input saturations. For a cascaded design the problem can be overcome. In all the examples studied in the first part of this thesis we were able to perform the analysis in the original error co-ordinates. what is most important in our eyes is not the triangular structure of the open-loop system. Also the extension to other problems is straightforward. the gain tuning turned out to be not very difficult. In case the linearity assumption on Ø Þ½ Þ¾ is not satisfied. A triangular structure for the open-loop system not necessarily has to be assumed beforehand. but the triangular structure of the closed-loop system.4. the open-loop system not necessarily has to be in triangular form to obtain triangular closed-loop dynamics. However.6. First of all. no assumption on Ø Þ½ Þ¾ is needed. this triangular structure for the closed-loop system readily follows starting from a triangular open-loop system. the main contributions in Part I are: ¯ ¯ ¯ ¯ we introduce the cascaded design approach. These controllers yield global uniform asymptotic stability and also deal with input saturations. In all the examples studied in the first part of this thesis. we present controllers that solve both the state. At first glance this might seem to be a restriction to the general applicability of the cascaded approach. the usage of the linearity assumption on Ø Þ½ Þ¾ is to guarantee boundedness of the solutions of the cascaded system. This is exactly what the cascaded approach does. . is interested in an À½ design. one simply has to solve two linear problems instead of solving partial differential equations that come from the nonlinear À½ control problem. the connecting term Ø Þ½ Þ¾ could be upper bounded by a linear function of Þ½ .4. we present controllers for solving both the state. which does not require the system to be transformed into a specific structure. since we could rely on linear techniques. Notice that all analysis can be done in the original error co-ordinates. ´ µ ´ µ ´ µ ´ µ To summarize. similarly.and output-feedback tracking problem for chained-form systems with two inputs. Whenever one is able to show boundedness of solutions. leading to a clear structure of the closed-loop dynamics and to much simpler expressions for the control laws (in the original co-ordinates) than achieved so far by means of backstepping. Another way to overcome the problem would be using different co-ordinates. However. one way to proceed might be to use the original cascaded theorem (Theorem 2. Using backstepping. These controllers yield global uniform asymptotic stability. Using as new co-ordinates the difference between the actual value and the desired value of the virtual control assures a linear connecting term in a backstepping design. as is done in backstepping. This is not the case. transformation of the open-loop system for obtaining a triangular structure suitable for backstepping is not necessary. As a result. This is where the cascaded design approach comes into play. All we focus on. leading to simpler controllers than found so far.3) instead of Corollary 2. If one. for instance. without using state-feedback or input transformations.and output-feedback tracking problem for mobile robots. For the systems under consideration. Conclusions at a triangular closed-loop system in a systematic way. we present controllers that solve the tracking problem for under-actuated ships by means of state-feedback. is a triangular structure for the closed-loop system. the nonlinear tracking control problem essentially reduced to two linear stabilization problems.

viewed from above with a camera. For this example a suitable problem formulation of the adaptive state-tracking problem is given and a solution is presented. Once a system is written in a triangular form that is suitable for backstepping.and outputfeedback case. Secondly. we illustrate difficulties in formulating the adaptive state-tracking problem for nonlinear systems with unknown parameters by means of an example. As an extra difficulty we took into account the possibility that both the camera position and orientation are unknown. In case of unknown camera orientation a class of adaptive controllers is presented. we considered a robot manipulator operating in the plane.and output-feedback problem a “start-integration time” exists such that global asymptotic stability is guaranteed.2. Further research 137 In the second part of this thesis we studied three specific problems. and of which an image is displayed at a screen. Thirdly we addressed the problem of adaptive state tracking control for nonlinear systems. To summarize. Furthermore. a control law can be built step by step. We showed this difficulty by means of an illustrative example of a four-wheel mobile car with unknown length. 11. the main contributions of Part II are: ¯ ¯ ¯ we show global asymptotic stability of linear PID controllers when delaying the integral action. we addressed the visual servoing problem of planar robot manipulators under a fixed camera configuration. We showed that for both the state.11. we introduce classes of controllers that solve the visual servoing of planar robots under a fixed camera position for both the state. Finally. In case the camera orientation is known. We were able to regulate the tip of the robot manipulator to a specified point at this screen using only position measurements. It turned out that formulating the adaptive state-feedback tracking control problem is not straightforward. we addressed the practically important problem of global set-point stabilization of robot manipulators with PID control. These classes also contain saturated controllers. We also presented criteria on how to choose the start-integration time and the control gains. as well as certain intrinsic camera parameters (like scale factors. We formulated the adaptive state-feedback tracking control problem in such a way that it reduced to the tracking problem in case the parameters are known. we showed in simulations the potential advantages of our linear scheme in comparison to existing nonlinear controllers. adaptive controllers were presented. The fact that we . since specifying the reference state-trajectory can be in conflict with not knowing certain parameters.and output-feedback problem. the cascaded design approach can be seen as a one-step vectorial backstepping design where part of the dynamics can be forgotten. In case of unknown camera orientation. In a sense. That is. focal length and center offset). we introduced a class of visual servoing controllers for both the state. First. we presented a general methodology for solving the problem by means of Lyapunov techniques.2 Further research One of the key properties of the backstepping design methodology is its recursive nature.

¿ ¿ For all systems studied in the first part of this thesis it turned out that we could conclude global Ã-exponential stability. the cascaded design approach reduced the nonlinear tracking control problem to two linear stabilization problems. Conclusions do not have a Lyapunov function for showing stability of the overall closed-loop system can be seen as a shortcoming of the cascaded design approach (or at least of Theorem 2. This led us to a clear structure and gave us a simple strategy for tuning the gains. provided that one of the reference inputs was persistently exciting. 1999b).4. we could add an (almost) arbitrary term Ø Ü Ý to Ú .3. It is worth investigating whether the concept of uÆ -PE can be successful in weakening the PE condition for general chained-form systems and the under-actuated ship.3 difficult. instead of applying the cascaded theorem for concluding asymptotic stability. As it happened to be.6 this can be done for the mobile robot. one could view the cascaded design approach as one-step vectorial backstepping. and leads to a weakening of the persistence of excitation condition on one of the two reference inputs. Another interesting question concerns the actual controller design. namely the gain-tuning.4. It is worth investigating whether this idea leads to similar results for the chained-form system and the under-actuated ship.4.2). As a result. In a similar way changes to the control laws for the chained-form system or the under-actuated ship can be constructed. It would be interesting to perform this exercise and study the resulting performance and compare this to existing results. It is interesting to determine whether this is purely a coincidence or that it can be explained. In that way the cascaded design approach can be applied recursively (as in the proof of Proposition 4. Since the mobile car can be transformed to a chainedform system of order the stabilization problem for a mobile car can be solved in a similar way. since the design methodology as proposed also turns out to be an eye-opener to recognizing structures that had not been noticed before. since it makes a recursive application of Theorem 2. As long as the function can be bounded by a linear function of Ü Ý Ì the proof still follows the same lines. that we have more freedom in choosing the control laws. the reference input that has to be persistently exciting. the from a practical point of view more interesting methods like using filtered measurements. As mentioned in Remark 4. adding integral action or using À½ -control can be applied without any difficulties.3 could be extended by showing the existence of a Lyapunov function candidate satisfying Assumption A1 along trajectories of the overall closed-loop system. one could also apply backstepping and redesign the control law for the input that was used for first stabilizing a subsystem. For the mobile robot for instance. For the systems considered in the first part of this thesis. Another idea to weaken the persistence of excitation condition on one of the two reference inputs could be using the concept of uniform Æ -persistence of excitation (uÆ -PE) as introduced in (Lor´a et al.3). This extra freedom enables us to improve performance and it is worth investigating how this freedom can be used in designing a well performing control law. It would be interesting to find out whether Theorem 2. The method we used in this thesis is simply applying optimal control to the resulting linear systems. However. ´ µ As mentioned in the beginning of this section. Notice however. In that paper the stabilization problem for a chained-form system of ı order was solved using this concept. once a cascaded design leads us to a simple structure.138 Chapter 11. However. the cascaded design approach is more than that. This observation .2. also was the first input that we used for stabilizing part of the tracking error dynamics.

since that would imply only that the reference has to “keep on moving”. the adaptive tracking control problem is even more interesting to study. Further research 139 might also be a third way to overcome the “problem of the PE-condition”.e. then conclude GUAS of the overall system under a PE-condition on ÚÖ ). The main reason why this PE-condition on the reference input is annoying.11. In Chapter 10 it was shown. first use Ú to stabize the Ü -dynamics. A PE-condition on the other reference input would be less annoying. Therefore. before we are able to do so. in practice we always have to deal with the fact that certain parameters are not known exactly. As for the mobile robot this can indeed be done (i. . we need to have a correct problem formulation. solving the problem is even more interesting. However. the problem of following a prescribed trajectory is important from a practical point of view. As mentioned in the beginning of this chapter.. by means of an illustrative example. Obviously. arriving at a proper problem formulation is an interesting problem in itself that deserves attention. that the formulation of the general adaptive (state)-tracking problem is a problem in itself. it would be worth investigating if this idea can be applied to the tracking problem for the under-actuated ship as well. is that for the mobile robot and the under-actuated ship it makes it impossible to follow straight lines.2. then use for stablizing the simplified remaining dynamics. And once we have a proper problem formulation. Unfortunately.

Conclusions .140 Chapter 11.

¿ ¾ ½ ¼ ¼ ¼ . ¡ ¡ A proof of this result by means of the second method of Lyapunov is less known (see also (Morin and Samson 1997)). .1) For this system we can define the Hurwitz-determinants ¬ ¬ ½ ¬ ¬ ¬ ¬ ¬ ¬ ¬ ¬.1) is asymptotically stable if and only if the determinants are all positive. ¡  ¿¡ ¡  ¾ ¡  ½ ¼ . . .7. . . If we define ½ ¡½ ¾ ¡¾ ¡½ ¾ ¿ ¡¿ ¡½¡¾ ¼ . . . ¼ ½  Ñ ¼ ¼ Û (A.. . . first we consider the stability of the differential equation dÑ Ý dØÑ ´Øµ · dÑ ½ ½ dØÑ ½ ݴص · ¡ ¡ ¡ · Ñ ½ d Ý dØ ´Øµ · ÑÝ ´Øµ ¼ (A..3.1) can also be represented as  ½  ¾ ¼ Û ½ ¼ ¼ . .. ¬ ¬ ¬ ¬ ¬ ´ ½ ѵ where. . .3. Theorem 2..2) 141 . ¿ ´ ѵ as was shown by Parks (1962).. ¿ ¾ ¾  ½ ¬ ¬ ¾  ¾ ¬ ¬ ¾  ¿ ¬ ¬ ¾  ¬ ¬ . It was shown by Hurwitz (1895) that the system (A. . it is assumed to be zero. . . . . . .8 and Proposition 6. However. ¬. .. if an element appears in with .1.. ¬. .. . . .Appendix A Proofs We present the proofs of Theorem 2. .3. ¬ ¡ ½ ¼ ¼ ½ . the system (A.

.. . we can write ·× ·¾ Û ·¾ · × · Û · · (A. . . ¼ ¼ . . ¼   . From the structure displayed in (A.2) results in ¡ Ñ ÛÑ ¾ are all positive) along solutions Î   ¾ Û½ ½ ¾ ËÛ. that trans- Asymptotic stability then can be shown by invoking LaSalle’s theorem (LaSalle 1960). Þ (A. we define ÞÑ ¾ . . .142 Differentiating the Lyapunov function candidate Appendix A.. . and we would like ÞÑ ÞÑ ½ Since ÛÑ ½ ÛÑ ¾   Ñ ÛÑ .4) ½ Þ½  «½ Þ½   «¾ Þ¾   ¡ ¡ ¡   «Ò ÞÒ Ñ « ¾Ê ´ ½ ѵ The characteristic polynomial of the transformed system then becomes · «½ · ½ Ñ ½ · ¡ ¡ ¡ · «Ñ ½ · «Ñ ·¡¡¡ · Ñ ½ Since a state-transformation does not change the characteristic polynomial and we know from Parks (1962) that the characteristic polynomial of (A. and we would like ÞÑ ½ ÞÑ ¾ Þ Proceeding similarly.2) equals Ñ Ñ ½ · Ñ .3). .. . we define ÛÑ ½ ÛÑ ¾   Ñ ÛÑ Û ÛÑ ½ . Proofs Î ¾ ¾ ¾ ½ Û½ · ½ ¾ Û¾ · ¡ ¡ ¡ · ½ ¾ ¡ ¡ ¡ Ñ ½ ÛÑ ½ · ½ ¾ ¡ ¡ ¡ (which is positive definite if and only if the determinants of (A. .. . Therefore. we define all Þ and obtain an expression that looks like By this construction of the state-transformation.. we are guaranteed to meet the Ñ   final equations of (A. .2) into ¾  ½  ¾ Þ ½ ¼ .4) we know that the matrix Ë is nonsingular. . Inspired by the result of Parks (1962) we look for a state-transformation Þ forms the system (A. . we define ¼ ½ ÞÑ Since ÛÑ ÛÑ ÞÑ ½.3) ¼ To start with. Ñ ¿ . The only thing that remains to be verified is whether the equation for Þ½ holds.

