Textbook Intelligent Optimal Adaptive Control For Mechatronic Systems 1St Edition Marcin Szuster Ebook All Chapter PDF

Intelligent Optimal Adaptive Control for
Mechatronic Systems 1st Edition

Marcin Szuster
Visit to download the full and correct content document:
https://textbookfull.com/product/intelligent-optimal-adaptive-control-for-mechatronic-sy
stems-1st-edition-marcin-szuster/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Adaptive Dynamic Programming with Applications in

Optimal Control 1st Edition Derong Liu
https://textbookfull.com/product/adaptive-dynamic-programming-
with-applications-in-optimal-control-1st-edition-derong-liu/
Adaptive Critic Control with Robust Stabilization for

Uncertain Nonlinear Systems Ding Wang
https://textbookfull.com/product/adaptive-critic-control-with-
robust-stabilization-for-uncertain-nonlinear-systems-ding-wang/
Intelligent Control of Robotic Systems 1st Edition

Laxmidhar Behera
https://textbookfull.com/product/intelligent-control-of-robotic-
systems-1st-edition-laxmidhar-behera/
Reinforcement Learning for Optimal Feedback Control

Rushikesh Kamalapurkar
https://textbookfull.com/product/reinforcement-learning-for-
optimal-feedback-control-rushikesh-kamalapurkar/
Adaptive aeroservoelastic control 1st Edition Tewari
https://textbookfull.com/product/adaptive-aeroservoelastic-
control-1st-edition-tewari/
Intelligent Algorithms for Analysis and Control of

Dynamical Systems Rajesh Kumar
https://textbookfull.com/product/intelligent-algorithms-for-
analysis-and-control-of-dynamical-systems-rajesh-kumar/
Adaptive and Fault-Tolerant Control of Underactuated

Nonlinear Systems 1st Edition Jiangshuai Huang
https://textbookfull.com/product/adaptive-and-fault-tolerant-
control-of-underactuated-nonlinear-systems-1st-edition-
jiangshuai-huang/
Force and Position Control of Mechatronic Systems:

Design and Applications in Medical Devices Tong Heng
Lee
https://textbookfull.com/product/force-and-position-control-of-
mechatronic-systems-design-and-applications-in-medical-devices-
tong-heng-lee/
Optimal Control in Thermal Engineering 1st Edition

Viorel Badescu
https://textbookfull.com/product/optimal-control-in-thermal-
engineering-1st-edition-viorel-badescu/
Studies in Systems, Decision and Control 120
Marcin Szuster
Zenon Hendzel
Intelligent
Optimal
Adaptive Control
for Mechatronic
Systems
Studies in Systems, Decision and Control
Volume 120
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control- quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems,
Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace
Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems, and others. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enables both a wide and rapid
dissemination of research output.
More information about this series at http://www.springer.com/series/13304

Marcin Szuster Zenon Hendzel
•
Intelligent Optimal Adaptive

Control for Mechatronic
Systems
123
Marcin Szuster Zenon Hendzel
Department of Applied Mechanics Department of Applied Mechanics
and Robotics, Faculty of Mechanical and Robotics, Faculty of Mechanical
Engineering and Aeronautics Engineering and Aeronautics
Rzeszow University of Technology Rzeszow University of Technology
Rzeszow Rzeszow
Poland Poland
ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-3-319-68824-4 ISBN 978-3-319-68826-8 (eBook)
https://doi.org/10.1007/978-3-319-68826-8
Library of Congress Control Number: 2017956748
© Springer International Publishing AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
From Marcin Szuster
To my wife Sylwia, for supporting me all
the way.
.
.
.
From Zenon Hendzel

To my grandsons Julian, Jan, Dominik, …
To learn is not to know.
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Artificial Intelligence and Neural Networks . . . . . . . . . . . . . . . 4
1.2 Learning with a Critic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Scope of Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Object of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Two-Wheeled Mobile Robot . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Description of the Kinematics of a Mobile Robot . . . . 12
2.1.2 Description of the Dynamics of a Mobile Robot . . . . . 21
2.2 Robotic Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.1 Description of the Kinematics of a Robotic
Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 32
2.2.2 Description of the Dynamics of a Robotic
Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 41
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 49
3 Intelligent Control of Mechatronic Systems . . . . . . . . . . . . . . . . . . 51
3.1 Methods for Control of Nonlinear Systems . . . . . . . . . . . . . . . 51
3.2 Neural Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2.1 Random Vector Functional Link Neural Network . . . . 56
3.2.2 Neural Network with Gaussian-Type Activation
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 57
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 58
4 Optimal Control Methods for Mechatronic Systems . . . . . . . . . . . . 61
4.1 Bellman’s Dynamic Programming . . . . . . . . . . . . . . . . . . . . . 61
4.2 Linear-Quadratic Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Pontryagin’s Maximum Principle . . . . . . . . . . . . . . . . . . . . . . 71
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
vii
viii Contents
5 Learning Methods for Intelligent Systems . . . . . . . . . . . . . . . . . . . . 85

5.1 Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.1 Steepest Descent Algorithm . . . . . . . . . . . . . . . . . . . . 86
5.1.2 Variable Metric Algorithm . . . . . . . . . . . . . . . . . . . . 87
5.1.3 Levenberg–Marquardt Algorithm . . . . . . . . . . . . . . . . 88
5.1.4 Conjugate Gradient Method . . . . . . . . . . . . . . . . . . . 89
5.2 Learning with a Critic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.2.1 Q-Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . 92
5.3 Learning Without a Teacher . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.3.1 Winner-Take-All Networks . . . . . . . . . . . . . . . . . . . . 93
5.3.2 Winner-Take-Most Networks . . . . . . . . . . . . . . . . . . . 94
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6 Adaptive Dynamic Programming - Discrete Version . . . . . . . . . . . . 97
6.1 Neural Dynamic Programming . . . . . . . . . . . . . . . . . . . . . . . . 97
6.2 Model-Based Learning Methods . . . . . . . . . . . . . . . . . . . . . . . 101
6.2.1 Heuristic Dynamic Programming . . . . . . . . . . . . . . . . 102
6.2.2 Dual-Heuristic Dynamic Programming . . . . . . . . . . . . 106
6.2.3 Global Dual-Heuristic Dynamic Programming . . . . . . 117
6.3 Model-Free Learning Methods . . . . . . . . . . . . . . . . . . . . . . . . 120
6.3.1 Action-Dependent Heuristic Dynamic
Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7 Control of Mechatronic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.1 Tracking Control of a WMR and a RM with a PD
Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.1.1 Synthesis of PD-Type Control . . . . . . . . . . . . . . . . . . 130
7.1.2 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.1.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.2 Adaptive Tracking Control of a WMR . . . . . . . . . . . . . . . . . . 140
7.2.1 Synthesis of an Adaptive Control Algorithm . . . . . . . 141
7.2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
7.3 Neural Tracking Control of a WMR . . . . . . . . . . . . . . . . . . . . 148
7.3.1 Synthesis of a Neural Control Algorithm . . . . . . . . . . 148
7.3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.4 Heuristic Dynamic Programming in Tracking Control
of a WMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.4.1 Synthesis of HDP-Type Control . . . . . . . . . . . . . . . . 157
7.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.5 Dual-Heuristic Dynamic Programming in Tracking Control
of a WMR and a RM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Contents ix
7.5.1 Synthesis of DHP-Type Control . . . . . . . . . . . . . . . . 175

