You are on page 1of 13

Computers and Electrical Engineering 28 (2002) 241–253

www.elsevier.com/locate/compeleceng

Adaptive control of dynamic mobile robots with nonholonomic constraints
Farzad Pourboghrat *, Mattias P. Karlsson
Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL 62901-6603, USA Received 16 November 1999; accepted 22 August 2000

Abstract This paper presents adaptive control rules, at the dynamics level, for the nonholonomic mobile robots with unknown dynamic parameters. Adaptive controls are derived for mobile robots, using backstepping technique, for tracking of a reference trajectory and stabilization to a fixed posture. For the tracking problem, the controller guarantees the asymptotic convergence of the tracking error to zero. For stabilization, the problem is converted to an equivalent tracking problem, using a time varying error feedback, before the tracking control is applied. The designed controller ensures the asymptotic zeroing of the stabilization error. The proposed control laws include a velocity/acceleration limiter that prevents the robotÕs wheels from slipping. Ó 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Mobile robot; Nonholonomic constraint; Dynamics level motion control; Stabilization and tracking; Adaptive control; Backstepping technique; Asymptotic stability

1. Introduction Motion control of mobile robots has found considerable attention over the past few years. Most of these reports have focused on the steering or trajectory generation problem at the kinematics level i.e., considering the system velocities as control inputs and ignoring the mechanical system dynamics [1–3]. Very few reports have been published on control design in the presence of uncertainties in the dynamic model [4]. Some preliminary results on control of nonholonomic systems with uncertainties are given in Refs. [4–6]. Two of the most important control problems concerning mobile robots are tracking of a reference trajectory and stabilization to a fixed posture. The tracking problem has received solutions
*

Corresponding author. Tel.: +1-618-453-7026. E-mail address: pour@siu.edu (F. Pourboghrat).

0045-7906/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 5 - 7 9 0 6 ( 0 0 ) 0 0 0 5 3 - 7

Using backstepping technique we derive adaptive control laws that work even when the model of the dynamical system contains uncertainties in the form of unknown constants. The BrockettÕs condition essentially states that for nonholonomic systems an equilibrium solution can be asymptotically stabilized only by either a time varying. they gave several examples on how the stabilization problem can be solved for a mobile robot at the kinematics level.242 F. at the dynamics level. piecewise continuous control. the work presented in Refs. and hence the inertia. a discontinuous. Pourboghrat. particularly the mass.2]. Several other researchers have later used this reduced order state equation in their studies. in Ref. Our work is based on. [7]. including stabilization of general systems to an equilibrium manifold and stabilization of differentially flat and Caplygin systems to an equilibrium point. where the robots are to move a variety of parts with different shapes and masses. They also showed how a solution to the tracking problem could be extended to work even for the stabilization problem. The basic idea is to have a reference car that generates a trajectory for the mobile robot to follow. justifying an adaptive control approach. In Refs.P. was considered. nonlinear velocity control inputs were defined that made the tracking error go to zero as long as the reference car was moving. [1. [9]. M. Dynamic model of mobile robot Here. 1). In Ref. they proposed an adaptive controller for a number of important nonholonomic control problems. In Ref. Here. or a dynamic state feedback. we consider a three-wheeled mobile robot moving on a horizontal plane (Fig. In these cases.2. which states that for nonholonomic systems a single equilibrium solution cannot be asymptotically stabilized using continuous static state feedback [9. In addressing the above problem. [4]. 2. [9]. the robotÕs mass and inertia may vary up to 10% or 20%. In Ref. Using the reduced state equation in Ref. The assumption for the uncertainty in robotÕs parameters. as defined in their paper. they formulated a reduced order state equation for a class of nonholonomic systems. [10] a smooth feedback control was presented for the kinematics control problem resulting in a globally marginally stable closed loop system. They also designed a smooth feedback control for a dynamical state-space model resulting in a Lagrange stable closed loop system. The mobile robot features two differentially driven rear wheels and a castor front wheel. The desired trajectory was then tracked using sliding mode control. and timevarying piecewise continuous control. The radius of . A two dimensional Lyapunov function was utilized in Ref. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 including classical nonlinear control techniques [1. [2]. [1.10]. can be justified in real applications such as in automotive manufacturing industry and warehouses. we present adaptive control schemes for the tracking problem and for the problem of stabilization to a fixed posture when the dynamic model of the mobile robot contains unknown parameters. The problem of stabilization about a fixed posture has been shown to be rather complicated. the problem of controlling nonholonomic mechanical systems with uncertainties. In Ref.7]. resulting in discontinuous control signals. they used input–output linearization to make a mobile platform follow a desired trajectory. Their solutions included time-varying control. The mobile robot was shown to be exponentially stable for a class of quadratic Lyapunov functions.2]. and can be seen as an extension of. This is due to violating the BrockettÕs condition [8]. [3] to prescribe a set of desired trajectories to navigate a mobile robot to a specified configuration.