.6) We can also write the system (A. ¼ Û ½ ´Øµ ¼ ¼ . . . . we obtain ¾ ¿ ¾ ½ Û½ ¿ ¼ . . ½ £ £ . When we apply the change of co-ordinates Þ ¾ ¼ .. .5) Û½ Þ½ · ¿ Þ¿ · Þ ½ ½ · Knowing this state-transformation and (A. . . . .7 and Theorem 2. . ¼ . Therefore.7 clearly « ( 143 ½ Ñ). . . . . . .. . .. . Þ · ´ ´Øµ   ½µ ¼ ¼ . ½ .3.. A.. . which is described by ¾   ½   ¾ ´Øµ   ¿   ´Øµ ¿ Þ ´Øµ ¼ ¼ .7 and 2. . .8. Before we can prove Theorem 2.A.. .. . . Ñ ¿ ½ Þ½ · ¿ Þ¿ · ¿ . .. ¼ . . ¼ ´Øµ .5) we can start proving Theorems 2. .. Û½ Þ½ · ؽ ¿ Þ¿ · ؽ Þ · ´  ½Þ½   ¾Þ¾   ¡ ¡ ¡   ÒÞÒµ · ؽ ¿Þ¾ · ؽ Þ ·   ½Þ½   ¿ Þ¿   · ´Ø½ ¿   ¾ µÞ¾ · ´Ø½   µÞ So obviously · (A... . .3..  ½  ¾ . . . . .. .6) as ¾ Þ ½ ´Øµ ¼ ¼ . . . .1.3. . . . . . ¼ ¼ ¾ . .8 we need to remark one thing about this transformation. .12). . £ .1 Proof of Theorem 2. . . ¼ . We need to show global uniform exponential stability (GUES) of the system (2. . . When we define Ì Ë  ½ . ¼ Û · ´ ´Øµ   ½µ . .. . ... ¼ ¿ ËÛ as defined before.. .7 Proof... ¼   . . Proof of Theorem 2.. .. Þ (A. ¼ . we know that Û½ Þ½ · ؽ ¿ Þ¿ · ؽ Þ · Û¾ Þ¾ · ؾ Þ · ؾ Þ · But also Û½   ½ Û½   ¾ Û¾ (notice that ½ ½ ). . .   Ñ ¼ ¼ ½ ¼ ½ ¼ .3. .. . ¼ .3..3. .  ½  ¾ ¼ . . .

....9) is uniformly completely observable (UCO). . . .144 which (using Appendix A. . . . .. it follows immediately from Corollary 2.4 that the pair (A. . .9) is UCO.. If Ø is persistently exciting. ..3.. . . ... ¼ . ¼ . ¿ .8) along solutions of (A. Differentiating (A. . ¼ ´Øµ   Ñ ´Øµ ¼ ¼ ½ ¼ ¼£ (A. . . Proofs ½ ½ ) can we rewritten as ¾   ½   ¾ ´Øµ ¼ Û ´Øµ ¼ ¼ .. . . .. ¼ . . . ´µ .. . .8) which is positive definite if and only if · ½ Ñ ½ ·¡¡¡·   ¾Û½ ½ ¾ Ñ ½ · Ñ is a Hurwitz-polynomial... .. . ¼ ´Øµ   Ñ ´Øµ ¼ ¼ Û (A. . . . . . ¿ ¢ ½ . which completes the proof.7) is globally uniformly exponentially stable (GUES) if the pair ¼¾   ½   ¾ ´Øµ ¼ ´Øµ ¼ ¼ . It is well-known (Khalil 1996) that the origin of the system (A.7) Consider the Lyapunov function candidate Î ¾ ¾ ¾ ½ Û½ · ½ ¾ Û¾ · ¡ ¡ ¡ · ½ ¾ ¡ ¡ ¡ Ñ ½ ÛÑ ½ · ½ ¾ ¡ ¡ ¡ Ñ Ñ ÛÑ ¾ (A. ¼ .. . .7) results in Î which is negative semi-definite. . . .

4. . . ´µ ´ µ ´ µ Then the result follows immediately from Corollary 2. . . . .. . . ¿ ¾ ¼ ¼ . ¼   . .14) are Hurwitz.. .  Ð . .2.3.3.9). .   . . Þ½ · ¼ ¼ .. . . . ´ µ . ..6. . . . . . Proof of Theorem 2. .7 that the systems Þ½ ½ Ø Þ½ and Þ¾ ½ Ø Þ¾ are globally uniformly exponentially stable (GUES).. . ¿ ´Ø Þ ßÞ ½ Þ¾ µ ´Ø Þ µ ¾ where Þ½ and ¢ ܾ ܾ Ü¿ £ ÜÒ Ì Þ¾ ¢   ܾ Ü¿   Ü¿ ÜÒ   ÜÒ £Ì Since Ø is persistently exciting (PE) and Ð are such that the polynomials (2. .. .8 145 A. .. .  ¾  ¿ .A. .3. ¼  Ð¿ ¼ ßÞ  Ð¾ ¾ ¼ .     ¿ ¼ ¼ . . Þ¾ ½ ´Ø Þ µ ½ ¼ ßÞ ¼ ¼ ¼ .8 Proof. .25)) and the fact that a LTV system which is globally uniformly asymptotically stable (GUAS) is also GUES (Theorem 2. . Þ¾ . .2 Proof of Theorem 2.3. . .. The system (2.. . .13) can be written as ¾  ¾  ¿ Þ½ ¼ ¼ ¾ . .  Ð Þ¾ . . we know from Theorem 2. (since Ø Þ½ Þ¾ satisfies (2.

13d) (A.10a) and the fact that ¾¾ ѽ½ Ô ½ · ½½ ѽ½ ¿ ¾¾ ¼.13e) (A.1 Before we prove Proposition 6. First. Let the following conditions be given: ½ ¾ ¼ Define and ( ¿ ´ · ½ · ½½   ¾¾µ ´ ¾¾ · ѽ½ ¿µ ´ ½ · ½½   ¾¾µ Ѿ¾ ½½ ¼ ѽ½ Ѿ¾ ¾¾   ½½ (A.13c) (A.1.13a) (A.1 we first prove the following lemma: Lemma A. Proofs A. Then obviously ½ · ½½ ѽ½ ¼ Ô ¾¾ ѽ½ ѽ½ ¾¾ ѽ½ ¾   ´ ½ · ½½ µ Ѿ¾ · ¿ ´ ¾¾   ½   ½½ µ Ѿ¾ · ¿ ¼ ½½ ½½ .11a) (A.146 Appendix A.3 Proof of Proposition 6.3. we remark that from (A.12) Դܵ Then and ѽ½ ܾ   ´ ½ · ½½ µÜ · ¿ are well-defined.3. and furthermore ¼ ¼ ¼ ¼ ¼ ¼ ¼ Ô´¼µ and   ¾¾   ѽ½ ѽ½   ¾¾ Ѿ ½½ (A.13b) (A.10d) ) by means of · which is similar to saying that and ½ · ½½ ѽ½ ¿ ѽ½ (A.10b) (A.13f) · ѽ½ · ѽ½ · ¾¾ Proof.10a) (A. ѽ½ ¼ we have: Consider the polynomial (A.10c) (A.3.11b) are the roots of the polynomial (A.12).

13). such that Ô ½½ we obtain that and are well-defined by means of (A.10b). From (A. we obtain ¾ ܽ ¿ ܾ Ü¿ Ü ¾ Ѿ¾ ´ ·Ñ½½ ѽ½ ¾¾ Ѿ¾ ´ ·Ñ½½ ѽ½ ¾¾ ¼ ¼ ¾¾   ½½   ѽ½ ´   ¾¾ ÖÖ   Ѿ¾ ´Ñ  µ ÖÖ Ñ¾¾ µ ѽ½ Ñ   Ѿ¾   Ѿ¾ ´   µ ÖÖ Ñ¾¾ ´½½  µ ÖÖ ¾¾ µ ·   ¼ ÖÖ Ñ¾¾ ´ ¾¾ ´Ñ½½ µ´·¾¾  Ñ½½ µ ÖÖ Ñ¾ µ ¾¾ ½½ µ Ѿ¾ ´ ·Ñ½½ µ´Ñ½½   ¾¾ µ ÖÖ   ÖÖ ¼   ¾ · ¾¾ µ ¾¾ ´Ñ½½ ¿ ¾ ܽ ¿ ܾ Ü¿ Ü (A.3. (A. from the intermediate value theorem we know that a constant exists. due to (A.A.3. such that Ô Ñ½½ .14) can be Ù Ú Þ½ Þ¾ ¿  ´ · µ   Ñ ÖÖ ´Øµ Ñ ½½ ¾¾ Ѿ¾ ´ ·Ñ½½ ѽ½ ´Ñ¾ ½½ ½ ¼  Ñ µ´ ·Ñ µ ÖÖ ´Øµ · µ ½½ ¼ ½ ¾¾ ¾¾ ¾¾    ÖÖ ´Øµ   ¿¾ ¿ ¼ ¼ ѽ½ ÖÖ ´Øµ ¿¾ ¼ ÖÖ ´Øµ ¼ Ù Ú Þ½ Þ¾ ¿ Using the change of co-ordinates ¾ Ù Ú Þ½ Þ¾ ¿ ¾ Ѿ¾ ¾¾ ¼ ¼ ¼  Ñ ¼ ¼ ½ ¼ ¾¾ ¾¾   ¼  ½   ¼   ¼ ½ ¼   ܽ ܾ Ü¿ Ü which. Proof of Proposition 6. ¾¾ ½ · ½½ and also a .11.1 147 Therefore.12) is given by ¾ Ù Ú Þ½ Þ¾ ¿ ¾ ·   Ñ Ñ Ö ´Øµ  Ñ Ö ½ ½½ ¾¾ ½½ ½½ ¾ ½ ¼ ·Ñ¾¾ ѽ½ ÖÖ   Ѿ¾ ¾¾ ´Øµ   Ñ ¼ ¼ ¼ ½  ÖÖ ´Øµ ¿ ½½ ѽ½ ÖÖ ÖÖ ´Øµ ¼ ¼ ´Øµ¿ ¾ Ù ¿ Ú Þ½ Þ¾ (A. 6.13)) Lyapunov-function candidate ΠѾ ¾¾ ܾ · Ѿ ´ ¾¾   ѽ½ µ´Ñ½½   ¾¾ µ ܾ · ´Ñ½½   ¾¾ µ ¾¾ ¾Ñ½½ ¾¾ ½ ¾Ñ½½ ¾¾´Ñ¾ · ¾¾ µ ¾ ¾´ · ѽ½ µ´   ½½ ´ ¾¾   ѽ½ µ ܾ · ¾´ · ѽ½ µ´   µ µ Ü¿ · ¾ .10) and ´µ ¼ ´µ ¼ ¼ ¼ we can conclude the inequalities (A. the closed-loop dynamics (A. As a result ѽ½ . is well-defined.11) and use (A.11).3. Ѿ¾ .15) Differentiating the positive definite (cf. ¾¾ ѽ½ Proof of Proposition 6.1.13a). The closed-loop system (6.14) If we define expressed as ¾ and ¾ as in (A.

Proofs Î   ¾¾   ´Ñ½½Ñ  ´Ñ¾µ´ ¾¾· ѽ½ µµÑ¾¾ ܾ   ´ ´Ñ½½µ´   ·¾¾ µµ ܾ   ¾ ¿   ½½ ½½ ¾¾   ´  ´ ¾¾   ѽ½ µ µ ܾ µ´ · ѽ½ which is negative semi-definite (cf. ´µ ´ ´Øµ µ is UCO. exponentially stable (GUES) if the pair where ´ ´µ µ ¼ ¾ ´Øµ Ѿ¾ ´ ·Ñ½½ ѽ½ ¾¾ Ѿ¾ ´ ·Ñ½½ ѽ½ ¾¾ ¼ ¼ ¼ Ö ´Ñ ½½ ѽ½ ´Ñ¾ ½½   ¾¾   ½½   ѽ½ ´   ¾¾ ÖÖ   Ѿ¾ ´Ñ  µ ÖÖ Ñ¾¾ µ ѽ½ Ñ ¾¾   Ѿ¾   Ѿ¾ ´   µ ÖÖ Ñ¾¾ ´½½  µ ÖÖ Ñ¾¾ ´ ·Ñ½½ µ´ ¾¾  Ñ½½ µ µ ÖÖ ÖÖ   ¼ ¾ · ¾¾ µ ¾¾ ´Ñ½½ µ Ѿ¾ ´ ·Ñ½½ µ´Ñ½½   ¾¾ µ ÖÖ   ÖÖ ¼   ¾ · ¾¾ µ ¾¾ ´Ñ½½ ¾¾ µ´ ¾¾ ѽ½ µÑ¾¾ · ¾¾ µ ¿   Õ ´ ´Ñ½½ µ´     ¾¾ µ · µ Õ ´   ´ ¾¾ µ´  Ñ½½ µ ·Ñ½½ µ If ÖÖ Ø is persistently exciting.15) yields Appendix A.13)).148 along solutions of (A. .15) is globally uniformly Ø is uniformly completely observable (UCO). It is well-known (Khalil 1996) that the origin of the system (A.4 that the pair which completes the proof. (A. it follows from Corollary 2.3.