7.5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
7.6 Globalised Dual-Heuristic Dynamic Programming in
Tracking Control of a WMR and a RM . . . . . . . . . . . . . . . . . 195
7.6.1 Synthesis of GDHP-Type Control . . . . . . . . . . . . . . . 196
7.6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
7.7 Action Dependent Heuristic Dynamic Programming in
Tracking Control of a WMR . . . . . . . . . . . . . . . . . . . . . . . . . 214
7.7.1 Synthesis of ADHDP-type Control . . . . . . . . . . . . . . 215
7.7.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.8 Behavioural Control of WMR’s Motion . . . . . . . . . . . . . . . . . 224
7.8.1 Behavioural Control Synthesis . . . . . . . . . . . . . . . . . . 228
7.8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
7.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
7.9.1 Selection of Value of the Future Reward Discount
Factor c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
8 Reinforcement Learning in the Control of Nonlinear
Continuous Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
8.1 Classical Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . 256
8.1.1 Control Synthesis, Stability of a System,
Reinforcement Learning Algorithm . . . . . . . . . . . . . . 256
8.1.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
8.2 Approximation of Classical Reinforcement Learning . . . . . . . . 266
8.2.1 Control System Synthesis, Stability of the System,
8.2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
8.3 Reinforcement Learning in the Actor-Critic Structure . . . . . . . 271
8.3.1 Synthesis of Control System, System Stability,
8.3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
8.4 Reinforcement Learning of Actor-Critic Type in the Optimal
Adaptive Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
8.4.1 Control Synthesis, Stability of a System,
x Contents
8.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286

8.5 Implementation of Critic’s Adaptive Structure in Optimal
Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
8.5.1 Control Synthesis, Critic’s Learning Algorithm,
Stability of a System . . . . . . . . . . . . . . . . . . . . . . . . . 287
8.5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
9 Two-Person Zero-Sum Differential Games and H1 Control . . . . . 299
9.1 H1 control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
9.2 A Two-Person Zero-Sum Differential Game . . . . . . . . . . . . . . 301
9.3 Application of a Two-Person Zero-Sum Differential Game
in Control of the Drive Unit of a WMR . . . . . . . . . . . . . . . . . 302
9.3.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
9.4 Application of a Neural Network in the Two-Person
Zero-Sum Differential Game in WMR Control . . . . . . . . . . . . 308
9.4.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
10 Experimental Verification of Control Algorithms . . . . . . . . . . . . . . 317
10.1 Description of Laboratory Stands . . . . . . . . . . . . . . . . . . . . . . 317
10.1.1 WMR Motion Control Stand . . . . . . . . . . . . . . . . . . . 317
10.1.2 RM Motion Control Stand . . . . . . . . . . . . . . . . . . . . 319
10.2 Analysis of the PD Control . . . . . . . . . . . . . . . . . . . . . . . . . . 321
10.2.1 Analysis of the WMR Motion Control . . . . . . . . . . . . 321
10.2.2 Analysis of the RM Motion Control . . . . . . . . . . . . . 326
10.2.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
10.3 Analysis of the Adaptive Control . . . . . . . . . . . . . . . . . . . . . . 329
10.3.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
10.4 Analysis of the Neural Control . . . . . . . . . . . . . . . . . . . . . . . 333
10.4.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
10.5 Analysis of the HDP Control . . . . . . . . . . . . . . . . . . . . . . . . . 337
10.5.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
10.6 Analysis of the DHP Control . . . . . . . . . . . . . . . . . . . . . . . . . 347
10.6.2 Analysis of the RM Motion Control . . . . . . . . . . . . . 352
10.6.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
10.7 Analysis of the GDHP Control . . . . . . . . . . . . . . . . . . . . . . . 357
Contents xi

10.7.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
10.8 Analysis of the ADHDP Control . . . . . . . . . . . . . . . . . . . . . . 362
10.8.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
10.9 Analysis of Behavioral Control . . . . . . . . . . . . . . . . . . . . . . . 366
10.9.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
10.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
Chapter 1
Introduction
From the earliest days, man has desired to create machines that would reflect on
the human condition i.e. machines that would demonstrate autonomy, capacity for
movement and that would learn and adapt to changing environmental conditions.
For many centuries it was not possible for man to build machines and devices that
would have at least some of these characteristics. It was not until the 20th century that
the rapid development of knowledge in fields such as automatic control, computer
science, electronics and manufacturing processes allowed for the construction of
robots i.e. machines with complex mechanical structures supplied with appropriate
control software that could perform certain tasks previously done by humans. Fur-
ther progress continued in such areas as robotics and mechatronics i.e. disciplines
that are concerned with the mechanics, design, control and operation of robots. Cur-
rent development of science and technology challenges the scientific community to
provide innovative engineering solutions and encourages undertaking research on
optimal solutions for mechatronic systems which include, inter alia, wheeled mobile
robots and robotic manipulators. The issues relating to mechatronics are of inter-
disciplinary nature and require the knowledge of multiple disciplines. In particular,
wheeled mobile robots are nonlinear, nonholonomic mechatronic systems composed
of interacting mechanical, electrical, electronic components and software.
The class of robots that perform behaviors with a high degree of autonomy includes
wheeled mobile robots, whose primary task is to execute a collision-free motion in
their surrounding environment. Such robots can be used for inspection, patrol, main-
tenance, search and transportation tasks or for exploration of unknown environments
e.g. other planets [4]. The last few years have seen a significant growth of inter-
est in mobile robotics and increased research on wheeled mobile robots and legged
robots. This is due to the improvement of mechanical components (sub-assemblies)
used for building robots and development of microprocessor-based software systems
that enable precise control of complex mechanical systems and reduce design and
manufacturing costs of such devices. These factors are driving the increased interest
© Springer International Publishing AG 2018 1
M. Szuster and Z. Hendzel, Intelligent Optimal Adaptive Control
for Mechatronic Systems, Studies in Systems, Decision and Control 120,
https://doi.org/10.1007/978-3-319-68826-8_1
2 1 Introduction
in personal mobile robots and service robots that have an enormous development
potential and offer a wide range of applications. There is demand for mobile devices
that could replace humans at tedious, strenuous, dangerous tasks or tasks which pose
health risks, and would provide assistance to elderly or disabled persons.
Robotic manipulators are the second class of robots which are used for other
applications. They resemble the human arm in structure and in many of its functions.
Manipulators are increasingly used to replace humans on repetitive and arduous tasks
such as operation of hydraulic presses and production lines. Due to their high preci-
sion of movement and repeatability, manipulators are used in advanced mechanical
processing of machine parts or assembly operations e.g. assembly of electronic cir-
cuits. Not only does the use of robotic manipulators ensure high performance and
repeatability of tasks carried out but it also leads to increased efficiency and, in
consequence, reduction of production costs. Industrial manipulators are frequently
used in high-volume production, particularly in automotive, electro-mechanical and
chemical industries, and environments where human exposure to toxic and explosive
gases occurs.
The early development of wheeled mobile robots dates back to the 1950s with the
establishment of the first fully automated factories in the United States where mobile
robots were utilized to automate manufacturing tasks such as material transfer and
handling. However, the work on further development of robots was discontinued due
to control-related difficulties, in particular, insufficient control techniques. It was not
until the 1970s when major progress in automation of transport occurred, driven by
growing demand for robotic solutions in electronics, pharmaceutical and automotive
industries. The adopted solutions were cost-effective, which spurred interest in robo-
tization of transport operations and increased research efforts. Construction of the
first servo-controlled electric-powered teleoperator in the 1940s marked the starting
point for development of robotic manipulators [23]. The first programmable robotic
arm was created in 1954. This followed years of rapid development of new robotic
manipulator designs and expansion of their capabilities and areas of application.
The increased interest in robotics has resulted in a large volume of scientific publi-
cations on robot design, kinematics and dynamics, control and trajectory generation.
One of the key elements of a robot is the control system which, together with
the sensory system, determines the capabilities and usability of a device. Currently,
complex robot control systems are mostly software programs executed by a process-
ing unit linked to a sensory system and actuators of a robot. Given the nature of
operations, a complex robot control system uses, inter alia, control algorithms for
trajectory generation and execution.
A desired robot trajectory is executed by the tracking control system that generates
control signals making the selected point attached to a robot follow the movement
of a virtual point according to predefined parameters of motion.
Initially, tracking control systems employed simple controllers that consisted of
proportional (P), derivative (D) and integral (I) elements. However, they did not
deliver the required control performance with regard to the systems described by
nonlinear dynamic equations i.e. robotic manipulators and wheeled mobile robots.
Therefore, other types of control systems have been introduced. They consist of the
1 Introduction 3
aforementioned controllers and additional algorithms for the generation of control