The dynamic model for the above wheeled-mobile robot is given by Refs. / q 3. and distance between the rear wheels. The reference trajectory qr ðtÞ ¼ ½xr ðtÞ. _ ¼ ½x _. the wheels is denoted r and the length of the rear wheel axis is 2l. as well as its derivative. /r ðtފT is generated by a reference vehicle/robot whose equations are 8 _ r ¼ vr cos /r <x _ ¼ vr sin / ð3Þ y : _r /r ¼ xr . q ¼ ½x. yr ðtÞ. y . wheel radius. respectively. The triplet vector function qðtÞ ¼ ½ xðtÞ. and that m and I denote the mass and the moment of inertia of the mobile robot.F. That is. at any time. Mobile robot configuration. y ðtÞ. We assume that. b2 ¼ l=ðrI Þ. q ¼ ½x. /ðtފT denotes the trajectory (position and orientation) of the robot with respect to a fixed workspace frame. The assumption that the signs of b1 and b2 are known is practical since b1 and b2 represent combinations of the robotÕs mass. coming from the assumption that the wheels do not slip. Eq. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 243 Fig.11]. moment of inertia. Pourboghrat. the robotÕs posture. Inputs to the system are two torques T1 and T2 . M. 8 k > sin / þ b1 u1 cos / x¼m <€ k € y ¼ Àm cos / þ b1 u1 sin / ð1Þ > :/ € ¼ b2 u2 _ sin / À y _ cos / ¼ 0 x ð2Þ where b1 ¼ 1=ðrmÞ. y . Also.P. given by k ¼ Àm/ unknown constants with known signs. u1 ¼ T1 þ T2 and u2 ¼ T1 À T2 are the control inputs. [10. at any given time. Here. and k is the _ ðx _ cos / þ y _ sin /Þ. 1. /ŠT . provided by two motors attached to the rear wheels. _ ŠT . are available for feedback. y _. (2) is the nonholonomic constraint. it is assumed that b1 and b2 are Lagrange multiplier. /ŠT describes the robotÕs configuration (posture) at that time. Tracking problem definition The tracking problem consists of making the trajectory q of the mobile robot follow a reference trajectory qr .