Appendix B Backstepping controller for an under-actuated ship The expression for Ù¾ ½ ¾ ½ ¾ ¿ « ­ ´Ù Ú Ö Þ½ Þ¾ Þ¿ ص of (6.17) is: ¾ ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾   ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ѽ½ Ù Ú ¾ Ö Ù Þ¾   ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾ · ¾Ñ½½Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾ Þ¾ ¿   ¾Ñ½½ Ù Ú ¾ Þ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾   ѽ½ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½ ¾ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¿ Þ¾ Þ½   ѽ½ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½   ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿   ¾Ñ½½Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ѽ½ Ù¾Ú ½ ¾Ö¾Þ½   ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾ · ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ¾ ¿ ¾   ¾Ñ½½ Ù Ú ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾   ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾ ¿ · ѽ½ Ù Ú ¾ Þ¾ Þ¾   ¾Ñ½½Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾   ¾Ñ½½Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾ · ѽ½ Ù¾Ú ¾ Ö Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾   ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¿ · ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¾ ¾ ¾   ѽ½ Ù Ú ¾ Ö Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾ ¾ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ ¾ · ѽ½ Ù¾Ú ½ ¾Ö¾Þ½ · ѽ½ Ù Ú ½ ¾Ö¾Þ½ Þ½ · ѽ½ Ù Ú¾« ¾ ¾Ö¾Þ¾ · ѽ½ Ù¾Ú « ¾ ¿Ö¾Þ¿ Þ¾ ¾ · ¾Ñ½½ Ù Ú « ¾ ¿Ö¾Ù Þ¿ Þ¾ · ѽ½ Ù Ú « ¾ ¾Ö¾Þ¾ Þ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú « ¾ ¾Ö¾Þ¿ Þ½ Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ½ ¾ Ö¾Þ½   ѽ½ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½ · ѽ½ Ù Ú ¾Ö¾ · ѽ½ Ù¾Ú ¿Þ¿ «Ö¾ ¾Þ¾ · ¾Ñ½½ Ù Ú ¿Ù Þ¿ «Ö¾ ¾Þ¾ · ѽ½ Ù Ú Ú ¾«Ö¾ ¾Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾«Ö¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿ ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾«Ö¾Þ¿ Þ½   Ѿ¾ Ù¾Ú ¿Þ¿ «Ö¾ ¾Þ¾ ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú Ú «Ö ¾Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿   Ѿ¾ Ù Ú ¾ Þ¾ ¾«Ö¾Þ¿ Þ½ ¾ · Ѿ¾ Ù Ú ¾Þ¾ Ö Ù   Ѿ¾ Ù Ú ¾ Þ¾ Ö Ù ¾ ¿   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ¾Ñ¾¾Ù Ú ¾ Þ¿ Þ½ Þ¾ ¿ ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ Ö Ù Þ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ ¾ ¾ ¾   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö ¾ ¾ Ù ¾ Þ½ Þ¿ · Ñ¿¿ ¿ ¾ Ù Þ¾ · Ñ¿¿ ¿ ¾ Þ¿ ¾ Ö¾ ¿ ¾ Ñ¿¿ ¿ ­ ¾ Ö¾ Þ¿ Þ¾ Ú · Ñ¿¿ ¿ « Ö¾ Ù Þ¿ ­Þ¾ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿ ­ Ö Þ¿ Þ½ Ù   Ñ¿¿ ¿ ­ Ö Þ½ Ú ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½   Ñ¿¿ ¿ ­ ¾Ö¾ Þ½ Ú ¾ ¾ ¿ ¾ · Ñ¿¿ ¿ ­ Ö Þ¿ Þ¾ Ú · Ñ¿¿ ¿ « Ö Ú ­Þ¾ Þ¿ ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ­ Ö Þ¾ · Ñ¿¿ ¿« ¿Ö¾ ¾Þ¾ ­Þ¿ ¾ ¾   Ñ¿¿ ¿ « Ö Ù Þ¿ ­Þ½ · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ½ Þ¿ ­Þ¾ ¿ ¾ ¿ ¾ · Ñ¿¿ ¿ « Ö ¾ Þ¾ ­Þ¿ Þ¾ · Ñ¿¿ ¿ « Ö Ú ­Þ¿ Þ¾   Ñ¿¿ ¿ « Ö¾Ù Þ¿ ­Þ½   Ñ¿¿ ¿« ¿Ö¾Ú ­Þ½ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿ « Ö Ù Þ¿ ­Þ¾ · Ñ¿¿ ¿ « Ö Ù Þ¿ ­Þ¾ ¿ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿ « Ö ¾ Þ½ Þ¿ ­Þ¾   Ѿ¾ Ù Ú ­ Ö Þ¾ Þ¾ ¿ ¾ ¾ ¿ ¾ · Ñ¿¿ ¿ « Ö ¾ Þ¿ Þ½ ­Þ¾   Ñ¿¿ ¿ « Ö ¾ Þ¾ ­Þ½   Ñ¿¿ ¿ ­ ¿Ö¾Þ½ Þ¿ Ù · Ñ¿¿ ¿ Ö ­ ¾Ö¾Þ¾ Þ¾ ¾ · Ñ¿¿ ¿ Ö ­ ¾Ö¾ Þ½ Þ½   Ñ¿¿ ¿« ¿Ö¾ ¾ Þ½ Þ¿ ­ ¾ ¾ · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ ­Þ¾ · Ñ¿¿ ¿ « Ö¾Ù Þ¿ ­Þ¾   Ñ¿¿ ¿ ­ ¿Ö¾Þ¾ Ù · Ñ¿¿ ¿ ­ ¾Ö¾Þ½ Þ¿ Ù ¿ ¾ ¾ ¾ ¾   Ñ¿¿ ¿ ­ Ö Þ½ Þ¿ Ù · Ñ¿¿ ¿ Ö ­ Ö Þ¾ ¾ ¾ · Ñ¿¿ ¿ Ö ­ ¾Ö¾ Þ½ · ѽ½ Ù Ú ­ ¾ Ö¾Þ½ · Ñ¿¿ ¿ ­ ¾ Ö¾Þ¿ Þ½ Ù · Ñ¿¿ ¿­ ¾ Ö¾Þ¾ Ù · Ñ¿¿ ¿ ­ ¾ Ö¾Þ¾ Ù   Ñ¿¿ ¿ ­ ¿Ö¾Þ¾ Ù · Ñ¿¿ ¿ ­ ¾Ö¾Þ½ Þ¿ Ù · ѽ½ Ù Ú ­ ¾Ö¾ Þ¾ Þ¾ · ѽ½ Ù Ú ­ ¾Ö¾Þ½ Þ½ · ѽ½ Ù Ú ­ ¾Ö¾ Þ¾ Þ¾ · ѽ½ Ù Ú ­ ¾Ö¾Þ½ Þ½ ¾ ¾ · ѽ½ Ù Ú ­ ¾Ö¾ Þ¾ · ѽ½ Ù Ú ­ ¾ Ö¾Þ½ ¾ · ѽ½ Ù Ú ­ ¾Ö¾ Þ½ Þ½ · ѽ½ Ù Ú ­ ¾Ö¾Þ¾ ¾ ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½ · ѽ½ Ù Ú ­ ¾Ö¾ Þ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾ Þ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½ Þ½ ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½ Þ½   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾ ¾ · ѽ½ Ù Ú ­ ¾ Ö¾Þ½ · ѽ½ Ù Ú ­ ¾ Ö¾Þ¾ Þ¾ ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾ Þ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½ Þ½   Ѿ¾ Ù Ú ­ ¾Ö¾Þ½ Þ½   Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ ­Þ½ · ѽ½ Ù Ú ­ ¾ Ö¾Þ½ Þ½ ¾ · ѽ½ Ù Ú ­ ¾Ö¾Þ½   Ѿ¾ Ù Ú ­ ¾Ö¾Þ¾ Þ¾   Ñ¿¿ ¿ ­ ¿Ö¾Þ¿ Þ½ Ù   Ñ¿¿ ¿­ ¿ Ö¾Þ¿ Þ½ Ù · ѽ½ Ù Ú ­ ¾Ö¾Þ¾ Þ¾ · Ñ¿¿ ¿ ­ ¾Ö¾Þ¾ Þ¿ Ú ¾ · Ñ¿¿ ¿ ­ ¾ Ö¾Þ¾ Þ¿ Ú · ѽ½ Ù Ú ­ ¾Ö¾Þ¾ ¾ ¾   ѽ½ Ù¾Ú ¾Þ¾ Ö   ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ ¾ ¾ ¾ ¾ ¾   ѽ½ Ù Ú ½ Ö Þ¿ Þ¾ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ¿ Þ½ ¾   ѽ½ Ù Ú ¾Þ¾ Ö Ù · ѽ½ Ù Ú ¾ Þ¾ Ö Ù ¾ ¿ · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¿ ¾   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ¾ Ö Ù Þ¾ Ö Ö Ö  ´ Ñ¿¿ · Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö 149 .

150 Ö Ö Ö Appendix B. Backstepping controller for an under-actuated ship   Ѿ¾ Ù Ú ½ ¾Ö¾Ù Þ½   Ѿ¾ Ù Ú ½ ¾ Ö¾Ù Þ½ ¾   Ѿ¾ Ù Ú ½ ¾ Ö¾Þ½ Þ½   Ѿ¾ Ù Ú¾« ¾ ¾ Ö¾Þ¾ ¾   Ѿ¾ Ù¾Ú « ¾ ¿Ö¾ Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¾ Þ¾ ¾ ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ ¾ Ö¾Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾ Þ¾ Þ¿ Þ½ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ ¾ Þ¾ Þ¾ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ ¾ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Ú Þ¾   Ñ¿¿ ¿ « ¿Ö¾Ú ½Ù   Ñ¿¿ ¿ « ¿Ö¿Ú ½Þ¾ ¾ · Ñ¿¿ ¿« Ö¾Ù Þ¿ ½ ½Þ¾ · Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ¾ ¾   Ñ¿¿ ¿ « Ö¾Ù Þ¿ ½ ½Þ½ · Ñ¿¿ ¿« Ö Ù¾Þ¿ ¾ ¾ ¾ · Ñ¿¿ ¿« ¿Ö¾Ù Þ¿ ¾ Þ¾ · Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾ ¾ ¾ ¾ ¾   Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾   Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾ ¾   Ñ¿¿ ¿ « Ö¿Ù Þ¿ ½Þ½ · Ñ¿¿ ¿ ¾ Ù Þ¾ ¾ · Ñ¿¿ ¿« Ö¾Ù Þ¿ ½ ½Þ¾ · ¾Ñ¿¿ ¿«¾ Ö¿Ù Þ¿ ¾Þ¾ ¾   Ñ¿¿ ¿ « ¾Ö Ú ½ ¾ Þ½ Þ¿ · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ Ù   Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ ½ Ù   Ñ¿¿ ¿« ¿ Ö¿Ú ½Þ½ Þ¿ · ¾Ñ¿¿ ¿ « ¿Ö Ú ¾ Ù Þ¿ · Ñ¿¿ ¿« ¾Ö¾Ú ¾ Þ¾ Þ¿ ¾   ¾Ñ¿¿ ¿ « ¿Ö Ú ¾ Ù Þ¿   Ѿ¾ Ù¾Ú ¿ Þ¿ «Ö¾ ¾Þ¾ ¾ ¾   Ѿ¾ Ù Ú¾ ¾«Ö¾ ¾Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿   Ѿ¾ Ù Ú ¾ Þ¾ ¾«Ö¾Þ¿ Þ½ ¾ · Ѿ¾ Ù Ú ¾Þ¾ Ö Ù   Ѿ¾ Ù Ú ¾ Þ¾ Ö Ù ¾ ¿   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ¾Ñ¾¾Ù Ú ¾ Þ¿ Þ½ Þ¾ ¿ ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ Ö Ù Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ · ¾Ñ¾¾Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ Ö Ù Þ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾   ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ¾ ¿ ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾   Ѿ¾ Ù Ú ¾Ö¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½ ¾ ¾   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾   Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½   Ѿ¾ Ù¾Ú ½ ¾Ö¾Þ½ ¾   Ѿ¾ Ù¾Ú ½ ¾ Ö¾Þ½   Ѿ¾ Ù Ú ½ ¾Ö¾Þ½ Þ½   Ѿ¾ Ù Ú « ¾ ¾Ö¾Ú Þ¾   Ѿ¾ Ù¾Ú « ¾ ¿Ö¾ Þ¿ Þ¾ ¾   ¾Ñ¾¾ Ù Ú « ¾ ¿ Ö¾Ù Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾ Þ¾ Þ¾ ¾ ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ ¾ Ö¾Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾ Þ¾ Þ¿ Þ½ ¿   Ѿ¾ Ù Ú ¾Ö¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾   Ѿ¾ Ù¾Ú ¾ Ö Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾ · ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù¾Ú ¾ Ö Þ¾ ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾Þ¾ Þ¾ ¾ ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ ¾ Ö¾Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½   Ѿ¾ Ù¾Ú ¿ Þ¿ «Ö¾ ¾Þ¾   ¾Ñ¾¾Ù Ú ¿Ù Þ¿ «Ö¾ ¾Þ¾ ¾ ¾   Ѿ¾ Ù Ú Ú ¾«Ö¾ ¾Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿   Ѿ¾ Ù Ú ¾ Þ¾ ¾«Ö¾Þ¿ Þ½ ¾ · Ѿ¾ Ù¾Ú ¾Þ¾ Ö   Ѿ¾ Ù¾Ú ¾ Þ¾ Ö ¾ ¿   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ¾Ñ¾¾Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ ¾ ¾ ¾   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö   Ѿ¾ Ù¾Ú ½ ¾ Ö¾Þ½   Ѿ¾ Ù¾Ú ½ ¾Ö¾Þ½ ¾   Ѿ¾ Ù Ú ½ ¾Ö¾ Þ½ Þ½   Ѿ¾ Ù Ú¾« ¾ ¾Ö¾ Þ¾ ¾   Ѿ¾ Ù¾Ú « ¾ ¿Ö¾Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿   ¾Ñ¾¾ Ù Ú « ¾ ¿Ö¾Ù Þ¿ Þ¾ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾   ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¿ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ ¾ Þ¾ Þ¾ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ · Ѿ¾ Ù¾Ú ¾ Þ¾ Ö   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¿ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾   Ѿ¾ Ù¾Ú ¾ Ö Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ · ѽ½ Ù Ú ¾Ö¾   Ѿ¾ Ù¾Ú ¿ Þ¿ «Ö¾ ¾Þ¾   ¾Ñ¾¾ Ù Ú ¿ Ù Þ¿ «Ö¾ ¾Þ¾   Ѿ¾ Ù Ú¾ ¾«Ö¾ ¾Þ¾ ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿ ¾   Ѿ¾ Ù¾Ú ¾ Þ¾ Ö · ѽ½ Ù¾Ú « ¾ ¿Ö¾Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ¾ Þ¾ · ѽ½ Ù Ú « ¾ ¾Ö¾ Þ½ Þ¿ Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ½ ¾ Ö¾Þ½ ¾ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½ · ѽ½ Ù Ú ¾ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¾ ¿ · ѽ½ Ù Ú ¾ ¾Þ¾ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½   ѽ½ Ù Ú ½ ¾Ö¾ Þ¿ Þ¾ Þ½ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¿ Þ¾ Þ½   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½   ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ ¾   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½   ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ ¾ ¾ · ¾Ñ½½Ù Ú ½ Ö ¾ Þ¾ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ · ѽ½ Ù Ú ½ ¾Ö¾ Ù Þ½ ¾ · ѽ½ Ù Ú ½ ¾ Ö¾Ù Þ½ · ѽ½ Ù Ú ½ ¾Ö¾Þ½ Þ½ ¾ · ѽ½ Ù Ú « ¾ ¾Ö¾ Ú Þ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ¿ Þ½ · ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù¾Ú ¾ Ö Þ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾   ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ¾ ¿ ¾ · ¾Ñ¾¾Ù Ú ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ¾ Þ½ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ ¾ Þ¾ Þ¾ ¿ · ¾Ñ¾¾Ù Ú ¾ Þ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · ¾Ñ¾¾Ù Ú ½ Ö ¾ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾   Ѿ¾ Ù Ú ¾Ö¾ · ѽ½ Ù Ú « ¾ ¾Ö¾ Þ½ Þ¿ Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ½ ¾Ö¾Þ½ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½ · ѽ½ Ù Ú ¾Ö¾ · ѽ½ Ù¾Ú ¿Þ¿ «Ö¾ ¾Þ¾ · ѽ½ Ù Ú Ú ¾«Ö¾ ¾Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾«Ö¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿ ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾«Ö¾Þ¿ Þ½   ѽ½ Ù Ú ¾Þ¾ Ö Ù ¾   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½   ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ ¾ ¾ · ¾Ñ½½Ù Ú ½ Ö ¾ Þ¾ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ · ѽ½ Ù Ú ½ ¾Ö¾Ù Þ½ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö .

151 ¾ · ѽ½ Ù Ú ½ ¾ Ö¾Ù Þ½ · ѽ½ Ù Ú ½ ¾Ö¾Þ½ Þ½ · ѽ½ Ù Ú¾« ¾ ¾Ö¾ Þ¾ · ѽ½ Ù¾Ú « ¾ ¿Ö¾Þ¿ Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ Þ¾   ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ ¾ ¾ ¾   ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¾ ¿ · ѽ½ Ù Ú ¾ ¾Þ¾ Þ¾   ¾Ñ½½Ù Ú ¾ Þ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½   ѽ½ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¿ Þ¾ Þ½   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¿ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½   ¾Ñ½½Ù Ú ¾ Þ¾ ¾ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ѽ½ Ù Ú « ¾ ¾Ö¾Þ¾ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ Ö Ù · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Ö Ù Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾   ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ѽ½ Ù Ú ¾ Ö Ù Þ¾   ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ · ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¿ ¾ · ѽ½ Ù Ú ¾ ¾Þ¾   ¾Ñ½½Ù Ú ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾   ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ½ ¾ Ö¾Þ½ ¾   ѽ½ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½ · ѽ½ Ù Ú ¾Ö¾ · ѽ½ Ù¾Ú ¿ Þ¿ «Ö¾ ¾Þ¾ · ѽ½ Ù Ú¾ ¾«Ö¾ ¾Þ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾«Ö¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿ ¾ ¾ ¾ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ · ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ · ѽ½ Ù¾Ú ½ ¾Ö¾Þ½ · ѽ½ Ù¾Ú ½ ¾Ö¾Þ½ ¾ · ѽ½ Ù Ú ½ ¾Ö¾Þ½ Þ½ · ѽ½ Ù Ú « ¾ ¾Ö¾Ú Þ¾ · ѽ½ Ù¾Ú « ¾ ¿ Ö¾Þ¿ Þ¾ · ¾Ñ½½ Ù Ú « ¾ ¿Ö¾Ù Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú « ¾ ¾ Ö¾Þ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½   ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ½ Þ¿   ¾Ñ½½ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾   ѽ½ Ù Ú ½ Ö ¾ Þ¿ Þ½ · ¾Ñ½½ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ · ѽ½ Ù Ú ½ Ö ¾ Þ¾ Þ½ · ѽ½ Ù Ú ¾ Þ¾ ¾ ¾ ¾   ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾ Þ½ Þ¿ Þ¾ ¾ ¾ · ѽ½ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¾ ¿ · ѽ½ Ù Ú ¾ ¾Þ¾ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ½ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾   ѽ½ Ù Ú ½ ¾Ö¾Þ¾ Þ¿ Þ½   ѽ½ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ ¾ ¾ · ѽ½ Ù¾Ú ¾ Þ¾ Ö · ѽ½ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ½ Þ¿ Þ¾ ¿ ¿   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾ ¾ · ѽ½ Ù¾Ú ¾ Ö Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ ¾ ¾   ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ · ѽ½ Ù Ú ¾ ½ Ö Þ¾ Þ¿   ѽ½ Ù¾Ú ¾ Ö Þ¾   ѽ½ Ù Ú ¾ ½ Ö Þ½ Þ¾ · ¾Ñ½½ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¿ ¾ · ѽ½ Ù Ú ¾ ¾Þ¾   ¾Ñ½½ Ù Ú ¾ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ Þ¾ ¾   Ñ¿¿ ¿ Ö ¾ Þ¾ Ö Ù · Ñ¿¿ ¿Ö ¾ Þ¾ Ö Ù ¾ ¿ · Ñ¿¿ ¿ Ö ¾ ¾Þ¿ Þ½ Þ¾   ¾Ñ¿¿ ¿ Ö ¾ Þ½ Þ¿ Þ¾ ¿ ¿   ¾Ñ¿¿ ¿Ö ¾ Þ½ Þ¿ Þ¾   ¾Ñ¿¿ ¿ Ö ¾ Þ¿ Þ½ Þ¾ ¿ ¾   ¾Ñ¿¿ ¿Ö ¾ Þ¿ Þ½ Þ¾ · Ñ¿¿ ¿ Ö ¾ Ö Ù Þ¾ ¾ ¾ · Ñ¿¿ ¿ Ö ¾ ½ Ö Þ½ Þ¾   ¾Ñ¿¿ ¿ Ö ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¾ ¿ ¾ · Ñ¿¿ ¿ Ö ¾ ¾Þ¾ Þ¾ · Ñ¿¿ ¿Ö ¾ ¾Þ¾   ¾Ñ¿¿ ¿Ö ¾ Þ¾ · Ñ¿¿ ¿ Ö ¾ Þ¾ Þ¾ · Ñ¿¿ ¿Ö ¾ Ö¾ · Ñ¿¿ ¿ Ö ¿Ù Þ¿ «Ö¾ ¾ Þ¾ · Ñ¿¿ ¿ Ö ¿ Ù Þ¿ «Ö¾ ¾Þ¾ ¾ ¾ · Ñ¿¿ ¿ Ö Ú ¾«Ö¾ ¾Þ¾ · Ñ¿¿ ¿ Ö ¾ Þ¾ ¾«Ö¾ ¾ · Ñ¿¿ ¿ Ö ¾ Þ¾ ¾ «Ö¾Þ½ Þ¿ · Ñ¿¿ ¿ Ö ¾ Þ½ Þ¿ Þ¾ · Ñ¿¿ ¿ Ö ¾ Þ¿ Þ½ Þ¾ · Ñ¿¿ ¿Ö ¾ Þ¿ Þ½ Þ¾ ¾ ¾   Ñ¿¿ ¿ Ö ½ ¾Ö¾Þ¾ Þ¿ Þ½   Ñ¿¿ ¿Ö ½ ¾Ö¾Þ¿ Þ¾ Þ½ ¾   Ñ¿¿ ¿ Ö ½ ¾Ö¾Þ¿ Þ¾ Þ½   Ñ¿¿ ¿Ö ½ Ö ¾ Þ¾ Þ½ ¾ ¾ · ¾Ñ¿¿ ¿ Ö ½ Ö ¾ Þ¾ Þ½ · Ñ¿¿ ¿ Ö ¾ ½ Ö Þ¿ Þ¾ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö ¾ · Ñ¿¿ ¿Ö ¾ ½ Ö Þ¾ Þ¿   Ñ¿¿ ¿ Ö ¾ Ö Ù Þ¾   Ñ¿¿ ¿ Ö ¾ ½ Ö Þ½ Þ¾ · ¾Ñ¿¿ ¿ Ö ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¾ ¾   Ñ¿¿ ¿ Ö ¾ ½ Ö Þ¾ Þ¿   Ñ¿¿ ¿Ö ¾ ½ Ö Þ¿ Þ¾ ¾ ¾ · Ñ¿¿ ¿Ö ¾ ¾Þ½ Þ¿ Þ¾ · Ñ¿¿ ¿Ö ¾ ¾Þ¿ Þ½ Þ¾ ¿ ¾   ¾Ñ¿¿ ¿ Ö ¾ Þ¾ Þ¾   Ñ¿¿ ¿ Ö ½ ¾Ö¾ Þ¾ Þ¿ Þ½ ¾ · Ñ¿¿ ¿Ö ¾ Þ½ Þ¿ Þ¾   Ñ¿¿ ¿ Ö ½ Ö ¾ Þ½ Þ¿ ¾ ¾ · Ñ¿¿ ¿Ö ¾ Þ¾ ¾«Ö¾Þ¿ Þ½ · Ñ¿¿ ¿« ¿Ö¾ ¾ Þ¿ Þ½ ½ ½Þ¾ ¾   Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ ½ ½ Þ½ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¿ Þ½ Ù ¾ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¿ Þ½ Ù   Ñ¿¿ ¿« ¿ Ö¾ ¾Þ¿ Þ½ ½ Ù ¾ · Ñ¿¿ ¿« ¿ Ö¾ ¾Þ¿ Þ½ Ù   Ñ¿¿ ¿ « ¿Ö¾ ¾ Þ¿ Þ½ Ù ¾   Ñ¿¿ ¿ « ¿Ö¾ ¾ Þ¿ Þ½ Ù · Ñ¿¿ ¿« ¾ Ö ¾ Ú¾ ¿ ¾ ¿   Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ ½   Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ Ú ¾ ¾   Ñ¿¿ ¿ « ¿Ö¾ ¾Þ½ Þ¿ ½Ù · Ñ¿¿ ¿ « ¿Ö ¾ Þ¿ Þ½ Ù ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¿ Þ½ Þ¾ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¿ Þ½ Þ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¿ Ö ¾ Þ¿ Þ½ Ù · Ñ¿¿ ¿« ¾Ö ¾ Þ¿ Þ½ Ú ¾ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¾ Ö ¾ Þ½ Þ¿ ½ · ¾Ñ¿¿ ¿« ¾Ö ¾ Þ½ Þ¿ ½ Þ½ ¾ · Ñ¿¿ ¿« ¿ Ö¾ ¾Þ½ Þ¿ ½ ½Þ¾ ¾ ¾ · Ñ¿¿ ¿« ¿ Ö¾ ¾Þ½ Þ¿ ½ ½Þ¾   Ñ¿¿ ¿« ¿ Ö¾ ¾Þ½ Þ¿ ½ ½ ¿ ¾ ¾   ¾Ñ¿¿ ¿ « ¾Ö ¾ Þ½ Þ¿ ½Þ½ · Ñ¿¿ ¿« ¾Ö¾ ¾ Þ½ Þ¿ Ù ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ½ Þ¿ Ù · Ñ¿¿ ¿ «¾ Ö¿ ¾ Þ¿ Þ½ ¾ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¾ Ö ¾ Þ¿ Þ½ ½ · Ñ¿¿ ¿«¾ Ö¿ ¾ Þ¾ ¿ ¾ ¾ ¾   Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ½ Þ¿ Þ¾   Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ½ Þ¿ ¾ ¾ ¾ ¾   ¾Ñ¿¿ ¿ « ¿Ö¿ ¾Þ½ Þ¿ ½ Þ½ · Ñ¿¿ ¿ «¾ Ö¿ ¾ Þ½ Þ¿ · Ñ¿¿ ¿« ¾ Ö Ú ½ ¾ Þ½ Þ¿ · Ñ¿¿ ¿ « ¾Ö Ú ½ ¾ Þ¿ Þ½ · Ñ¿¿ ¿« ¿ Ö¾Ú ½ ½ Þ¾ Þ¿ · Ñ¿¿ ¿ « ¿Ö¾Ú ½ ½ Þ¿ Þ¾ ¾ · Ñ¿¿ ¿« ¾ Ö Ú ½ ¾ Þ¾   Ñ¿¿ ¿ « ¾Ö Ú ½ ¾ Þ¿ Þ½ ¿ ¾ ¾ ¾ ¾   Ñ¿¿ ¿ « ¾Ö ¾ Þ¿ Þ½ ½ · Ñ¿¿ ¿« ¿Ö ¾ Þ½ Þ¿ Ù ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ½ Þ¿ Þ¾ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ½ Þ¿ Þ¾ ¾ ¾ ¾ · Ñ¿¿ ¿« ¿ Ö ¾ Þ½ Þ¿ Ù · Ñ¿¿ ¿« ¾Ö ¾ Þ½ Þ¿ Ú ¿ ¾ ¿ ¾   Ñ¿¿ ¿ « ¿Ö ¾ Þ½ Þ¿ Ù   Ñ¿¿ ¿ « ¿Ö ¾ Þ½ Þ¿ Ù ¾ ¾ · Ñ¿¿ ¿«¾ Ö¿Ù¾Þ¿ · Ñ¿¿ ¿ « Ö¾Ù Þ¿ ½ ½ Þ¾ ¿   Ñ¿¿ ¿« ¿ Ö¾ ¾Þ½ Þ¿ ½ Ù   Ñ¿¿ ¿ « ¾Ö ¾ Þ½ Þ¿ Ú ¿ ¾ ¿ ¾   Ñ¿¿ ¿« ¿ Ö ¾ Þ¿ Þ½ Ù   Ñ¿¿ ¿ « ¿Ö ¾ Þ¿ Þ½ Ù ¿ ¾ ¿ ¾   Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¿ Þ½ Þ¾   Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¿ Þ½ Þ¾ ¾ ¿   Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¿ Þ½ Þ½ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¿ Þ½ Þ½ ¾ ¾ ¾   Ñ¿¿ ¿« ¿ Ö¿ ¾Þ¿ Þ½ ½ · Ñ¿¿ ¿ « ¿Ö ¾ Þ¾ Ù Þ¿ ¿ ¾   Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ½ Þ¿ Þ¾ · Ñ¿¿ ¿« ¾Ö¾ Ú ¾Ù ¾ ¾ ¾ · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¿ Þ½ ½ ½Þ¾   Ñ¿¿ ¿ « Ö Ù¾Þ¿ ¾ ¾ ¾ ¿ ¾   Ñ¿¿ ¿« Ö Ù¾Þ¿ ¾   Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Þ¿   ¾Ñ¿¿ ¿Ö ½ Ö ¾ Þ¾ Þ½   ¾Ñ¿¿ ¿Ö ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾   Ñ¿¿ ¿Ö ½ Ö ¾ Þ¿ Þ½ · Ñ¿¿ ¿Ö ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾ · Ñ¿¿ ¿ Ö ½ Ö ¾ Þ½ Þ¿ · ¾Ñ¿¿ ¿ Ö ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · Ñ¿¿ ¿ Ö ½ Ö ¾ Þ¿ Þ½ · Ñ¿¿ ¿Ö ½ ¾ Ö¾Ù Þ½ ¾ · Ñ¿¿ ¿ Ö ½ ¾Ö¾Ù Þ½ · Ñ¿¿ ¿ Ö ½ ¾ Ö¾Þ½ Þ½ · Ñ¿¿ ¿ Ö « ¾ ¾Ö¾Ú Þ¾ · Ñ¿¿ ¿Ö « ¾ ¿Ö¾Ù Þ¿ Þ¾ ¾ · Ñ¿¿ ¿ Ö « ¾ ¿Ö¾Ù Þ¿ Þ¾ · Ñ¿¿ ¿ Ö « ¾ ¾Ö¾Þ¾ Þ¾ ¾ ¾ · Ñ¿¿ ¿ Ö « ¾ ¾Ö¾Þ½ Þ¿ Þ¾ · Ñ¿¿ ¿ Ö « ¾ ¾Ö¾Þ¿ Þ½ Þ¾ ¾ ¾ ¾ ¾ · Ñ¿¿ ¿ Ö ½ ¾ Ö¾Þ½   Ñ¿¿ ¿« ¿Ö Ù Þ¿ ½ ¾ Þ½ ¾ ¾ · Ñ¿¿ ¿ « ¿Ö¾Ù¾Þ¿ ¾   Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½ ¾ ¾ · Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ Ú · ¾Ñ¿¿ ¿ «¾ Ö¿Ù Þ¿ ¾Þ½ ¾ ¾   Ñ¿¿ ¿« ¿ Ö¾ ¾ Þ¾ Ù · Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½   Ñ¿¿ ¿« Ö¾Ù Þ¿ ½ Ù · Ñ¿¿ ¿ ½ Ö ¾ Þ½ Þ¿ Ú ¾ ¿   Ñ¿¿ ¿ ¾ Ù ¾ Þ¿ Þ½ · ¾Ñ¿¿ ¿ Ù ¾ Þ¾ ¾ · ¾Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ ½ Þ½ Þ¿ ¾ · ¾Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ ½ Þ¿ Þ½ · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¾ ½ ½Þ¿ Þ¾   Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¾ ½ ½Þ½ ¿ ¿   ¾Ñ¿¿ ¿« ¾Ö ¾ Þ¾ ½ Þ¿ Þ½   ¾Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ ½Þ½ Þ¿   Ñ¿¿ ¿« ¿ Ö¾ ¾Þ¾ ½ Ù · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¾ Ù ¾   Ñ¿¿ ¿« ¿ Ö¾ ¾ Þ¾ Ù · Ñ¿¿ ¿ « ¿Ö¾ ¾Þ¾ Ù ¾   Ñ¿¿ ¿« ¿ Ö¾ ¾Þ¾ ½Ù   Ñ¿¿ ¿ « ¿Ö¿ ¾Þ¾ ½ ¾ ¾ ¿ ¾   Ñ¿¿ ¿« ¿ Ö¿ ¾Þ½ Þ¿ ½ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ½ Þ¿ ¾ ¾ · Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½ · Ñ¿¿ ¿ « Ö¾Ù Þ¿ ½ ½ Þ¾ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö .

152 Ö Ö Appendix B. Backstepping controller for an under-actuated ship Ö Ö · Ñ¿¿ ¿« ¿ Ö Ù Þ¿ ½ ¾ Þ¾   Ñ¿¿ ¿ « Ö¾Ù Þ¿ ½ ½ Þ½ ¾   Ñ¿¿ ¿ ½ ¾ Ö¾Ù Ú   Ñ¿¿ ¿ ½ ¾ Ö¾Þ½ Ú ¾ ¾   Ñ¿¿ ¿ ½ ¾Ö¾Þ½ Ú · Ñ¿¿ ¿ ½ ¾Ö¾Þ¾ Þ¿ Ú ¾ ¾ · Ñ¿¿ ¿ ½ ¾Ö¾Þ¾ Þ¿ Ú · Ñ¿¿ ¿ ½ ¾ Ö¾Þ¿ Þ¾ Ú ¾ · Ñ¿¿ ¿ ½ ¾Ö¾Þ¿ Þ¾ Ú · Ñ¿¿ ¿ ½ Ö ¾ Þ¾ Ú ¾ · Ñ¿¿ ¿ ½ Ö ¾ Þ¾ Ú   Ñ¿¿ ¿ ½ Ö ¾ Þ¾ Ú ¿ ¾ ¿   ¾Ñ¿¿ ¿ Ù ¾ Þ¾ · Ñ¿¿ ¿ ¾Ù ¾ Þ¾   ¾Ñ¿¿ ¿ Ù ¾ Þ¾ · Ñ¿¿ ¿ ¾ Ù Þ½ Þ¿ · Ñ¿¿ ¿ ¾ Ù Þ¿ Þ½ · Ñ¿¿ ¿ ¾ Ù Þ½ Þ¿ ¾ · Ñ¿¿ ¿ ¾ Ù Þ¿ Þ½   Ñ¿¿ ¿ ¾ ¾Ù¾Ö   Ñ¿¿ ¿ ¾ Ù Þ¾ ¾ ¿   Ñ¿¿ ¿ ¾ Ù Þ¾ · Ñ¿¿ ¿ ¾ Ù¾ Ö · ¾Ñ¿¿ ¿ Ù ¾ Þ½ Þ¿ ¿ ¾ · ¾Ñ¿¿ ¿ Ù ¾ Þ¿ Þ½ · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ½ ¾ ¾   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¾ Þ¿   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¿ Þ¾ ¾ ¾ · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ½   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¾ Þ¿ ¾ ¾   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¿ Þ¾   Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ½ ¾ ¾ · Ñ¿¿ ¿ ¾ ¾Ù ½Ö Þ¾ Þ¿ · Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ¿ Þ¾ ¾   Ñ¿¿ ¿ ¾ Ù Þ½ Þ¿   Ñ¿¿ ¿ ¾ Ù Þ¿ Þ½   Ñ¿¿ ¿ ¾ ¾Ù Ù Ö ¾ ¾   Ñ¿¿ ¿ ¾ ¾Ù ½ Ö Þ½ · Ñ¿¿ ¿ ¾ ¾Ù ½Ö Þ¾ Þ¿ ¾ · Ñ¿¿ ¿ ¾ Ù Ù Ö · Ñ¿¿ ¿ ½ Ö ¾ Þ½ Þ¿ Ú · Ñ¿¿ ¿ ½ Ö ¾ Þ¿ Þ½ Ú · Ñ¿¿ ¿ ½ Ö ¾ Þ¿ Þ½ Ú ¾ ¾   Ñ¿¿ ¿ ½ Ö ¾ Þ¾ Ú   Ñ¿¿ ¿ ½ Ö ¾ Þ½ Þ¿ Ú ¾ ¾   Ñ¿¿ ¿ ½ Ö ¾ Þ½ Þ¿ Ú   Ñ¿¿ ¿ ½ Ö ¾ Þ¿ Þ½ Ú ¾ ¾   Ñ¿¿ ¿ ½ Ö ¾ Þ¿ Þ½ Ú · Ñ¿¿ ¿ ¾Ù ¾ Þ½ Þ¿ ¾ ¿ · Ñ¿¿ ¿ ¾ Ù ¾ Þ¿ Þ½   ¾Ñ¿¿ ¿ Ù ¾ Þ½ Þ¿ ¿ ¾   ¾Ñ¿¿ ¿ Ù ¾ Þ¿ Þ½   Ñ¿¿ ¿ ¾Ù ¾ Þ¾ ¾ ¾   Ñ¿¿ ¿ ¾Ù ¾ Þ½ Þ¿   Ñ¿¿ ¿ ¾Ù ¾ Þ¿ Þ½ ¿ ¿ · ¾Ñ¿¿ ¿ Ù ¾ Þ¾ · ¾Ñ¿¿ ¿ Ù ¾ Þ½ Þ¿ ¿ · ¾Ñ¿¿ ¿ Ù ¾ Þ¿ Þ½   Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ¿ Þ¾   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ½ · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¾ Þ¿ · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¿ Þ¾   Ñ¿¿ ¿ ¾ Ù Ù Ö   Ñ¿¿ ¿ ¾ Ù ½ Ö Þ½   Ñ¿¿ ¿ ¾ Ù¾ Ö · Ñ¿¿ ¿ ¾ ¾ Ù¾Ö · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¾ Þ¿ · Ñ¿¿ ¿ ¾ Ù ½ Ö Þ¿ Þ¾ · Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ½   Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ¾ Þ¿   Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ¿ Þ¾ ¾ ¾ · Ñ¿¿ ¿ ¾ ¾Ù ½ Ö Þ¿ Þ¾ · Ñ¿¿ ¿ ¾Ù ¾ Þ¾ ¾ ¾ · ¾Ñ¿¿ ¿ «¾ Ö¿Ù Þ¿ ¾Þ½ · ¾Ñ¿¿ ¿ «¾ Ö¿Ù Þ¿ ¾Þ½ ¾ ¾ · ¾Ñ¿¿ ¿ «¾ Ö¿Ù Þ¿ Ù · Ñ¿¿ ¿« ¿Ö Ù Þ¿ ½ ¾ Þ½ ¾ · Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½ · Ñ¿¿ ¿ «¾ Ö¿Ú¾ · Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ½   Ñ¿¿ ¿ ¾ ¾ Ù ½Ö Þ¾ Þ¿   Ñ¿¿ ¿ ¾ Ù Þ½ Þ¿   Ñ¿¿ ¿ ¾ Ù Þ¿ Þ½ ¾ ¾ · Ñ¿¿ ¿ ¾Ù ¾ Þ½ Þ¿ · Ñ¿¿ ¿ ¾Ù ¾ Þ¿ Þ½ ¿   ¾Ñ¿¿ ¿ Ù ¾ Þ½ Þ¿ · ¾Ñ¿¿ ¿ «¾ Ö¿Ù Þ¿ ¾Þ¾ ¿   ¾Ñ¿¿ ¿ Ù ¾ Þ¿ Þ½ · Ñ¿¿ ¿ ¾ ¾ Ù Ù Ö ¿   Ñ¿¿ ¿ ½ ¾Ö¾Ù Ú   Ñ¿¿ ¿« ¿Ö ¾ Þ¾ Ù Þ¿ ¾ ¾ ¾   Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ¾ · ¾Ñ¿¿ ¿ «¾ Ö¿ ¾ Þ½ Þ¿ Þ½ ¿ ¾   Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Þ¿ Þ¾   Ñ¿¿ ¿« ¾Ö¾ ¾ Þ¾ Þ½ ¾   ¾Ñ¿¿ ¿« ¿ Ö¿ ¾Þ¾ ½Þ¿ Þ½ · ¾Ñ¿¿ ¿ «¾ Ö¿ ¾ Þ¾ Þ½ Þ¿ ¾ ¾ · ¾Ñ¿¿ ¿ «¾ Ö¿ ¾ Þ¾ Þ¿ Þ½ · Ñ¿¿ ¿ « ¿Ö¾ ¾ Þ¾ ½ ½Þ¿   Ñ¿¿ ¿ « Ö¾Ù¾Þ¿ ¾ · Ñ¿¿ ¿ « Ö¾Ù¾Þ¿   Ñ¿¿ ¿ « Ö¾Ù Þ¿ ½ Ù   Ñ¿¿ ¿« Ö¿Ù Þ¿ ½Þ¾ ¾   ¾Ñ¿¿ ¿« ¿ Ö¿ ¾Þ¾ ½Þ½ Þ¿ · Ñ¿¿ ¿ « ¿Ö ¾ Þ¾ Ù Þ¿ ¾ ¾ ¾ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Þ¿ · Ñ¿¿ ¿« ¾ Ö¾ ¾ Þ¾ Þ¿ Þ¾ ¿   Ñ¿¿ ¿ « ¿Ö ¾ Þ¾ Ù Þ¿ · ¾Ñ¿¿ ¿«¾ Ö¿Ú ¾Þ½ Þ¿ · ¾Ñ¿¿ ¿ «¾ Ö¿Ú ¾ Þ¿ Þ½ · ¾Ñ¿¿ ¿ «¾ Ö¿Ú Ù Þ¿ ¿   Ñ¿¿ ¿ « ¿Ö¾Ú ½ ½Þ½ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Þ½ ¾ ¾ ¾ · Ñ¿¿ ¿ « ¾Ö ¾ Þ¾ ½ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Ù ¾   Ѿ¾ Ù¾Ú « ¾ ¿Ö¾Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¾ Þ¾ ¾ ¾   Ѿ¾ Ù Ú « ¾ ¾Ö¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú « ¾ ¾ Ö¾Þ¿ Þ½ Þ¾ ¾ · Ѿ¾ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½ · Ñ¿¿ ¿ « ¾Ö¾Ú ¾ Þ¿ Þ¾ ¾ · ¾Ñ¿¿ ¿ «¾ Ö¿Ú ¾ Þ¾ · Ñ¿¿ ¿ « ¾Ö¾ ¾ Þ¾ Ù ¿ ¾ ¾ ¿   Ñ¿¿ ¿ « ¾Ö ¾ Þ½ Þ¿ ½   Ñ¿¿ ¿« ¾ Ö ¾ Þ¿ Þ½ Ú ¾ ¾ · Ñ¿¿ ¿ « ¾Ö¾Ú ¾ Þ½   Ñ¿¿ ¿« ¾Ö Ú ½ ¾ Þ¾ ¾ · Ñ¿¿ ¿ « ¾Ö¾Ú ¾Ù   Ñ¿¿ ¿« ¿Ö Ù Þ¿ ½ ¾ Þ¾ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö ¾ ¾   Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½ · Ñ¿¿ ¿ « ¿Ö¾Ù¾Þ¿ ¾ ¾ ¾ · ¾Ñ¿¿ ¿« ¿Ö¾Ù Þ¿ ¾Ù   Ñ¿¿ ¿ « ¿Ö Ù Þ¿ ½ ¾ Þ½   Ñ¿¿ ¿ « Ö¾Ù¾Þ¿ ½ · Ñ¿¿ ¿ « Ö¾Ù¾Þ¿   Ñ¿¿ ¿ « Ö¾Ù¾Þ¿ ¾   ¾Ñ¿¿ ¿ « Ö¾Ù Þ¿ ¾ Ù · ¾Ñ¿¿ ¿« Ö¾Ù Þ¿ Ù   Ñ¿¿ ¿ « Ö¾Ù¾Þ¿ ½ ¾   Ñ¿¿ ¿ « Ö¿Ù Þ¿ ½Þ¾   Ñ¿¿ ¿ « Ö¿Ù Þ¿ ½Þ½ ¾ ¾   Ñ¿¿ ¿ « Ö¿Ù Þ¿ ½Þ½   Ñ¿¿ ¿ « Ö¿Ù Þ¿ ½Þ½ ¾ ¾ · ¾Ñ¿¿ ¿« Ö Ù Þ¿ ¾ Ù · Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾ ¾ ¾ · Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾ · Ñ¿¿ ¿ « Ö Ù¾Þ¿ ¾ ¾ ¾ ¾ ¾   ¾Ñ¿¿ ¿ « Ö Ù Þ¿ ¾ Ù   Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾ ¾ ¾   Ñ¿¿ ¿ « ¿Ö¾Ù Þ¿ ¾ Þ¾   Ñ¿¿ ¿ « ¿Ö¾Ú ½ Ù · Ñ¿¿ ¿ « ¿Ö¾Ú Ù   Ñ¿¿ ¿« ¿Ö¾ Ú ¾Ù   Ñ¿¿ ¿ « ¿Ö¾Ú ¾ Ù · Ñ¿¿ ¿ « ¿Ö¾Ú Ù   Ѿ¾ Ù Ú ¾Ö¾ ¾   Ñ¿¿ ¿ ¾Ù ¾ Þ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ · ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ Ö ¾ Þ¾ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ ¾ ¾ ¾   ¾Ñ¾¾ Ù Ú ½ Ö ¾ Þ½ Þ¿ Þ½   Ѿ¾ Ù Ú ½ Ö ¾ Þ¿ Þ½   Ѿ¾ Ù Ú ½ ¾ Ö¾Ù Þ½   Ѿ¾ Ù Ú ½ ¾Ö¾Ù Þ½ ¾ ¾ ¾   Ѿ¾ Ù Ú ½ ¾Ö¾Þ½ Þ½   Ѿ¾ Ù Ú ½ ¾ Ö¾Þ½ ¾ ¾ · ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾   Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ Ö Ù Þ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ½ Þ¾   ¾Ñ¾¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ Þ¾ ¾ ¾ ¿ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¾ · ¾Ñ¾¾ Ù Ú ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¾ Þ¿ · Ѿ¾ Ù Ú ¾ ½ Ö Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ½ Þ¿ Þ¾ ¾ ¾   Ѿ¾ Ù Ú ¾ ¾Þ¿ Þ½ Þ¾   Ѿ¾ Ù Ú ¾ ¾Þ¾ Þ¾ ¿ · ¾Ñ¾¾ Ù Ú ¾ Þ¾ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ½ Þ¿ Þ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¿ Þ½ Þ¾ · Ѿ¾ Ù Ú ½ ¾ Ö¾Þ¾ Þ¿ Þ½ ¾ ¾ · Ѿ¾ Ù Ú ½ ¾Ö¾Þ¿ Þ¾ Þ½ · Ѿ¾ Ù Ú ½ ¾ Ö¾Þ¿ Þ¾ Þ½ ¾ ¾   Ѿ¾ Ù Ú ¾ Þ¾ · Ñ¿¿ ¿Ö ¾ Þ¾ · ѽ½ Ù¾Ú ¿ Þ¿ «Ö¾ ¾ Þ¾ Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ñ½½ Ù Ú ¿ Ù Þ¿ «Ö¾ ¾ Þ¾ · ѽ½ Ù Ú ¾ ¾ «Ö¾ ¾ Þ¾ ¾ ¾ ¾ ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ «Ö¾ Þ½ Þ¿ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ «Ö¾ Þ¿ Þ½   ѽ½ Ù¾ Ú ¾ Þ¾ Ö ¾ ¾ ¾ ¾ ¾ · ѽ½ Ù Ú ¾ Þ¾ Ö · ѽ½ Ù Ú ¾ Þ¿ Þ½ Þ¾ · ѽ½ Ù Ú ¾ Þ¾ ¾ ¾ ¾ · Ñ¿¿ ¿ « Ö¿ Ù¾ Þ¿ · ¾Ñ¿¿ ¿ «¾ Ö¿ Ù Þ¿ ¾ Þ½ ¾ ¾   Ñ¿¿ ¿ « ¾Ö ¾ Ú¾ · Ñ¿¿ ¿ Ö ¾ ¾Þ½ Þ¿ Þ¾ ¿ ¿ ¾ · ¾Ñ¿¿ ¿ « Ö Ú ¾ Ù Þ¿   ¾Ñ¿¿ ¿ « Ö Ú ¾ Ù Þ¿ ¾ ¾   Ñ¿¿ ¿ « ¾Ö¾Ú ¾ Þ¾ Þ¿   Ñ¿¿ ¿« ¾ Ö¾Ú ¾ Þ¿ Þ¾   Ñ¿¿ ¿ « ¾Ö¾Ú ¾ Þ½   Ñ¿¿ ¿ « ¿Ö¿Ú ½Þ¿ Þ½ ¾ ¾ ¾ · ¾Ñ¿¿ ¿ « Ö¿ Ú Ù Þ¿ µ ´­ ¾ Ö¾ Þ½ · ­ ¾ Ö¾ Þ¾ · ­ ¾ Ö¾ Þ¾ Þ¾ ¿ ¾ ¿ ¾ ¾ ¾ · ­ Ö Þ½ Þ½ · Ù Þ¿ «Ö ¾ Þ¾ · Ù Þ¿ «Ö ¾ Þ¾ ¾ ¾ ¾ ¾ ·Ú «Ö¾ ¾ Þ¾ · ¾ Þ¾ ¾ «Ö¾ · ¾ Þ¾ ¾ «Ö¾ Þ½ Þ¿ ¾ ¾ ¾ · ¾ Þ¾ «Ö¾ Þ¿ Þ½   ¾ Þ¾ Ö Ù · ¾ Þ¾ Ö Ù ¾ ¾ ¿ ¿ · ¾ Þ¿ Þ½ Þ¾   ¾ ¾ Þ½ Þ¿ Þ¾   ¾ ¾ Þ½ Þ¿ Þ¾ ¿ ¿ ¾   ¾ ¾ Þ¿ Þ½ Þ¾   ¾ ¾ Þ¿ Þ½ Þ¾ · ¾ Ö Ù Þ¾ ¾ ¾ ¾ · ¾ ½ Ö Þ½ Þ¾   ¾ ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ¾ ½ Ö Þ¿ Þ¾ ¾ · ¾ ½ Ö Þ¾ Þ¿   ¾ Ö Ù Þ¾   ¾ ½ Ö Þ½ Þ¾ ¾ ¾ ¾ ¿ ¾ · ¾ ¾ ½ Ö Þ¾ Þ¿ Þ¾ · ¾ Þ¾   ¾ ¾ Þ¾ · ¾ Þ¾ Þ¾ ¾ ¾ ¾ ¾ ¾   ¾ ½ Ö Þ¾ Þ¿   ¾ ½ Ö Þ¿ Þ¾ · ¾ ¾ Þ½ Þ¿ Þ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¿ · ¾ Þ½ Þ¿ Þ¾ · ¾ Þ¿ Þ½ Þ¾ · ¾ Þ¾ Þ¾   ¾ ¾ Þ¾ Þ¾ · ¾ Þ½ Þ¿ Þ¾ · ¾ Þ½ Þ¿ Þ¾ · ¾ Þ¿ Þ½ Þ¾ · ¾ Þ¿ Þ½ Þ¾ ¾ ¾ ¾   ½ ¾Ö¾Þ¾ Þ¿ Þ½   ½ ¾Ö¾ Þ¿ Þ¾ Þ½   ½ ¾ Ö¾Þ¿ Þ¾ Þ½ ¾   ¾ ½ Ö ¾ Þ¾ Þ½   ½ Ö ¾ Þ¾ Þ½   ½ Ö ¾ Þ½ Þ¿ ¾ ¾   ¾ ½ Ö ¾ Þ½ Þ¿ Þ½   ½ Ö ¾ Þ¿ Þ½ · ¾ ½ Ö ¾ Þ¾ Þ½ ¾ ¾ ¾ ¾ · ½ Ö ¾ Þ¾ Þ½ · ½ Ö ¾ Þ½ Þ¿ · ¾ ½ Ö ¾ Þ½ Þ¿ Þ½ ¾ ¾ ¾ ¾ ¾ ¾ · ½ Ö ¾ Þ¿ Þ½ · ½ Ö Ù Þ½ · ½ Ö Ù Þ½ ¾ ¾ ¾ ¾ ¾ ¿ ¾ · ½ Ö Þ½ Þ½ · « ¾ Ö Ú Þ¾ · « ¾ Ö Ù Þ¿ Þ¾ ¿ ¾ ¾ ¾ ¾ ¾ ¾ ¾ · « ¾ Ö Ù Þ¿ Þ¾ · « ¾ Ö Þ¾ Þ¾ · « ¾ Ö Þ½ Þ¿ Þ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ · « ¾ Ö Þ¿ Þ½ Þ¾ · ½ Ö Þ½   ½ Ö Þ¾ Þ¿ Þ½ · ¾ Þ¾ ¾ ¾ · Öµ ·¾ · Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö .

Canudas de Wit. vol. Millman.. M. 4549–4553. pp. 1995.Bibliography A. and M. Khosla. Barb˘ lat.” Automatica. S. USA: Birkh¨ user. USA. “Control of linear time-varying systems by the gradient algorithm. C. 3475–3480. R. W. “Visual servoing in the task-function e framework: A contour following task. H. “Control of a nonholonomic mobile robot: backstepping kinematics into dynamics.” in Differential Geometric Control Theory. “Asymptotic stability and feedback stabilization. Nijmeijer.” Journal of Intelligent and Robotic Systems. Reyhanoglu. Fierro and F. I.” IEEE Transactions on Automatic Control. 1997. “Practical stabilization of nonlinear systems in chained form. Brockett. M. vol. P. pp. H. “Exponential stabilization of nonholonomic systems via discontinuous control.” in Proceedings of the 36th Conference on Decision and Control. McClamroch. W. e e A. Ortega. 1998.” in Proceedings of the 33rd Conference on Decision and Control. FL. 24. pp. LA. 1959. and H. 125–132. 3805–3810. 1–12. 1746–1761. 34. 1. pp. Berghuis. vol. and N.” in Proceedings of the 34th Conference on Decision and Control.” Systems and Control Letters. Sussmann. R. Coste-Mani` re. Eds. Brockett.-S. USA. no. Chen. Bloch. 153 . 12. USA. pp. Byrnes and A. USA. 1989. 37. no. 437–442. 181–191. Tahoe City. pp. Isidori. and P. pp. L. vol. 741–746. I. 2. Boston. R. 267–270. “Syst` mes d’´ quations diff´ rentielles d’oscillations non lin´ aires. Couvignou. Lewis. “New results and examples in nonlinear feedback stabilization. “Control and stabilization of nonholonomic dynamic systems.” in Proceedings of the Third IFAC Symposium on Nonlinear Control Systems Design (NOLCOS’95).” Revue de a e e e e math´ matiques pures et appliqu´ es. J. Escobar. a C. 1983. 4. 11. E. no. “Regulation and tracking of the nonholonomic double integrator: A field-oriented control approach. vol. 12. pp. and H. pp. K. 1994. California. Lake Buena Vista. MA. R. 1995. G. no. California. Reyhanoglu. 1992. San Diego. 1995. pp. R. Astolfi. M. New Orleans.