signals, where object dynamics is taken into account. Such control systems com-
prise robust systems that include parametric uncertainties related to the modeling of
controlled object dynamics, and adaptive systems that can adjust the parameters of
a control algorithm to variable robot operating conditions.
Over the last twenty years, the rapid development of artificial intelligence methods
such as artificial neural networks, fuzzy logic systems and genetic algorithms, cou-
pled with the development of microprocessor technologies, has enabled the design
of more complex control systems. Artificial neural networks are the most commonly
used controller structures in robotics, as they are adaptable and can be used for any
nonlinear mapping. Fuzzy logic systems are also found useful, as they allow for
development of control systems for complex objects on the basis of all available
scientific information. Control systems that use artificial intelligence methods have
the ability to adjust systems’ parameters to variable operating conditions of a device
and can perform complex tasks, thereby demonstrating a high level of autonomy
and machine intelligence. Application of artificial neural networks to control algo-
rithms often involves the use of learning with a teacher methods. The downside of
such methods is that the reference signal to be mapped by the network needs to
be predefined. Another approach to artificial neural networks involves the recently
developed reinforcement learning methods. These methods take inspiration from
animal and human behavior, where learning process consists in interaction with the
environment and assessment of environment’s response to a particular action given a
defined objective function. Neural dynamic programming algorithms are examples
of reinforcement learning algorithms.
Development of trajectory generation algorithms for mobile robots is also inspired
by the study of animal locomotion. Such observations have led to the development of
methods for trajectory planning e.g. behavioral control. This approach allows for the
design of trajectory generation systems with a simple structure and high degree of
autonomy that results from their capacity for adaptation to unknown environmental
conditions when carrying out complex tasks. The popularity of these issues is due to
the fact that such algorithms have practical applications and are used for a variety of
tasks e.g. reaching a target point in an unknown environment, investigation of conta-
minated areas, exploration of planets, or other human-defined tasks such as mowing,
patrolling, area search, cleaning. Artificial intelligence methods, in particular rein-
forcement learning algorithms, demonstrate considerable potential for application in
trajectory generation systems of mobile robots, which is reflected in current research.
Generation and execution of mobile robots’ trajectory are among the essential
areas of robotics. So far, no universal methods have been devised to solve these
highly complex problems. Solutions of the trajectory generation and tracking control
problems presented in the scientific publications are merely of academic nature, and
only a few have been implemented. Most of the proposed methods for trajectory
generation algorithms are offline solutions that have no real applications. Current
research efforts focus on development of practical mechatronic methods for motion
planning and control of wheeled mobile robots and robotic manipulators.
4 1 Introduction
Despite the rapid development of algorithms for trajectory generation and execu-
tion, there are still problems to be addressed. These include the execution of robots’
planned motion with regard to the dynamics of a controlled object, and trajectory
planning supported with the operation of the sensory system and its application in
behavioral control so as to execute complex tasks.
1.1 Artificial Intelligence and Neural Networks
The term “artificial intelligence”(AI) was coined by John McCarthy in 1956 [21], to
define an interdisciplinary field of science whose primary goal of research is access
and incorporation of human intelligence into machines. There are many different
definitions of artificial intelligence in the available literature. According to the def-
inition proposed by R.J. Schalkoff artificial intelligence seeks to solve problems
by emulating natural behavior and cognitive processes of humans with the use of
computer programs [21]. Artificial intelligence methods include, in particular, fuzzy
logic systems, evolutionary algorithms and artificial neural networks.
Research into evolutionary algorithms [11, 21] is nature-inspired. In other words,
it consists in the imitation of natural processes, where every living organism has its
own unique genetic material that stores information about inherited traits. In repro-
duction, the offspring inherits a complete set of traits encoded in genes. However,
in the phase of chromosomal crossover i.e. exchange of genetic material between
maternal and paternal chromosomes, mutations may occur which results in the off-
spring inheriting traits from both parents as well as having individual ones that differ
from those found in either parent. Should such set of traits facilitate the organism’s
adaptation to environmental conditions, it will pass on the genetic material to its
offspring. This approach has been used for solving optimization problems, where
a population of individuals is identified as a pool of potential solutions to a given
problem. The degree of their environmental adaptation (so-called fitness) is mea-
sured by means of an adequate fitness function. These individuals exchange their
genetic material i.e. chromosomal crossover and mutations occur, and thus generate
new solutions that are then selected in a fitness-based process to maximize values of
the fitness function. Hence, successive generations of descendants (offspring) offer
solutions that converge towards the optimal solution.
Development of fuzzy logic systems was initially undertaken in 1965 with the
introduction of fuzzy sets proposed by Lotfi Zadeh [11, 12, 21], which in an ill-
defined way associate numerical values with linguistic variables such as “very big”,
“big”, “small”. Fuzzy logic operations resemble human reasoning which is rather
approximate than precise in nature. The area of mathematics that deals with fuzzy
information is called “fuzzy set theory”. Its key element is fuzzy logic which is used
in fuzzy control and modeling.
Artificial neural network approach seeks alternative solutions to various compu-
tational problems through application of knowledge related to the brain structure and
brain functions of living organisms [11, 12, 16, 17, 21, 24]. It is an interdisciplinary
1.1 Artificial Intelligence and Neural Networks 5
field linked to automatic control, biocybernetics, electronics, applied mechanics and