Assuming that the angles Note that since T matrix is nonsingular. Then. and are expressed as _ cos / þ y _ sin / v¼x _ x¼/ ð6Þ 0 1 ð4Þ 4. since the actual control variables u1 and u2 do not appear in Eq. v ~ðtÞ ¼ 0. under the Assumption A1 . we define the equivalent trajectory tracking error as ~ e ¼ Tq cos / sin / 0 where e ¼ ½e1 . (4) can be written as. qr . and vr and xr are the reference translational (linear) velocity and the reference rotational (angular) velocity. xr . e3 ŠT . x _ . and T ¼ @ À sin / cos / 0 A. respectively. the goal is to design a controller to force the tracking error e ¼ ½e1 . the derivative of the trajectory tracking error given in Eq. e3 ŠT to zero. where q ~ðtÞ ¼ qr ðtÞ À qðtÞ is defined as ¼ uðq. Using backstepping technique. In the sequel we refer to e as the trajectory tracking error. Pourboghrat. Assumption A1 . Using the nonholonomic constraint (2). 8 _ 1 ¼ e2 x À v þ vr cos e3 <e _ ¼ Àe1 x þ vr sin e3 ð5Þ e : 2 _ 3 ¼ xr À x e where v and x are the translational and rotational velocities of the mobile robot. Also let us define v written as ~ v ¼ vd þ v ~ x ¼ xd þ x ð7Þ . For the tracking problem it is assumed that the reference velocities vr and xr do not both go to zero simultaneously. Tracking controller design Here.  The  tracking problem. (5).P. as well as their derivatives are available and that they all are bounded. [1]. vr . Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 The subscript ‘‘r’’ stands for reference. respectively. q u2 the trajectory tracking error. with vd and xd the trajectory tracking error e converges to ~ as virtual control errors. e is nonzero as long as q /r and / are given in the range ½Àp. pŠ. is to find a feedback control law u1 _ r Þ such that lim t!1 q _ r. M. v and x can be ~ and x zero asymptotically. That is. we have the equivalent trajectory tracking error e ¼ 0 only if q ¼ qr .244 F. That is. The purpose of the tracking controller is to force the equivalent trajectory tracking error e to 0. e2 . [1]. As in Ref. it is assumed that at any time either lim t!1 vr ðtÞ90 and/or lim t!1 xr ðtÞ90. we consider variables v and x as virtual controls. Let vd and xd denote the desired virtual controls for the mobile robot. We assume that vr and xr . 0 0 1 ~ 6¼ 0. e2 .

c1 . any differentiable control law that makes the kinematics model of the mobile robot track a desired trajectory can be used instead of Eq. e2 . Taking the time derivative of V1 .F. If Assumption A1 holds. (1). M. as ~ and x In view of Eqs. Proof. we find the time derivatives of v _ þ€ _ Àv _ ¼v _ d ¼ b1 u1 À v _d ~ _ Àv _d ¼ € _ sin // _ cos // x cos / À x y sin / þ y v €Àx _ ¼x _ d ¼ b2 u2 À x _d ~ _ Àx _d ¼ / x ð13Þ ð12Þ ð11Þ . xr Þe1 xd ðvr . and c2 are positive constants and b 1 1 2 estimate of b2 ¼ 1=b2 . c2 . but with the advantage. Clearly V1 is positive definite and V1 ¼ 0 only if e ¼ 0. i. Observe that our approach from here on is general for any vd and xd (with well defined first derivatives). strictly positive on R  R-ð0. as we are going to prove later. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 245 Let us choose the virtual controls vd and xd . xr . e3 Þ ¼ xr þ k2 vr e2 þ k3 ðvr . Now. then the adaptive control scheme (9) makes the origin e ¼ 0 uniformly asymptotically stable. Pourboghrat. (7) and (8). xr Þ sin e3 ð8Þ where k2 is a positive constant and k1 ðÁÞ and k3 ðÁÞ are bounded continuous functions with bounded first derivatives. using Eqs. xr . that it can be used to track any reference trajectory as long as Assumption A1 holds. Consider the following Lyapunov function candidate V1 ¼ Á 1 1À 2 ð1 À cos e3 Þ e1 þ e2 2 þ 2 k2 ð10Þ where k2 is a positive constant.P. (8) is similar to the control law proposed by Ref. e3 Þ ¼ vr cos e3 þ k1 ðvr . we obtain _1 ¼ e1 ð À v þ vr cos e3 Þ þ e2 vr sin e3 þ 1 sin e3 ðxr À xÞ V k2 Furthermore. consider the following adaptive control scheme: ^ ðÀc1 v ~ þ e1 þ v _ dÞ u1 ¼ b 1   1 ^ _d ~ þ sin e3 þ x u2 ¼ b2 À c2 x k2 _ ^ ~ þ e1 þ v ~ðÀc1 v _ dÞ b 1 ¼ Àc1 signðb1 Þv   1 _ ^ ~ À c2 x _d ~ þ sin e3 þ x b2 ¼ Àc2 signðb2 Þx k2 ð9Þ ^ is an estimate of b ¼ 1=b1 and b ^ is an where c1 . 0Þ.e. e1 . we have 1 _1 ¼ Àk1 e2 À k3 sin2 e3 À v ~ sin e3 ~e1 À x V 1 k2 k2 ~ . (2) and (6). (8). Eq. Result 1. [1]. as vd ðvr .