” in Proceedings of the 35th Conference on Decision and Control. 1393–1399. I. 15. Jankovi´ . Sun. Hurwitz. John Wiley and Sons Ltd. Lefeber. submitted to the 39th Conference on Decision and Control. “Ueber die Bedingungen. 1997. Fossen and J.” Systems and Control Letters. “Nonlinear tracking of underactuated surface vessels. 30–39. 1. Fossen. Nijmeijer.” Mathematische Annalen. Z. pp. Kokotovi´ . Tch´ and H. Nijmeijer and T. 975–980. Jiang. 265–279. unter welchen eine Gleichung nur Wurzeln mit negativen reelen Theilen besitzt.-P.” 1999. Japan. 1723–1735. pp. I. 35. pp. 46. Z. “Passive nonlinear observer design for ships using Lyapunov methods: Experimental results with a supply vessel. pp.-M.” in Proceedings of CONTROLO’98. Sydney. Fliess. Eds.-P.” in Proceedings of the 3rd European Control Conference. “A recursive technique for tracking control of nonholonomic systems in chained form. I.” Automatica. T. E. Jiang. pp. Australia. pp. Z. “Saturated stabilization and tracking of a nonholonomic mobile robot.-P. Ioannou and J. 273–284.” IEEE Control Systems Magazine. Kobe. USA: PrenticeHall. 255–262.” IEEE Transactions on Automatic Control. United Kingdom: Springer-Verlag.-P. London. W. J. vol. “Design of trajectory stabilizing feedback for driftless flat systems. “Observer-controller design for nonholonomic systems. “Iterative design of time-varying stabilizers for multi-input systems in chained form. “Constructive Lyapunov stabilization of c c nonlinear cascade systems. vol. 44. 1. 1999b. E. J.154 Bibliography M. Robust Adaptive Control. Nijmeijer. Guidance and Control of Ocean Vehicles. Martin. and A. Third Portuguese Conference on Automatic Control. pp. T. 1882–1887. Hager. V. 1998. D. Z. pp. vol. 1995. . 207–228. 1999. Jiang and H.-P. a mobile car. Jiang and H. 1994. and P.” in New trends in nonlinear observer design. vol. 7. Jiang and H. Nijmeijer. “Tracking control of mobile robots: a case study in backstepping. Lefeber. and H. A. Jakubiak. Nijmeijer.-P. pp. 1. C. 1996. R. 28. Nijmeijer. Z. vol. pp. K. 33. Z. Sepulchre. Nijmeijer. pp. “Robot hand-eye coordination based on stereo vision. Upper Saddle River. 1895. 12. Submitted to Systems and Control Letters. S. H. Jiang. Godhavn. Lefeber. no. G. no. “Stabilization and tracking of a nonholonomic mobile robot with saturating actuators. M.” IEEE Transactions on Automatic Control. Fossen.” 2000. no. 1996. vol. 1996. Strand. 41. Levine. no. “Observer based tracking controllers for on. 244 in Lecture Notes in Control and Information Sciences. 3– 16. Morse. vol. Rouchon. NJ. and H.” Automatica. 315– 320. Portugal.-P. no. P. vol. no. A. P. 1996.. J. 1995. 2. Rome. and P. Coimbra. E. Chang. 1999a.

I. Prentice-Hall. New Haven. vol.” IEEE Control Systems Magazine. MN. vol. NJ. McClamroch. A. 759–766. 6. F. Nonlinear and Adaptive Control Design. Kanellakopoulos. pp. 208–210. and Control.” IEEE Transactions on Robotics and Automation. and T. 12. A. 559–574. H. vol. pp. 16. 520–527. I.” IRE Transactions on Circuit Theory. “Fixed-eye direct visual feedback control of planar robots. Kapoor. Nonlinear Systems. 285–307. Kimura. Marquez. no. N. vol.” in Proceedings of the 1996 IEEE International Conference on Robotics and Automation.” International Journal of Adaptive Control and Signal Processing.” Robotica. USA. no. 1987. 1990. G. “Fixed camera visual servoing of planar robots. R. Koditschek. Krsti´ . 1982. “A stable tracking control method for an autonomous mobile robot. vol. Linear Systems. no.” Journal of Systems Engineering. Kolmanovsky and N. LaSalle. vol. no. 20–36. Khalil. “Hybrid feedback laws for a class of cascade nonlinear control systems. vol. Y. Noguchi. K. 5. pp. Shirkey. pp. no. 34. 4. Upper Saddle River. “Developments in nonholonomic control problems. 1996. . Kolmanovsky and N. 1980. 1995. “Preliminary results about robust Lagrange stability in adaptive regulation. McClamroch. second ed. Daoutidis. 239–248. 384–389. and M. pp. 1996. “Adaptive techniques for mechanical systems. R. no. 4. pp. I. Spong. no. “A tuning procedure for stable PID control of robot manipulators. John Wiley and Sons. Kelly and A. 1. 259–265. and P. 7. USA. Minneapolis. Phoenix. pp. Kern. Y. vol. Teel. 3. 1996. c c Series on Adaptive and Learning Systems for Signal Processing. R. USA: Prentice-Hall. pp.” in Proceedings of the 38th Conference on Decision and Control. pp.” IEEE Transactions on Automatic Control.” Automatica. Kanayama. 5. Arizona. H.” in Proceedings of the 5th Yale Workshop on Adaptive Systems. J.-P. pp.Bibliography 155 Z. 6. pp. Kojic and A. pp.” IEEE Transactions on Automatic Control. P. T. Communications. Kailath. 141–148. CT. 1998. “An anti-windup design for linear systems with input saturation. Miyazaki. 27. Kelly. 1995. Praly. 1996. vol. 1992. E. vol. M.” in Proceedings of the 1990 IEEE International Conference on Robotics and Automation. USA. Annaswamy. 9. R. “Adaptive control of nonlinearly parametrized systems with a triangular structure. Kelly. “Some extensions of Lyapunov’s second method. H. Jiang and L. 1271– 1282. “Uniform controllability of a class of linear time-varying systems. 5. and P. 1999. 41. 1995. 2643–2649. 1960. Kokotovi´ . 13. W. D. 1995. Kelly. “Robust asymptotically stable visual servoing of planar robots. P.

pp.” in Proceedings of the 5th European Control Conference.” in Stability and Stabilization of Nonlinear Systems. 5. 1996. 1997a. “Tracking control of an under-actuated ship. 1998. E. A.” in Proceedings of the 38th IEEE Conference on Decision and Control. “Adaptive tracking control of a wheeled mobile robot. Nijmeijer.” Journal A. A. A. Lefeber. 1998. E.” IEEE Transactions on Automatic Control. Karlsruhe. Y. Lefeber. University of Twente. Kelly.” Memorandum 1339. USA. 1999a. Enschede. “Bounded tracking controllers for the chaotic (forced) Duffing equation. Aeyels. University of Twente. abstract only. F.” in 17th Benelux Meeting on Systems and Control. no. vol. 144–152. “Output feedback tracking of nonholonomic systems in chained form.” Memorandum 1404. pp. Nijmeijer. 1996. pp. Z¨ rich. The Netherlands. “Globally bounded tracking controllers for rigid robot systems. and H. and A. 1995. 563–568.” in Preprints COSY Workshop on Control of Nonlinear and Uncertain Systems. J. Petersburg. Switzerland. E. Nijmeijer. 1999b. Nijmeijer. J. The Netherlands. “Adaptive and filtered visual servoing of planar robots.156 Bibliography A. 246 in Lecture Notes in Control and Information Sciences. 1999. Lefeber and H. E. “Bounded tracking controllers for the chaotic (forced) Duffing equation. paper 772. R. The Netherlands. 495–497. Nijmeijer. pp. 3. W. Lefeber and H. Lefeber and H. 35– 39.. 1997d. Nijmeijer. A. vol. St. Nijmeijer. “On the possible divergence of the projection algorithm. Nijmeijer. “Bounded global tracking of rigid robot manipulators using composite controllers. E. The Netherlands. Ortega. p. 3. A. and H. Nijmeijer. Lefeber and H.” in Proceedings of the 1st International Conference on Control of Oscillations and Chaos (COC’97). E. 1997b. A. Mierlo. Arizona. 415–418. Lefeber and H.” in Proceedings of the Fourth IFAC Symposium on Nonlinear Control Systems Design (NOLCOS’98). Eds. “Adaptive tracking control of nonholonomic systems: an example. van der Schaft. Robertsson. in preparation.” 2000a. “Bounded tracking controllers for the chaotic (forced) Duffing equation. 220. vol. J. J. Russia. A. 39. “Linear controllers for tracking chained-form systems. 1998. Faculty of Applied Mathematics. J. D. Brussels. United Kingdom: Springer-Verlag. paper 455 (TH-E G3). Enschede. London. 2. and H. Belgium. E. “(Adaptive) control of chaotic and robot systems via bounded feedback control. Germany. . Lefeber and J. Nijmeijer. vol. Enschede. vol. R. Enschede. 2. Robertsson. E. Nijmeijer. 183–199. A. pp. Lefeber. Lefeber and H.” MSc thesis. Lefeber. Polderman. University of Twente. The Netherlands. LamnabhiLagarrigue. u A. Lefeber and H.” in Proceedings of the 4th European Control Conference. Phoenix. and H. no. Lefeber. 2094–2099. K. Faculty of Applied Mathematics. A. Pettersen. Lefeber. “An example of a bounded tracking controller. pp. 40. no. J. A. E. pp. 1997c.

saturated controls. Tomei. Ed. A. pp. 201–208.” IEEE Transactions on Automatic Control. M. 1604–1609. INRIA. 1996.” in Proceedings of the 33rd Conference on Decision and Control. P. Germany. R. 700–716. pp. Sastry. Bordeaux. A. MA. Sastry. 1990. 121–126. pp. Karlsruhe. W. no. 1995. 11. A. Submitted to Stability and Control: Theory and Applications. vol. San Antonio. F. TX. Nonlinear Control Design. and S. Lake Buena Vista. pp. 243–264. From Routh-Hurwitz stability criterion to time-varying exponential stabilizers. 1994.” IEEE Transactions on Automatic Control. Murray. 10. Nijmeijer. P. 44. and H. pp. Lor´a. M. A. 1993. Lefeber. M. 1992. Oelen and J. 333–340. S. pp. vol. 500. F. 1634–1652. A. Teel. E. 5. 1997. 1999. USA. Cambridge.” in Proceedings of the 5th European Control Conference. France. vol.. “Control of nonlinear chained systems. “Visually guided robotic tracking and grasping of a moving object with a robotic hand-eye system. G. Lei and B. Micaelli and C. no. pp.” Rapport de recherche 2097. R. no. England: Prentice-Hall. and F. A. Lor´a. “Global asymptotic stability of robot manipulators ı with linear PID and PI¾ D control. Annaswamy. Miyazaki and Y.” in Robotics Research: The Fifth International Symposium. pp. .” in Proceedings of the 36th Conference on Decision and Control. 2. F. “A new notion of persistency-of-excitation for UGAS on ı NLTV systems: Application to stabilisation of nonholonomic systems. USA. Ghosh. Praly. and A.” IEEE Transactions on Automatic Control. 38. “Robust tracking control of two-degrees-of-freedom mobile robots. and stabilization for feedforward systems. 4. no. Samson. Robertsson. Walsh. “Stabilization and tracking for nonholonomic control systems using time-varying state feedback. Masutani. paper no. 109–114. Mazenc and L. 1559– 1578. E. FL. “Linear controllers for exponential tracking of systems in chained form. Morin and C. “Adaptation in the presence of a general nonlinear parameterization: An error model approach. 1993. K. Murray and S.” 1999a. Nijmeijer. Lefeber. M. “Nonholonomic motion planning: Steering using sinusoids. pp. Marino and P. USA: MIT Press.Bibliography 157 E. 41.-P. Hemel Hempstead Hertfordshire. Mazenc and L. “Adding an integration and global asymptotic stabilization of feedforward systems. 9.” in Proceedings of the 32nd Conference on Decision and Control. 1999b. Samson. Fliess. and H. 618–623.” Control Engineering Practice. vol. Loh.” in IFAC Nonlinear control systems design. R. “Adding integrations. 2000b. California. San Diego. 1993.” International Journal of Robust and Nonlinear Control. van Amerongen. Sophia-Antipolis. Praly. pp. “Robustness of sensory feedback contorl based on imperfect jacobian. USA. vol. 1994. “Trajectory tracking for unicycle-type and two-steering-wheels mobile robots. Panteley. Skantze.

221–226. 29.” 2000. “Feedback control of a nonholonomic wheeled cart in cartesian space. E. “Underactuated ship tracking control. 18.” Systems and Control Letters.” Systems and Control Letters. vol. C. pp. “A new proof of the Routh-Hurwitz stability criterion using the second method of Liapunov. no. Nantes. 1996. pp. Prieur and L.” in Proceedings of the 38th Conference on Decision and Control. 1998. vol. 1996. K. Panteley. 1991.” 1999. Rugh. W. Panteley. vol. Egeland. P. Nijmeijer. Panteley and A. “Exponential stabilization of an underactuated surface vessel. Teel. pp. “Control of chained systems application to path following and time-varying point-stabilization of mobile robots. “Exponential tracking control of a moı bile car using a cascaded approach. Linear System Theory. K.” in Proceedings of the IFAC Conference on System Structure and Control.” IEEE Transactions on Automatic Control. 1999. Ortega. Y. pp. 1998. Sacramento. Panteley and A. E. “Passivity properties for stabilization of nonlinear cascaded systems. 4. J. no. Prentice-Hall.158 Bibliography R.” in Proceedings of the IFAC Workshop on Motion Control. Japan. 1995. 1962. 1996. A. E. pp. and H. 64–77. USA. Submitted to Automatica. Pettersen and H. 1. pp. and A. Kobe. C. E. 99-160. “Explicit design of time-varying stabilizing control laws for a class of controllable systems without drift.” in Proceedings of the 1991 IEEE International Conference on Robotics and Automation. pp. Information and System Sciences Series. Samson and K.-B. 131–138. Submitted. Nijmeijer. CNRS. E. France. 58. Lor´a. 694–702.” Proceedings of the Cambridge Philosophical Society. Y. “Global practical stabilization and tracking for an underactuated ship — a combined averaging and backstepping approach. 1998. pp. France.” in Proceedings of the 35th Conference on Decision and Control. pp.” 1999. Arizona. .” Automatica. M. A. Pomet. second ed. pp. C. Pettersen and O. Ait-Abderrahim. “Uniting local and global controllers. Technical Report no. Nijmeijer. lab. Lor´a. USA. 967–971. Kobe. Parks. Japan. Lefeber. 1136–1141. 147–158. 1214–1219. “Control and stabilization of an underactuated surface vessel. “On global uniform asymptotic stability of nonlinear time-varying ı systems in cascade. theory and experiments. d’Automatique de Grenoble. Lor´a. “Growth rate conditions for stability of cascaded time-varying ı systems and their applications. Pettersen and H. K. Samson. vol. Praly. C.” in Proceedings of the 35th Conference on Decision and Control. vol. Reyhanoglu. J. Y. 59–64. Grenoble. no. pp. 1991. 1992. Lor´a. 2371– 2376. 33. 40. Phoenix. 423–424. “New tools for UGAS and ULES of nonlinear timeı varying systems: a Æ -persistency of excitation approach. 2.