medicine as well.
The study of nervous system activity of living organisms has contributed to the
structure design and development of methods for artificial neural networks, in par-
ticular computational techniques, where in a problem-solving process the network’s
output can be configured through learning, as opposed to an algorithm which is
written in form of a program. These methods can be applied to problems where the
answer is not known and where traditional algorithms fail due to difficulties and
constraints. This is particularly the case for correlation analysis tasks such as diag-
nostics, classification, forecasting and recognition. Neural networks operate based on
parallel distributed processing (connectionism) where a large number of single and
simple units (neurons), each of which demonstrates slow activity, perform complex
computational tasks as they process data simultaneously. Information processing is
spread, thus a partial damage to the structure of network may not have a negative
impact on its functioning. Other characteristics of neural networks include its capac-
ity for generalization and approximation of nonlinear mapping, and low sensitivity
to errors. That is why these algorithms are used in input-output relations (where rela-
tionships are not fully known) and also for solving identification and optimization
problems, where input data are inaccurate or partially incorrect. Due to their capa-
bilities, neural networks are commonly applied in different areas of knowledge e.g.
robotics, where they are used for the modeling and control of the systems described
by nonlinear dynamic equations - such as robots. The simplest single-layer neural
network with a single output consists of a weight vector (where weights adjust in the
learning process) and a neuron activation function vector. The cross product of these
two vectors is the output of the network.
1.2 Learning with a Critic
Development of the academic field related to artificial neural networks dates back
to 1940s when McCulloch and Pitts [14] published the first works that introduced
a systematic description of biological neural networks, which was followed by for-
mulation of the perceptron model. The term “perceptron”was first used in 1957 by
Frank Rosenblatt [20] to describe the machine he designed. It was in part an electro-
mechanical and electronic device that recognized alphanumeric characters and used
an innovative learning algorithm for system programming. The rapid development
of neural networks was virtually put on hold for over a decade due to the 1969 pub-
lication by Minsky and Papert [15] which highlights the limitations of perceptrons.
It was found, however, that these limitations did not concern neural networks with
nonlinear activation functions. Moreover, the development of learning methods for
multilayer neural networks based on backward propagation of errors, which took
place in the early 1980s, led to further dynamic development of neural networks. It
was established that learning with a teacher method did not apply to all cases, due to
lack of the reference signal. New concepts were developed i.e. unsupervised learning
6 1 Introduction
and learning with a critic [1, 3, 16, 22]. Learning with a critic is a kind of supervised
learning, which is also called reinforcement learning. It consists in interaction with
the environment by taking actions and evaluating the effects of these actions. The
main advantage is that there is no requirement for the output values to be presented;
it shall be sufficient to determine whether the undertaken action produces relevant
results in terms of desired system behavior. Given the action undertaken by the sys-
tem has a positive effect with regard to the adopted evaluation criterion, a similar
behavior tendency for analogous future situations is strengthened (i.e. reinforced),
otherwise it is weakened. Learning with a critic requires the use of an adaptive algo-
rithm, referred to as critic, that evaluates the current control strategy. Learning with
a critic is more versatile in application than a typical learning with a teacher method,
as in the learning with a critic approach the values of desired output signals may be
unknown. However, this method requires a properly defined system for evaluating
the undertaken actions in terms of a desired input-output behavior. The practical
implementation of this neural network learning method is more complex compared
to supervised learning. There are many different algorithms based on the learning
with a critic method e.g. neural dynamic programming algorithms [1, 5, 6, 18, 19],
Q-learning algorithm [1, 3, 7, 9], R-learning [1] and SARSA [1, 3].
1.3 Scope of Study
The purpose of this book is to prove the viability of reinforcement learning techniques
and in particular the family of techniques known as adaptive dynamic programming,
in optimal control of discrete and continuous mechatronic systems. This work dis-
cusses the problem of tracking control of dynamic objects such as wheeled mobile
robot and robotic manipulator. It also addresses the issue of a defined trajectory gen-
eration with regard to a wheeled mobile robot moving in an unknown environment
with static obstacles.
This type of a sequential decision making process may be resolved by applying
the theory of dynamic programming. Dynamic programming, which is used in opti-
mal control, is an alternative method to the calculus of variations. It was developed
in the 1950s by R. Bellman [2]. This method can be used for solving optimal control
problems with regard to nonlinear, non-stationary systems. Despite its elegant math-
ematical form, the practical applications of dynamic programming are somewhat
limited. For this procedure to be applied, a description of the controlled system is
required, and with the large dimension of the state vector or the large number of
discrete steps “curse of dimensionality” occurs.
Adaptive dynamic programming method is applied by defining an iterative
sequence of value functions and control signals, and is based on the principle of
optimality, temporal difference learning, and uses two parametric structures (actor
and critic). It comprises an approximate solution to the Hamilton–Jacobi–Bellman
equation [13].
1.3 Scope of Study 7
The extension of this subject is the differential game theory, construed as an

element of decentralized control, widely used for complex mechatronic systems. The
application of neural dynamic programming allows for the solution of two-player
zero-sum game.
The theory of two-player zero-sum differential games and H∞ control follow
from the solution to the Hamilton–Jacobi–Isaacs (HJI) equation [25], which is a
generalization of the Hamilton–Jacobi–Bellman equation.
The issue of tracking control of wheeled mobile robots or robotic manipulators
is a difficult one, due to the nonlinear character of the equations that describe the
dynamics of such devices, whose mathematical model parameters might be unknown
and time-varying, depending on the operating conditions.
In mechanics, tracking control relates to solving an inverse dynamics problem
in order to calculate the control signal that would compensate nonlinearity of the
controlled object. The theory of robust and adaptive systems was used in the synthe-
sis of wheeled mobile robot tracking control algorithms. Stability of the proposed
solutions is ensured by Lyapunov stability theory.
The problem of generating collision-free trajectories, in an unknown environment
with static obstacles, consists in designing a feasible trajectory of an arbitrary point
on a wheeled mobile robot in real-time, based on environment feedback provided by
the robot’s sensory system. This requires the acquisition of signals from proximity
sensors and proper interpretation of information to gauge the distance of obstacles,
and subsequently the generation of a feasible trajectory to be executed by a robot
that ensures the performance of the adopted task.
The designed tracking control and trajectory generation algorithms were numer-
ically tested in Matlab-Simulink computing environment, with the use of a mobile
robot model, robotic manipulator model and a measurement test environment. The
experimental verification of algorithms’ performance was carried out on lab stations
dedicated for motion analysis of a wheeled mobile robot and robotic manipulator.
The station for motion analysis of a wheeled mobile robot consists of the Pio-
neer 2-DX robot with a power supply and a PC equipped with dSpace DS1102
DSP controller board, dSpace Control Desk experiment management software and
Matlab - Simulink computing package. The station for motion analysis of a robotic
manipulator consists of the Scorbot-ER 4pc manipulator with a power supply and a
PC equipped with dSpace DS1104 DSP controller board, dSpace Control Desk and
Matlab-Simulink software. The stations are adapted for scientific research such as
rapid prototyping of control systems for dynamic objects and their verification with
the use of physical objects controlled in real-time. The control system is represented
in form of code, written in Matlab-Simulink computing environment, which is com-
piled into object code compatible with dSpace Control Desk software that controls
the experiment being run on dSpace DSP controller board. Once the experiment is
launched, the board generates control signals in real-time, which is dictated by the
adopted algorithm. The signals are generated based on the available measurement
signals from the controlled object’s sensory system. Simultaneously, data acquisi-
tion continues. This approach serves as a rapid and convenient research method with
regard to the designed algorithms.
8 1 Introduction
In most cases, the process of experimental research on designed algorithms

requires adjustments to the values of certain control parameters and further numerical
tests once the parameters are tuned.
The quality of motion generated with the proposed tracking control algorithms
for a wheeled mobile robot, with the application of neural dynamic programming
methods, has been compared with the quality of the motion generated with the use
of a PD controller, adaptive control algorithm and neural control system, which are
described in publications [8, 10]. Furthermore, this book focuses on tracking con-
trol algorithms for the robotic manipulator with the application of selected neural
dynamic programming methods. To evaluate the quality of tracking control, perfor-
mance indexes have been defined based on the tracking errors relating to the desired
trajectory.
This book discusses several important aspects of mechatronics. Chapter 2 includes
a description of controlled systems i.e. a two-wheeled mobile robot and a robotic
manipulator. The kinematics and dynamics of a control object is presented with
example solutions to inverse kinematics and inverse dynamics problems. Chapter 3
presents intelligent control methods for nonlinear systems illustrated with an exam-
ple of neural control algorithm. It also provides basic information on artificial neural
networks in terms of the implementation of tracking control algorithms for dynamic
systems. Neural networks are further explained in the chapters to follow. Chapter 4
examines optimal control methods for dynamic systems such as Bellman’s dynamic
programming, linear-quadratic regulator and Pontryagin’s maximum principle. It
also provides an example of discrete-time linear system control with the use of opti-
mal methods. Chapter 5 discusses learning methods for intelligent systems such as
supervised learning, learning with a critic and learning without a teacher. Chapter 6
describes selected neural dynamic programming algorithms, in particular methods
requiring knowledge of the mathematical model of a controlled object, such as heuris-
tic dynamic programming algorithm, dual heuristic dynamic programming algorithm
and globalised dual heuristic dynamic programming algorithm, and methods that
do not require such knowledge i.e. action-dependent heuristic dynamic program-
ming algorithm. Chapter 7 presents the following tracking control algorithms for a
wheeled mobile robot: the PD controller, adaptive control algorithm, neural control
algorithm, control algorithm that uses neural dynamic programming structures - in
the configuration of heuristic dynamic programming, dual-heuristic dynamic pro-
gramming, globalised dual-heuristic dynamic programming and action-dependent
heuristic dynamic programming. This chapter also presents control algorithms for
a robotic manipulator such as: the PD controller, control algorithm that uses neural
dynamic programming structures - in the configuration of dual-heuristic dynamic
programming and globalised dual-heuristic dynamic programming. Descriptions of
individual control algorithms are illustrated with numerical test results which are
derived from tracking control simulations of a dynamic object. Furthermore, this
chapter provides an example of the use of neural dynamic programming algorithms
in behavioral control of a wheeled mobile robot. Chapter 8 discusses methods for con-
tinuous nonlinear systems control, in particular classic reinforcement learning and
its approximation, and reinforcement learning based on the actor-critic structure.
1.3 Scope of Study 9
Descriptions of individual control algorithms are supplemented with numerical test