and Eqs. V [12. V2 converges to a finite limit. ~ . using Eqs. again by BarbalatÕs lemma. and that k1 ðÁÞ and k3 ðÁÞ are bounded continuous functions with bounded first derivatives and are strictly positive on R  R-ð0. by uous. Since V2 is bounded from below and V ^ ^ are all bounded. Proceeding in the same manner. M. v must all converge to zero. i. x Also. Considering Eq. v 2 Furthermore. h In Section 3.e. e tend to zero as t ! 1. v ~ converge to zero ~. both vr e2 and xr e2 converge to zero. and k3 . e3 . To get a better understanding on how the control gains affect the response of the ~ are equal to zero as [1] ~ and x system.246 F. by _ 1 . e2 must converge to zero.13]. e equation can be written as _ 3 ¼ Àk2 vr e2 À k3 sin e3 e ð19Þ and its second derivative can be shown to be bounded. e1 . the third error BarbalatÕs lemma. (9) we get: ~ ¼b Àb where b 1 1 1 1 2 2 2 2 _2 ¼ Àk1 e2 À k3 sin2 e3 À c1 v ~2 6 0 ~2 À c2 x V ð15Þ 1 k2 _2 is negative semi-definite. Since e3 is differentiable and converges to _ 3 ! 0 as t ! 1. Clearly.. and x BarbalatÕs lemma. Pourboghrat. e3 . using the above results. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 Consider the Lyapunov function candidate 1 2 jb1 j ~2 jb2 j ~2 ~ 2Þ þ ~ þx V2 ¼ V1 þ ðv b þ b 2c1 1 2c2 2 2 ^ ¼ 1=b1 À b ^ and b ~ ¼b Àb ^ ¼ 1=b2 À b ^ . and from the above results. That is. e1 . k2 . e2 . k2 vr e2 and thus vr e2 zero and € e3 is bounded. b1 . e2 xr tend to zero. the second derivative of V2 can be written as _ 1 e2 €2 ¼ À2k1 e1 e2 ðxr þ k2 vr e2 þ k3 sin e3 þ x ~ Þ þ 2k1 e1 ðk1 e1 þ v ~Þ À k V 1 _ 2k3 k3 ^ ðÀc1 v ~ Þ À sin2 e3 À 2c1 v ~ðb1 b ~ þ e1 þ v _ dÞ À v _ dÞ þ cos e3 sin e3 ðk2 vr e2 þ k3 sin e3 þ x 1 k2 k2     1 ^ À c2 x _d _d Àx ~ b2 b ~ þ sin e3 þ x À 2c2 x ð16Þ 2 k2 which from the properties of k1 . we write the equations for the closed loop system when v . This in turn implies that e1 . Since V2 ðtÞ is differentiable and converges to some constant value and that V _2 ðtÞ ! 0 as t ! 1. Since e1 is differentiable and converges to zero and € e1 is bounded. and x not both tend to zero (by Assumption A1 ). and b ~. the first error equation can be written as _ 1 ¼ e2 xr À k1 e1 e The second derivative of e1 is _ 1 e1 _ r e2 þ xr ð À e1 x þ vr sin e3 Þ À k1 ðe2 xr À k1 e1 Þ À k € e1 ¼ x ð18Þ ð17Þ ð14Þ which can be shown to be bounded by once again using the properties of k1 . 0Þ. e2 . can be shown to be bounded. we demonstrated that the system is stable if k2 is a positive constant. the assumptions on vr and xr . (7)–(9) and (13). and hence. as well as. the assumption that vr and xr and their derivatives are _2 is uniformly continbounded. V2 . (5). V €2 is bounded. Hence. To show that e2 also goes to zero. e3 .P. (7) and (8). note that. However. since vr and xr do ~ ~.