“Stabilization of trajectories for systems with nonholonomic constraints. 5. 451–458. . Tilbury. J. Sontag. Sørdalen. Murray. 1989. J. 39. Walsh. 1995. Signals. 39. and Control. pp. Sastry. M. no. vol. pp. Brussels. 12. S. vol. and Y. R. 1. Val´ sek and N. pp. D. “Efficient pole placement technique for linear time-variant SISO aˇ systems. “Sufficient Lyapunov-like conditions for stabilization.” in Proceedings of the 3rd European Control Conference. vol. Wichlund. Sussmann. R. K. Olgac. M. H. 35–49. “A new feedback method for dynamics control of manipulators. “Uniting local and global controllers. vol. 4. vol. J. paper 959 (WE-E A3). Measurement.” International Journal of Control. and G. Laumond. 103. Y. and O. vol. Walsh. 1. “Control of vehicles with second-order nonholonomic constraints: Underactuated vehicles. 1994. D. pp. A.” IEE Proceedings — Control Theory and Applications. 849– 870. 1981. R.” IEEE Transactions on Automatic Control.” IEEE Transactions on Automatic Control. J. 142. Belgium. O. E. Takegaki and S. no. Egeland.” Transactions of the ASME Journal of Dynamic Systems. 1995. pp. 216–222. Rome. P. 1995. Sørdalen and O. Kapoor. Yang. 1997.” Mathematics of Control. M. 119–125.” in Proceedings of the 4th European Control Conference. Murray. Tsinias. 3086–3091. Teel. R. “Exponential stabilization of nonholonomic chained systems. “A general result on the stabilization of linear systems using bounded controls. and Systems. Teel and N. no. Italy. Egeland.” IEEE Transactions on Automatic Control. pp. 343–357. pp. pp. Arimoto. no. vol. G. 2. 1995. no. 40. and J. “Non-holonomic control systems: from steering to stabilization with sinusoids. 1994. A. 2411–2425.Bibliography 159 O. C. 62.

160 Bibliography .

the resulting closed-loop tracking error dynamics is considered. In Chapter 4 the kinematic model of a mobile car is considered. Due to the design this closed-loop tracking error dynamics has a cascaded structure. it is shown that the tracking error converges to zero. The behavior of the resulting controllers is illustrated by means of numerical simulations and in case of the ship by means of experiments on a scale model of an offshore supply vessel. Next. The tracking error at each time is defined as the difference between where the system is and where it should be.e. a car towing multiple trailers. a knife edge. systems that can be seen as a special interconnection of two stable subsystems.. no matter where the system is initialized nor at which time-instant. a trajectory that can be realized for the system under consideration. including a mobile robot. First the set-point control problem 161 . The problem now is to design a controller for the system which is such that the tracking error converges to zero. A large class of interesting mechanical systems can be transformed to the chained form. i. Next. based on the theory of cascaded systems. i. A new design methodology is presented. a vertical rolling wheel and a rigid spacecraft with two torque actuators.. All systems under consideration happen to have two inputs..e. This assumption simplifies the remaining dynamics considerably. In Chapter 6 the tracking problem for an under-actuated ship is dealt with.Summary The subject of this thesis is the design of tracking controllers for certain classes of mechanical systems. In the second part of the thesis some uncertainties are taken into account. it is assumed for the remaining dynamics that the stabilization of the first subsystem has worked out. a feasible reference trajectory is assumed to be given i.e. the remaining input is used for establishing asymptotic stability of the simplified remaining dynamics. whereas it has three degrees of freedom. Therefore. i. Having found in this manner control laws for the two inputs. By means of the theory of cascaded systems global uniform asymptotic stability of the tracking error dynamics is shown. In the cascaded design approach.e. first one of the inputs is used to stabilize a subsystem of the tracking error dynamics. In the first part an accurate mathematical model of the mechanical system under consideration is assumed to be given. The goal is to follow a certain specified trajectory. a ship with only two controls is considered.. concerning the models of the mechanical systems under consideration.. no matter where the system is initialized nor at which time-instant. The thesis consists of two parts. Chapter 5 is concerned with systems in chained form. This new approach is applied to three different models. i.e. This design strategy is applied to the three models mentioned above. it is assumed that for the first subsystem the state equals the reference state and the input equals the reference input.

but only after some period of time. One of the major difficulties is that some of the camera parameters are unknown.162 Summary for a rigid robot manipulator is studied in case the vector of gravitational forces is unknown. An operator determines a spot on the screen to which the tip of the manipulator should move and a controller has to be found which makes the manipulator do so. This is due to the fact that entirely specifying the reference trajectory is in conflict with not knowing certain parameters. The output of this camera is displayed at a screen. The fact that also the orientation angle of the camera is unknown leads to designing an adaptive controller to solve this problem. For this example of a four-wheel mobile robot with unknown length a formulation of the adaptive tracking problem is presented and also solved. Secondly. the visual servoing problem for a rigid robot manipulator is considered. but this time it is assumed that certain system parameters are unknown. the tracking problem of Part I is considered again. . The thesis ends with conclusions and recommendations for further research. It is shown that this approach leads to global asymptotic stability of the error dynamics provided that the integral action is not activated from the beginning. Since compensation for this vector is needed to achieve perfect regulation. Imagine that a rigid robot manipulator is moving in the horizontal plane and a camera is placed at the ceiling to watch the manipulator from above. By means of an example it is shown that the formulation of the adaptive tracking control problem is far from trivial. it is common practice to use a PID controller instead of a PD-controller with gravity compensation. Thirdly.

Deze nieuwe aanpak wordt toegepast op drie verschillende modellen. ongeacht waar of op welk tijdstip het regelsysteem ge¨nitialiseerd wordt. ı Deze ontwerpaanpak wordt toegepast op de drie eerder genoemde modellen. Dat betekent dat verondersteld wordt dat voor het eerste deelsysteem de toestand gelijk is aan de referentietoestand en de ingang gelijk is aan de referentie-ingang. Daarna wordt voor de resterende dynamica aangenomen dat de stabilisatie van het eerste deelsysteem gelukt is. wat wil zeggen dat de volgfout naar nul convergeert. wordt de resulterende volgfoutdynamica in gesloten lus bekeken. Nu er op deze manier regelwetten voor de twee ingangen gevonden zijn. Het doel is om een gespecificeerd traject te volgen. een mesblad. In het eerste deel wordt een wiskundig model van het mechanische regelsysteem met twee ingangen bekend verondersteld. Een grote klasse van interessante mechanische regelsystemen kan worden getransformeerd naar de “chained form”. Op elk tijdstip wordt de volgfout gedefinieerd als het verschil tussen waar het regelsysteem is en waar het zou moeten zijn (gezien het gewenste traject). In de cascade ontwerpaanpak wordt eerst een van de ingangen gebruikt om een deelsysteem van de volgfout-dynamica te stabilizeren.Samenvatting Dit proefschrift gaat over het ontwerpen van regelaars voor het volg-probleem voor bepaalde klassen van mechanische regelsystemen. Het gedrag van de resulterende regelaars wordt ge¨llustreerd met behulp van numerieke simulaties danwel. Met behulp van de theorie van cascade systemen wordt vervolgens globale uniforme asymptotische stabiliteit van de volgfout-dynamica aangetoond. Deze aanname vereenvoudigt de resterende dynamica aanzienlijk. ongeacht waar of op welk tijdstip het regelsysteem wordt ge¨nitialiseerd. In hoofdstuk 4 wordt het kinematische model van een mobiele robot beschouwd. Er wordt ı een nieuwe ontwerpmethode gepresenteerd die gebaseerd is op de theorie van cascade systemen. waarbij wordt aangenomen dat dat gewenste traject daadwerkelijk door het regelsysteem te volgen is. Dankzij de ontwerpaanpak heeft de volgfout-dynamica een cascade structuur. Hoofdstuk 5 gaat over regelsystemen in “chained form”. terwijl het drie vrijheidsgraden heeft. een wagen die een aantal opleggers trekt. systemen die gezien kunnen worden als een bijzondere verbinding van twee stabiele systemen. ı 163 . waaronder een robot karretje. Vervolgens wordt de ingang die nog over is gebruikt om asymptotische stabiliteit van de vereenvoudigde resterende dynamica te bewerkstelligen. In hoofdstuk 6 wordt het volg-probleem behandeld voor een schip met slechts twee stuurmiddelen. een verticaal rollend wiel en een ruimteschip met twee aandrijfmogelijkheden. Het probleem is nu om voor het regelsysteem een regelaar te ontwerpen die er voor zorgt dat de volgfout naar nul convergeert. Het proefschrift bestaat uit twee delen. dat wil zeggen.

heeft er toe geleid een adaptieve regelaar te ontwerpen om dit probleem op te lossen. Op een beeldscherm wordt weergegeven wat die camera ziet. Het doel is om een regelaar te ontwerpen die er voor zorgt dat de robotarm zich beweegt naar de plek die door iemand op het scherm is aangewezen. Een van de problemen die hierbij een rol speelt. Als eerste wordt gekeken naar het probleem om een stijve robotarm naar een vast punt te sturen waarbij de zwaartekrachtsvector onbekend verondersteld wordt. maar deze keer wordt aangenomen dat enkele parameters van het regelsysteem onbekend zijn. onder de voorwaarde dat de integrerende actie niet van het begin af aan. Het proefschrift wordt afgesloten met conclusies en aanbevelingen voor verder onderzoek. Als derde wordt het volg-probleem van deel I opnieuw bekeken. Dit komt doordat het volledig specificeren van het referentie-traject en het niet kennen van enkele parameters met elkaar in conflict is. . Het feit dat bovendien de orientatiehoek van de camera onbekend is. maar pas na enige tijd wordt geactiveerd. Aangezien deze vector gecompenseerd moet worden om perfecte positionering te krijgen is het gebruikelijk om een PID-regelaar te gebruiken in plaats van een PD-regelaar met compensatie voor de zwaartekracht. Voor dit voorbeeld van een vier-wielige mobiele robot met onbekende lengte wordt het adaptieve volg-probleem zowel geformuleerd als opgelost. is dat enkele van de parameters van de camera niet bekend zijn.164 Samenvatting in geval van het schip. Door middel van een voorbeeld wordt aangetoond dat het formuleren van een adaptief volg-probleem verre van triviaal is. Neem aan dat een stijve robotarm in een horizontaal vlak beweegt en dat een camera aan het plafond de robotarm van boven registreert. met behulp van experimenten op een schaalmodel van een zeewaardig bevoorradingsschip. Er wordt aangetoond dat deze aanpak leidt tot globale asymptotische stabiliteit van de foutdynamica. In het tweede deel van het proefschrift wordt rekening gehouden met een aantal onzekerheden in de modellen van de mechanische regelsystemen. Als tweede wordt het ‘visual servoing’ probleem voor stijve robotarmen bekeken.

His research subject is control of industrial systems. In the summer of that same year he was one of the six members of the team that represented The Netherlands at the International Mathematical Olympiad in Beijing. 2000 at the Department of Mechanical Engineering of Eindhoven Univerity of Technology. He started January 1st. In May 1996 he started a research project on “(adaptive) control of nonlinear mechanical systems with nonholonomic constraints”.E. After the summer of 1990 he started his studies at the Faculty of Mathematical Sciences at the University of Twente. China.ir. 165 . He was student-member of the VSNU-committee that was concerned with the educational visitation of the mathematics departments in The Netherlands. Nijmeijer. During the final years of his study he was involved in the optimal control of a solar-car and he graduated in April 1996 on the “(adaptive) control of chaotic and robot systems via bounded controls”. The results of this project can be found in this thesis.dr. developing material. In 1990 he finished the “Christelijk Gymnasium Sorghvliet” in The Hague. doing mathematics markets) and by being member of the editorial staff of Pythagoras. Rooda. J. He also was dedicated to popularizing mathematics by participating in VIERKANT voor Wiskunde (leading mathematics summer camps. During his studies he was an active member of the Mathematical Student Society Abacus. which was awarded the KIvI-regeltechniekprijs 1996. H. He is with the group of Systems Engineering of Prof.Curriculum Vitae Erjen Lefeber was born on the 7th of December 1972 in Beverwijk. During his last two years in Enschede he was elder of the Oosterkerk in Enschede.dr. supervised by Prof.

.

.Curriculum Vitae Deze pagina hoort er niet meer bij! Snij-lijnen zijn aangegeven voor formaat van 167 ½ ¢¾ cm.

Sign up to vote on this title
UsefulNot useful