results that are based on control simulations of a dynamic object with the use of the
presented method.The following chapter focuses on two-player zero-sum differential
games and H∞ control. The issues addressed in this book were used for the control
of mobile robot drive unit and for the tracking control of a wheeled mobile robot. The
results obtained in the simulation tests are provided. Chapter 10 includes a descrip-
tion of the laboratory stations used for the verification tests of control algorithms
and discusses the experiments run with the use of the control systems presented
in Chap. 7. Chapter 11 presents the summary of findings and conclusions from the
conducted research.
References
1. Barto, A., Sutton, R.: Reinforcement Learning: an Introduction. MIT Press, Cambridge (1998)
2. Bellman, R.: Dynamic Programming. Princeton University Press, New York (1957)
3. Cichosz, P.: Learning Systems. (in Polish) WNT, Warsaw (2000)
4. Fahimi, F.: Autonomous Robots - Modeling, Path Planning, and Control. Springer, New York
(2009)
5. Ferrari, S.: Algebraic and Adaptive Learning in Neural Control Systems. Ph.D. Thesis, Prince-
ton University, Princeton (2002)
6. Ferrari, S., Stengel, R.F.: An adaptive critic global controller. In: Proceedings of American
Control Conference, vol. 4, pp. 2665–2670. Anchorage, Alaska (2002)
7. Gaskett, C., Wettergreen, D., Zelinsky, A.: Q-learning in continous state and action spaces.
Lect. Notes Comput. Sci. 1747, 417–428 (1999)
8. Giergiel, J., Hendzel, Z., Zylski, W.: Modeling and Control of Wheeled Mobile Robots. (in
Polish) Scientific Publishing PWN, Warsaw (2002)
9. Hagen, S., Krose, B.: Neural Q-learning. Neural. Comput. Appl. 12, 81–88 (2003)
10. Hendzel, Z., Trojnacki, M.: Neural Network Control of Mobile Wheeled Robots. (in Polish)
Rzeszow University of Technology Publishing House, Rzeszow (2008)
11. Jamshidi, M., Zilouchian, A.: Intelligent Control Systems Using Soft Computing Methodolo-
gies. CRC Press, London (2001)
12. Kecman, V.: Learning and Soft Computing. MIT Press, Cambridge (2001)
13. Lewis, F.L., Vrabie, D., Syroms, V.L.: Optimal Control, 3rd edn. Wiley, New Jersey (2012)
14. McCulloch, W.S., Pitts, W.: A logical calculus of the ideas immanent in nervous activity. Bull.
Math. Biophys. 5, 115–133 (1943)
15. Minsky, M., Papert, S.: Perceptrons: an Introduction to Computational Geometry. MIT Press,
Cambridge (1969)
16. Osowski, S.: Neural Networks. (in Polish) Warsaw University of Technology Publishing House,
Warsaw (1996)
17. Osowski, S.: Neural Networks - an Algorithmic Approach. (in Polish) WNT, Warsaw (1996)
18. Powell, W.B.: Approximate Dynamic Programming: Solving the Curses of Dimensionality.
Willey-Interscience, Princeton (2007)
19. Prokhorov, D., Wunch, D.: Adaptive critic designs. IEEE Trans. Neural Netw. 8, 997–1007
(1997)
20. Rosenblatt, F.: The perceptron: a probabilistic model for information storage and organization
in the brain. Psychol. Rev. 65, 386–408 (1958)
21. Rutkowski, L.: Methods and Techniques of Artificial Intelligence. (in Polish) Scientific Pub-
lishing PWN, Warsaw (2005)
10 1 Introduction
22. Si, J., Barto, A.G., Powell, W.B., Wunsch, D.: Handbook of Learning and Approximate
Dynamic Programming. IEEE Press Wiley-Interscience, Hoboken (2004)
23. Spong, M.W., Vidyasagar, M.: Robots Dynamics and Control. (in Polish) WNT, Warsaw (1997)
24. Tadeusiewicz, R.: Neural Networks. (in Polish) AOWRM, Warsaw (1993)
25. Van Der Schaft, A.: L2 - Gain and Passivity Techniques in Nonlinear Control. Spriner, Berlin
(1996)
Chapter 2
Object of Research
This chapter describes the objects of research i.e. a two-wheeled mobile robot and
a robotic manipulator with three degrees of freedom. It presents the structure, com-
ponents and basic parameters of the controlled systems. The chapter discusses the
kinematics of both robots and provides a simulation for solving the inverse kine-
matics problem. The desired trajectory of a given system was generated, based on
a pre-planned path and velocity values of a selected point attached to a robot. The
generated trajectories were used for numerical tests and verification analyses of track-
ing performance with regard to the controlled systems. The chapter discusses also
the dynamics of both controlled systems and provides a simulation for solving the
inverse dynamics problem for a given trajectory generated earlier. This resulted in
waveforms of control signals for a given trajectory.
2.1 Two-Wheeled Mobile Robot
Pioneer 2-DX wheeled mobile robot (WMR) is a two-wheeled robot intended for
laboratory use. It is equipped with the so-called caster, which is a third self-adjusting
supporting wheel, whose dynamics was not taken into consideration, and is assumed
to be negligible. Mobile robots with a similar structure are known as two-wheeled
robots since they have two drive wheels. It is to be noted that in the last few years
new robots have been introduced that have no extra supporting wheel.
The Pioneer 2-DX model [1] is shown in Fig. 2.1a, while its dimensional repre-
sentation is shown in Fig. 2.1b.
The model is composed of:
• drive wheels 1 and 2,
• caster 3,
• frame 4.
© Springer International Publishing AG 2018 11