we choose x0 ¼ x2 r r control gains then become À Á 2 1=2 k1 ¼ 2n x2 r þ bvr k2 ¼ b À Á 2 1=2 k3 ¼ 2n x2 r þ bvr ð25Þ and the resulting control is now defined for any values of vr and xr . However. [2]. In order to prevent the mobile robot from slipping. xd Þ by constants ðvmax . a simple velocity/acceleration limiter may be implemented [1].F. Pourboghrat. either or both of the virtual reference velocities in Eq. amax Þ. if the reference trajectory is a straight line connected to a circle segment). This limits the virtual reference velocities ðvd . The systemÕs closed loop poles are now equal to the roots of the following characteristic polynomial equation: À Á ð23Þ ðs þ 2nx0 Þ s2 þ 2nx0 s þ x2 0 where n and x0 are positive real numbers. in a real application.g. xmax Þ and the virtual reference _ d are the virtual reference _ d and a ¼ x accelerations ða. the translational/rotational acceleration might become too large causing the robot to slip [1]. these parameters must be determined experimentally as the largest values with which the mobile robot never slips.. where a ¼ v accelerations. xd Þ. (8) might become too large for a real robot to attain in practice. if the tracking errors initially are large or if the reference trajectory does not have a continuous curvature (e. aÞ by constants ðamax . The corresponding control gains are k1 ¼ 2nx0 k2 ¼ 2 x2 0 À xr v2 r ð24Þ k3 ¼ 2nx0 With a fixed pole placement strategy (n and x0 are constant). The depend on the values of vr and xr . it is shown that the proposed algorithm works for any desired velocities.P. ðvd . we assume that vr and xr are constants. As in Ref. Hence. in practice. In practice. we get _ ¼ Ae e where Àk1 A ¼ @ Àxr 0 0 xr 0 Àk 2 v r 1 0 vr A Àk3 0 1 ð20Þ ð21Þ ð22Þ To simplify the analysis. M. In the above. One way to avoid this is by À letting 2the Áð1=2closed Þ þ bv with b > 0. as shown in Fig. . Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 247 Àk1 e1 þ ðxr þ k2 vr e2 þ k3 sin e3 Þe2 @ _ ¼ Àðxr þ k2 vr e2 þ k3 sin e3 Þe1 þ vr sin e3 A e Àk2 vr e2 À k3 sin e3 By linearizing the differential equation (20) around e ¼ 0. 2. the control gain k2 increases loop poles without bound when vr tends to zero.

amax ¼ 5 m/s2 and amax ¼ 25 rad/s2 . Pourboghrat. is taken as qð0Þ ¼ ½xð0Þ. Tracking control structure. M. An important advantage of adding the limiter is that it lowers the control gains indirectly only when the tracking errors are large. first the robot backs up and then heads toward the virtual reference robot moving on the straight line. Figs. however. /ð0ފT ¼ ½0. starting from qr ð0Þ ¼ ½xr ð0Þ. vmax ¼ 1:5 m/s. 5 and 6 show the simulation results for tracking a circular trajectory. indirectly. is initially at qð0Þ ¼ ½xð0Þ. the robot immediately heads toward the reference robot. /ð0ފT ¼ ½1. using M A T L A B / S I M U L I N K . but with somewhat different transient characteristics. which are assumed to be unknown to the controller. 0ŠT . The computer simulations for the above controller without the limiter. yr ð0Þ. Simulation results for tracking control problem Here. Moreover. The mobile robot. 2. Also selected are. i. Thus. p=2ŠT .P.. by using the limiter one can have higher control gains for small tracking errors to allow for better tracking. the robotÕs dynamic parameters are chosen as b1 ¼ b2 ¼ 0:5. The reference velocity is kept constant at vr ðtÞ ¼ 0:5 m/s. The reference trajectory is given by xr ðtÞ ¼ 0:5t. Simulation results for the case where the reference trajectory is a straight line are shown in Figs. starting at qr ð0Þ ¼ ½xr ð0Þ. 0. for large tracking errors. It then reaches it quickly and continues to track it.e. while letting the limiter to ‘‘scale down’’ the gains. Again. 3 and 4 for t 2 ½0. 10Š. but with known signs. yr ðtÞ ¼ 0:5t and /r ðtÞ ¼ p=4. All simulations have the common parameters of c1 ¼ c2 ¼ 100. produce similar results. . 0. defining a straight line. while for small tracking errors it does not affect the performance at all.248 F. to prevent the robot from slipping. which is moving on the circle. where / ¼ 0 indicates that the robot is heading toward positive direction of x. yr ð0Þ. y ð0Þ. /r ð0ފT ¼ ½0. 0. although not shown here. the damping factor n ¼ 1. The initial conditions for the mobile robot. The reference trajectory is a point moving counter clockwise on a circle of radius 1. 5. c1 ¼ c2 ¼ 10 and b ¼ 250. As it can be seen from these figures. /r ð0ފT ¼ ½1. xmax ¼ 3 rad/s. y ð0Þ. p=4ŠT . when too high a gain could cause the robot to slip. the results of computer simulation. 0ŠT . 0. however. are presented for a mobile robot with the proposed tracking control and with the velocity/acceleration limiter. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 Fig. as it is seen from these figures.

0. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 249 Fig. 3. . Without loss of generality. 6. M. such that lim t!1 ðqðtÞ À qd Þ ¼ 0. Mobile robot and reference trajectories in the ðx. y Þ plane.F. we may take qd ¼ ½0. Pourboghrat. q u2 posture qð0Þ. for any arbitrary initial robot law. 4. 0ŠT . Fig. tÞ.P. is to find a feedback control   u1 _ . given an arbitrary desired posture qd . ¼ uðq À qd . Stabilization problem definition The stabilization problem. Time history of the tracking errors.

Time history of the tracking errors. Simulation results when the reference trajectory describes a circle. Fig. The stabilization to a fixed posture problem is . M. The idea is to let the reference vehicle move along a path that passes through the point ðxd . Pourboghrat. 5.P. Instead of designing a new controller for the stabilization problem the same controller as for the tracking problem is used. Stabilization controller design Recall that there is no continuous static state feedback that can asymptotically stabilize a nonholonomic system about a fixed posture [8–10]. The approach to the problem taken here is the dynamic extension of that in Ref. 6.1.250 F. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 Fig. [2] where a kinematics model of the mobile robot is used. 6. yd Þ with heading angle /d .

gðe. Fig. By taking the time _ r are _ r is bounded. Therefore.P. i. e2 . vr ¼ x with gðe.F. . [2]. tÞ is bounded. it can be shown in the same way that v bounded. from the Section 5. 0ŠT .11] and the references therein. yr ðtÞ ¼ 0 and /r ðtÞ ¼ 0. When e tends to zero. tÞ have also been suggested in the literature. [2. and can be treated as. 6. y ð0Þ. and e3 are bounded. the assumptions made in Section 3 concerning the reference velocity are fulfilled. (26). /ð0ފT ¼ ½0. for all values on t. in this case _ r ¼ Àk4 xr þ gðe. tÞ also tends to zero. As in Ref. the tracking errors e1 . M. 1. then e must converge to zero.e.2. Therefore vr and the state xr also remain bounded. If vr is not equal to zero. Pourboghrat. which converges to zero and hence lead the mobile robot to the desired posture. 7. However. Mobile robotÕs trajectory in posture stabilization simulation. Simulation results for stabilization control problem Here. The design method is the same as derived for the tracking case. 7 and 8. a tracking problem (convergence of the tracking errors to zero) with the additional requirement that the reference vehicle should itself be asymptotically stabilized about the desired posture. the time-varying function gðe. Since vr and v derivative of Eq. tÞ. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 251 now equivalent to. Different time-varying functions gðe. see Refs. the simulation results for the stabilization problem are shown in Figs. Since. The mobile robot is initially at qð0Þ ¼ ½xð0Þ. The control parameters and system parameters are the same as for the simulations shown for the tracking problem and k4 ¼ 1. tÞ ¼ kek sin t ð27Þ ð26Þ where k4 > 0. the robotÕs position x must track xr . we let the reference vehicle move along the xaxis.