M. Szuster and Z. Hendzel, Intelligent Optimal Adaptive Control
for Mechatronic Systems, Studies in Systems, Decision and Control 120,
https://doi.org/10.1007/978-3-319-68826-8_2
12 2 Object of Research
Fig. 2.1 a Pioneer 2-DX wheeled mobile robot. b Pioneer 2-DX schematic diagram; main physical
dimensions
Pioneer 2’s total mass is m D X = 9 [kg], and its maximum capacity (payload) is
m L x = 20 [kg]. Drive wheels 1 and 2 are made of rubber and their main physical
dimensions are: radius r = r1 = r2 = 0.0825 [m] and width h w1 = h w2 = 0.037 [m].
The maximum velocity of point A, which is attached to the Pioneer 2-DX
WMR, is vmx = 1.6 [m/s]. The WMR is equipped with a sensory system con-
sisting of 8 ultrasonic sensors, affixed to the circumference on the front of the
frame. The sensors are denoted by su1 , . . . , su8 as marked in Fig. 2.1b. In the pro-
gramming phase, the maximum measuring range of rangefinders was limited to
dmx = 4 [m]. If the distance between an obstacle and the WMR is greater than
the adopted limit, a default value dmx is assumed. The minimum measuring dis-
tance from an obstacle is dmn = 0.4 [m]. If the distance between an obstacle and
the WMR is less than dmn , a default value dmn is assumed. Considering the above,
the distance measurements made by the ultrasonic sensors fall in the range dsi ∈
0.4, 4.0[m], i = 1, . . . , 8. The deviations of ultrasonic sensors’ axes from the
frame’s axis of symmetry are as follows: ωs1 = 90◦ , ωs2 = 50◦ , ωs3 = 30◦ , ωs4 =
10◦ , ωs5 = −10◦ , ωs6 = −30◦ , ωs7 = −50◦ , ωs8 = −90◦ .
An analysis of the WMRs motion is linked to the kinematics and dynamics of
these objects which is further discussed in the subchapters below.
2.1.1 Description of the Kinematics of a Mobile Robot
The issues related to the kinematics of the WMR include, inter alia, forward and
inverse kinematics problems. A problem-solving approach to forward kinematics
consists in the determination of the WMR’s position and orientation relative to a
stationary frame of reference, where the motion parameters of the drive units and
2.1 Two-Wheeled Mobile Robot 13
the geometry of the system are known. To solve the inverse kinematics problem, the
angular parameters of drive wheels rotation need to be determined, given that a pre-
defined path and velocity of a selected point of the system are known. An analysis
of the inverse kinematics problem is carried out to determine the trajectory of the
WMR that is to be executed by the tracking control system [7, 21].
Selection of a characteristic WMR’s point that is to follow the desired path depends
on the vehicle’s navigation system and the design of the control system. The desired
path for a selected WMR’s point is represented by rectilinear and curvilinear segments
[19] so as to ensure the completion of a defined task. There are many factors that
affect the path design such as the type of tasks to be carried out, the structure and
technical capabilities of the WMR. Section 7.8 provides a synthesis of the WMR’s
trajectory generation layer of the hierarchical control system. The layer generates
control signals that are necessary to determine the trajectory for the WMR’s selected
point. This process is carried out on an ongoing basis during the performance of tasks
such as “goal seeking with obstacle avoidance”, which is a combination of two task
types: “goal seeking”and “keep to the center of the empty space/obstacle avoidance”.
The kinematics of a selected point of the system is analyzed by means of kine-
matic equations that may be established through application of different methods
e.g. Denavit–Hartenberg convention [19] which employs homogenous coordinates
and transformation matrices. Another approach is to use the classic methods applied
in mechanics, in particular an analytical description of motion that is based on the
parametric equations of motion [21].
In the following considerations it is assumed that all the WMR’s components are
perfectly rigid and the motion is executed on a flat horizontal surface. The WMR
motion is described based on a model (see Fig. 2.2) corresponding to the structure of
the Pioneer 2–DX robot used in the verification of solutions that are presented later
in this work. The investigated WMR is a nonholonomic system with two degrees
Fig. 2.2 Schematic diagram of velocities of individual points on wheeled mobile robot’s frame
of freedom. Two independent variables were used for motion description i.e. drive
wheels 1 and 2 rotation angles, denoted by α1 and α2 , respectively. In the case where
angular velocity vectors of the WMR drive wheels take the same values, the robot’s
frame performs translational motion, whereas if angular velocity vectors of individual
wheels have different values, the robot’s frame is in plane motion (i.e. it moves in
the plane (surface) of motion defined by axes x y).
In the analysis of the inverse kinematics problem, it was assumed that the selected
point H on the WMR moves on a given path with a desired velocity. The execution
of trajectory is ensured by defining appropriate angular velocities for the rotation of
the WMR drive wheels 1 and 2. It is also assumed that the WMR wheels roll without
slipping and the velocity vector of the characteristic point A lies in the plane parallel
to the plane of motion and is directed perpendicularly to a line segment bounded by
points B and C [7, 21].
In the case where the velocities of points B and C are such that v B > vC , the
WMR’s frame is in the plane motion and point F shown in Fig. 2.2 is the instantaneous
center of rotation of the frame.
Projections of the WMR’s point A velocity vector on stationary coordinate system
axes x and y satisfy the relation
ẏ A = ẋ A tan (β) , (2.1)
where β – instantaneous angle of rotation of the WMR’s frame.

It follows from Eq. (2.1) that point A velocity vector is bounded by nonholonomic
constraints. Based on the geometry of the WMR the relation between points H and
A in the x y system can be determined
x H = x A + l3 cos (β) , (2.2)
y H = y A + l3 sin (β) . (2.3)
Differentiating the above relations with respect to time, we get
ẋ H = ẋ A + l3 β̇ sin (β) , (2.4)
ẏ H = ẏ A + l3 β̇ cos (β) . (2.5)
Given that v A is the value of point A velocity vector, then values of vector projections
on the coordinate system axes x y are as follows
ẋ A = v A cos (β) , (2.6)
ẏ A = v A sin (β) . (2.7)

Plugging relations (2.6) and (2.7) into the set of Eqs. (2.4) and (2.5), we obtain
ẋ H = v A cos (β) + l3 β̇ sin (β) , (2.8)
ẏ H = v A sin (β) + l3 β̇ cos (β) . (2.9)
It is assumed that point H moves in the x y plane, on a path that is analytically

expressed as
f (x H , y H ) = 0 , (2.10)
where by differentiating (2.10) with respect to time, we get
f˙ (x H , y H ) = 0 . (2.11)
Given the velocity of point A and the desired path of point H , the set of Eqs. (2.8),
(2.9) and (2.11) allows for the calculation of the change in the values of the following
parameters
x H = x H (t) ,
y H = y H (t) , (2.12)
β = β (t) .
Relations that describe the velocities of points B and C, when written as vectors,
take the following form
vC = v A + vC A ,
(2.13)
vB = vA + vB A ,
and in the scalar form

vC = v A + vC A ,
(2.14)
vB = v A + vB A ,
where
vC A = β̇l1 ,
(2.15)
v B A = −β̇l1 .
It is assumed that the WMR drive wheels roll without slipping, hence the following
relations are satisfied
v B = α̇1r1 ,
(2.16)
vC = α̇2 r2 ,
where r1 = r2 = r – radiuses of the WMR drive wheels.

By considering Eqs. (2.14)–(2.16), the angular velocities α̇1 and α̇2 of individual
WMR drive wheels can be written as

α̇1 1 1 l1 vA
= . (2.17)
α̇2 r 1 −l1 β̇
The inverse form of relation (2.17) takes the form

vA r 1 1 α̇1
= −1 −1 . (2.18)
β̇ 2 l1 l1 α̇2
Kinematic relations described by Eqs. (2.17) and (2.18) are considered later in this
work. To solve the inverse kinematics problem the values of the following variables
need to be calculated
x H = x H (t) ,
y H = y H (t) ,
β = β (t) , (2.19)
α̇1 = α̇1 (t) ,
α̇2 = α̇2 (t) .
by means of Eqs. (2.8), (2.9), (2.11) and (2.19), taking into account the desired path
for point H and the velocity of point A v A .
Given that distance l3 = 0, two trajectories of point A were generated relative to
the desired paths. The first being a loop-shaped path with loop radius R = 0.75 [m],
and the second being an 8-shaped path consisting of two combined loops, each of
radius R = 0.75 [m].
2.1.1.1 Simulation of Loop-Shaped Path Inverse Kinematics
The WMR’s point A motion on the desired loop-shaped path, from the initial position
at point S to the desired final position at point G, is divided into five characteristic
phases [7]:
(a) Start-up (motion on a rectilinear path):
v∗A vA
vA = t − t p , t p ≤ t < tr , α̇1 = α̇2 = , β̇ = 0 ,
tr − t p r
where v A – linear velocity of point A, t – time, v∗A – maximum desired linear velocity
of point A, t p – (motion) initial time, tr – end-time of start-up, α̇1 , α̇2 – desired
angular velocities of the WMR drive wheels rotation, r – drive wheel radius, β̇ –
instantaneous angular velocity of the WMRs frame rotation.
(b) Motion with the desired velocity, when v A = v∗A = const.:
vA
α̇1 = α̇2 = , tr ≤ t < tc1 , β̇ = 0 ,
r
where tc1 – start-time of curvilinear motion.
(c) Motion in a circular path with radius R, when v A = v∗A = const., R = 0.75 [m]:
vA vA
α̇1 = + l1 β̇ , α̇2 = − l1 β̇ , tc1 ≤ t < tc2 ,
r r
where tc2 – end-time of curvilinear motion, l1 – length that results from the WMR’s
geometry.
(d) Curve exit including the transition period, followed by motion on a rectilinear
path with constant velocity, when v A = v∗A = const.:
vA v
A
α̇1 = α̇ p1 − α̇ p1 − 1 − e−ςt , α̇2 = α̇ p2 − − α̇ p2 1 − e−ςt ,
r r
tc2 ≤ t < th ,
where th – braking start-time, ς – transition curves approximation constant, α̇ p1 , α̇ p2

– values of wheels’ angular velocities at the beginning of the transition period. Intro-
ducing approximation allows for the execution of system’s motion with smooth
changes being applied to parameters such as velocity and acceleration.
(e) Braking:
v∗A vA
v A = v∗A − (t − th ) , th ≤ t < tk , α̇1 = α̇2 = , β̇ = 0 ,
tk − t h r
where tk – end-time.
Equal lengths of time were adopted for the braking and the start-up phase. The
start-time values for the remaining phases are respectively: t p = 3 [s], tr = 5 [s], tc1
= 7.5 [s], tc2 = 20.5 [s], th = 23 [s], tk = 25 [s]. The maximum linear velocity value
of point A on the WMR is v∗A = 0.4 [m/s]. To eliminate discontinuities in angular
accelerations of trajectory with a triangular velocity profile, an approximation of the
velocity profile was generated by means of the following relation
1
v A = v∗A · −c (t−t )
, (2.20)
1 + exp 1 1 1 + expc1 (t−t2 )
where c1 = 12 [1/s] – sigmoid functions slope coefficients, t1 – motion mean start-

time, t1 = 0.5(tr + t p ) = 4 [s], t2 – braking mean start-time, t2 = 0.5(tk + th ) =
24 [s]. The diagrams of the assumed triangular velocity profile and its approximation
are shown in Fig. 2.3a, b, respectively.
Based on the desired velocity profile (Fig. 2.3b), the desired motion path of point
A on the WMR (Fig. 2.4a) and the geometry of the WMR, the inverse kinematics
problem was solved, thereby producing angular variables related to the WMR drive
wheels rotation. Figure 2.4b shows the diagram of rotation angles α1 and α2 [rad] of
the WMR drive wheels, Fig. 2.4c presents the angular velocities α̇1 and α̇2 [rad/s],
whereas in Fig. 2.4d the angular accelerations α̈1 and α̈2 [rad/s2 ] are presented.
Fig. 2.3 a WMR’s point A triangular velocity profile, b approximation of point A triangular velocity
profile
Fig. 2.4 a WMR’s point A desired motion path, b WMR drive wheels rotation angles α1 and α2 ,
c angular velocities α̇1 and α̇2 , d angular accelerations α̈1 and α̈2
2.1.1.2 Simulation of 8-Shaped Path Inverse Kinematics
The WMR’s point A motion on the desired 8-shaped path, from the initial position
at point S to the desired final position at point G, is divided into two characteristic
phases (I and II), each of which consists in a “drive” on a loop-shaped path. Motion
phases are separated by a stop at point P. Each phase consists of nine characteristic
sub-phases of motion:
(a) Start-up (motion in a rectilinear path):
v∗A vA
vA = t − t p1 , t p1 ≤ t < tr 1 , α̇1 = α̇2 = , β̇ = 0 ,
tr 1 − t p1 r
Another random document with
no related content on Scribd:
new leader, whose heart was not so easily moulded as that of
Montezuma. Respect for the emperor, the Spaniards replied, had
made them hitherto lenient toward his people, but if they remained
obstinate no further mercy would be shown, and not one Mexican
would be spared. “Two days hence not one Spaniard will be alive!”
was the retort.
Hostilities were thereupon resumed, and Cortés did not delay the
prearranged attempt to complete the capture of the approach to
Tlacopan. The presence of the imperial corpse had either a retarding
influence on the movements of the enemy, or else the Spaniards
sallied unexpectedly and fought with greater energy, for the four
remaining bridges were gained with little difficulty, and twenty
horsemen passed on to the shore, while the infantry and allies took
possession of the route, and began filling in the channels with débris,
so as to form a solid path, or to repair the bridges at the deeper
places. At this time a messenger arrived with the announcement
from the chiefs who were directing the siege of the fort that they
were willing to treat for peace. Leaving the forces in charge of
Velazquez, Cortés hurried with some horsemen to answer the
welcome summons. The chiefs proposed that if pardon was granted
them for past offences they would raise the siege, repair the bridges
and causeways, and return to peaceful intercourse. In order to
arrange the conditions they demanded the liberation of the captured
high-priest. This was at once agreed to, and after some discussion
messengers were despatched to different parts of the city, bearing
orders, it was said, to stay hostilities.
It is somewhat singular that the astute Cortés should have given
such ready credence to proposals so advantageous to himself. Yet
this appears to have been the case. Delighted with the happy
adjustment of affairs, he ordered prepared a grand supper; but he
had hardly seated himself at table before tidings reached him that
the Mexicans had returned to the attack on the causeway, largely
reinforced by land and water, and were regaining the bridges taken
that day. The conference had been a ruse to throw the Spaniards off
their guard, to obtain the release of the high-priest, from whom
besides much information was expected about the condition of the
besieged, and to gain time for bringing up reinforcements.[813]
Fearful that his retreat would yet be cut off, Cortés galloped back to
the causeway, threw himself on the enemy, recovered the bridges,
and was soon in hot pursuit of the flying Mexicans. He had not
proceeded far, however, when the Indians, who had rushed for
safety into the lake and the canals, were encouraged to return to the
attack and cut off the cavalry. With furious charges they drove the
guard from the bridges, and began to destroy them and remove the
filling.[814] The causeway swarmed again with foes, and the water
round it was alive with canoes, whence myriads of missiles were
directed against the horsemen as they pushed their way back. On
reaching the last causeway breach, nearest the city, the riders feared
they would be overwhelmed, for here the enemy was gathered in
masses and had destroyed the passage. Nothing was left for them
but to take to the water, midst a storm of stones and darts, while
lines of spears and javelins pressed against them from the land and
from canoes. The party was thrown in disorder, and one rider was
pitched from his saddle during the mêlée, obstructing the passage to
the rest.[815] Cortés remained the last to cover the retreat, and
single-handed now and then turned on the swarming warriors,
striking with the energy of despair. Eager to secure the great
general, the enemy pressed heavily upon him, and but for the stout
armor protecting himself and the horse he would certainly have
perished. As it was, he received two severe wounds in the knee,
besides many scratches. The last Spaniard having left the bank,
Cortés rang loud his San Pedro cry, and clearing the way he leaped
his heavily laden horse across the chasm, six feet in width, and
quickly left behind him the discomfited crowd. “Had not God helped
me,” he writes, “that moment would have been my last.” Indeed, it
was already rumored in the city that he was dead. It being found
impossible to hold the causeway bridges, a guard was left only at the
others, while the remainder of the troops returned to the fort, worn-
out and demoralized.[816]
Long since it had been agreed among the Spaniards that the city
must be evacuated; time and method were the only questions. The
former of these was now resolved on by the council: it should be this
very night. It was safer to meet the issue now than later. The enemy
was hourly reinforced. Perilous indeed was the undertaking to pass
with luggage, war stores, prisoners, and women over the broken
causeway in the darkness; but to remain was death. Botello, the
astrologer, had declared for this time, and so it was determined. For
Botello was wise and prudent, knowing Latin and the stars; he had
foretold the greatness of Cortés, and had recommended his night
attack on Narvaez, and general and soldiers believed in him.[817]
Had he lived a century or two later his words might have been
employed as the vox stellarum by the almanac makers. The
Mexicans had said that they would make it a time of sore distress,
any attempted escape of the intruders, a time when men must
struggle, and women would pray and weep; and if so, it were no
worse for the fugitives that black night should fling her mantle over
the bloody scene.
Since the Indians were supposed to have destroyed the crossing
at the causeway channels, a portable bridge was made with which to
effect the passage. Two more would probably have been made had
time and convenience permitted, but misfortune willed it otherwise. It
was agreed that a large portion of the effects must be left behind in
order not to encumber the march, but the gold demanded special
care. The royal officials, Mejía and Ávila, were charged to secure it,
and for this purpose a number of carriers were assigned, the general
giving also one of his own mares. Their convoy was intrusted to a
body of infantry, under Alonso de Escobar.[818] The secretary,
Hernandez, and the royal notaries were called to testify that all had
been done that was possible. There still remained a large quantity of
the bulky jewels belonging to the king, besides a mass of
unappropriated treasure, which could not be intrusted to carriers, or
for which no carriers were found, and rather than leave them to the
‘Indian dogs’ Cortés announced that the soldiers might take all they
wished—after permitting his favorites the first selection. He warned
them, however, that the more they took the more their safety would
be endangered. The adherents of Cortés do not appear to have
been eager to encumber themselves, and Bernal Diaz shared this
prudence in taking only four chalchiuite stones. The men of Narvaez
practised less restraint, and many loaded themselves with the metal.
Cortés was afterward charged with having appropriated a
considerable share of the wealth thus thrown open; he certainly had
funds with which to send for horses, war material, and supplies.[819]
Sandoval was appointed to lead the van, with two hundred
infantry and twenty horsemen, assisted by Ordaz, Andrés de Tapia,
and others. With him went fifty men under Captain Magarino to carry
the bridge. They were pledged to remain at their post to the last, and
were escorted by a select body of infantry and allies. For the middle
were destined the baggage and treasure, the prisoners and the sick,
under a large escort, supervised by Cortés himself, who, with Olid,
Morla, Ávila, and other captains, and a special force of one hundred
men, were to render aid where needed. The artillery was intrusted to
two hundred and fifty Tlascaltecs and fifty soldiers, and the rear was
placed in charge of Alvarado and Velazquez, with thirty horsemen
and about one hundred adherents of Cortés, with most of the men of
Narvaez. The allied forces, of whom a number appear to have
returned home during the inaction of Montezuma’s captivity, and who
had suffered greatly during the siege, must still have numbered
nearly six thousand men, including carriers, distributed among the
three divisions.[820] Among the prisoners Cortés enumerates the
legitimate son of Montezuma, and two of his daughters, probably
those bestowed on the general in marriage, King Cacama and his
younger brother and successor, and several other high personages.
[821] The sick were to be carried in hammocks and behind riders.
It is the evening of the 30th of June.[822] Fiery copper has been