Recent trends in Mobile robots. [4] Colbaugh R. 384–9. Samson C. [2] Canudas de Wit C. Sordalen OJ. A simple velocity/acceleration limiter was added to the controller. Barany E. editor. Pourboghrat. Utkin VI. Kimura Y. Nonlinear control design for mobile robots. that the robot actually turns around and backs up into the final posture. Noguchi T.P. . and /. namely. p. [3] Guldner J. Stabilization of nonholonomic mobile robot using Lyapunov functions for navigation and sliding mode control. Ohio. As it is seen from the figures. Note. for practical applications. and to improve the tracking performance. In: Zheng YF. vol. 8. Japan. 1428–34. tracking of a reference trajectory and stabilization to a fixed posture. Miyazaki F. Time plots of x. 121–56. Conclusions Two important control problems concerning mobile robots with unknown dynamic parameters have been considered. the stabilization about the final posture at the origin is achieved quite satisfactorily. A stable tracking control method for an autonomous mobile robot. Glass K. in this case.10(4):635–47. Adaptive Control of Nonholonomic Mechanical Systems. World Scientific. p. p. M. 1990. References [1] Kanayama Y.252 F. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 Fig. Several simulation results have been included to demonstrate the performance of the proposed adaptive control law. y. to avoid any slippage of the robotÕs wheels. Kobe. 1. Proceedings of IEEE International Conference on Robotics and Automation. Control-Theory Adv Technol 1994. Cincinnati. 7. Proceedings of 35th Conference on Decision and Control. 1996. An adaptive control law has been proposed for the tracking problem and has been extended for the stabilization problem. Khennouf H. 1993.

Kumar V. Sussmann HJ. Lewis FL. Int J Robot Res 1994. Control of mechanical systems with rolling constraints: application to dynamic control of mobile robots. IEEE Contr Sys Magaz 1995. Millman RS. Canudas de Wit C. USA. Kanellakopoulos I. McClamroch NH.D. Sweden and the M. . Kokotovic P.15(6):20–36.13(1):55–69.14(3) 149-163. Developments in nonholonomic control problems. [7] Sarkar N. d’Andrea-Novel B. [8] Brockett RW. respectively.E. In: Brockett RW. [6] Jiang ZP.P. 2207–12.F. 106–24. in 1995 and 1997. He is now with the Department of Electrical and Computer Engineering at Southern Illinois University at Carbondale (SIU-C) where he is an Associate Professor. IL 62901. Proceedings of 33rd IEEE Conf on Decision and Control.E. Boston. 181–91. Master Thesis. [11] Kolmanovsky I. IL. [13] Karlsson MP. 1997. Asymptotic stability and feedback stabilization.E. He has been employed at Orian Technology since 1997. [12] Krstic M. Berlin: Springer. editors. Nonlinear and Adaptive Control Design. His current interests include control algorithm development for mechanical and electrical systems and software development for embedded systems. Farzad Pourboghrat received his Ph. 1983. flexible structures and MEMS. Control of nonholonomic systems with applications to mobile robots.Õs Technical Center. Mattias Karlsson received the B. degree from Southern Illinois University. Differential Geometric Control Theory. degree in Electrical Engineering from the University of Iowa in 1984. Carbondale. MA: Birkhauser. 1991. p. New York: Wiley.S. [9] Bloch AM. He is currently an on-site consultant at Caterpillar Inc. IL. Lake Buena Vista. editor.S. p. His research interests are in adaptive and sliding control with applications to DSP embedded systems. Combining backstepping and time-varying techniques for a new set of adaptive controllers. mechatronics. p. IEEE Trans Automat Contr 1992. Pomet JB. Mossville. McClamroch NH. M. J Robot Sys 1997. Southern Illinois University. Karlsson / Computers and Electrical Engineering 28 (2002) 241–253 253 [5] Fierro R. FL. 1995. Reyhanoglu MR. Control of nonholonomic mobile robot: backstepping kinematics into dynamics. Control and stabilization of nonholonomic dynamic systems. Pourboghrat. degree from the University of Bor as. Yun X.37(11):1746–56. [10] Campion G. Controllability and state feedback stabilization of nonholonomic mechanical systems. Bastin G.E. 1994. Advanced Robot Control. Carbondale.