the sky that day; the sun blood-red and moon-like, turning day to
night, when night is so soon to be employed as day. As the hour
approaches, a fog sets in, which thickens into mist and denser
moisture until, to favor the Spaniards, providence turns it to a
drizzling rain,[823] thus to veil their movements, and make substantial
the silence of the city, the lake, the distant wood; and thereupon all
join fervently in the prayer of Father Olmedo and commend their
lives to almighty God.
About midnight the order is given to march.[824] Stealthily they

creep down the temple square and reach the Tlacopan road.[825]
The streets are wholly deserted. All is quiet, save the dull rumble of
tramping soldiers. The blessed rain, or some supernatural
interposition, seems to keep the whole city within doors. And if this
kind power will but have patience and not desert them for one brief
hour—ah, it is so easy for Omnipotence to help! Along the road like a
phantom the army moves. The van picks up the guard at the canal
crossings. The causeway is almost reached. Already they begin to
breathe freer; a feeling of intense relief steals into their breasts, and
—Mother of God! what noise is that? It is the piercing outcry of a
woman[826]—may the foul fiend seize her!—breaking upon the
stillness like a warning note from the watch-tower of Avernus. On the
instant the war-drum of the Tlatelulco temple sends forth its doleful
sound, chilling the fugitives to the very heart’s core. Quickly its tones
are drowned by the nearer, shriller trumpet-blasts and shouts of
warriors, echoed and reëchoed from every quarter.
Meanwhile the advance column had reached the sixth bridge
crossing,[827] the first to connect with the causeway, and had obliged
the Mexican picket to retire, after exchanging a few shots. The
portable bridge was here laid, and the van crossed with quickened
steps, followed by the centre with the baggage and artillery. At this
juncture the enemy fell upon the rear, rending the air with their yells,
sending their missiles fast and furious, while from the cross-roads
issued a swarm, with lance and sword, on Alvarado’s flank. Over the
water resounded their cries, and canoes came crowding round the
causeway to attack the forward ranks. To add to the horrors of the
tumult, several men and horses slipped on the wet bridge and fell
into the water; others, midst heart-rending cries, were crowded over
the edge by those behind. All the rest succeeded in crossing,
however, except about one hundred soldiers. These, it is said,
bewildered by the battle cries and death shrieks, turned back to the
fort, and there held out for three days, till hunger forced them to
surrender and meet the fate of sacrificial victims at the coronation
feast of Cuitlahuatzin.[828]
The half mile of causeway extending between the first and
second breaches was now completely filled with Spaniards and
allies, whose flanks were harassed by the forces brought forward in
canoes on either side. Dark and foggy as the night was, the outline
of the Indian crews could be distinguished by the white and colored
tilmatli in which many of them were clad, owing to the coldness of
the air. Fearlessly they jumped to the banks, and fought the
Spaniards with lance and javelin, retreating into the water the
moment the charge was over. Some crept up the road sides, and
seizing the legs of the fugitives endeavored to drag them into the
water. So crowded were the soldiers that they could scarcely defend
themselves; aggressive movements were out of the question.
Repeated orders had been transmitted to Magarino to hurry
forward the removal of his bridge to the second channel, and, seeing
no more soldiers on the opposite bank of the first opening, he
prepared to obey, but the structure had been so deeply imbedded in
the banks from the heavy traffic that his men labored for some time
in vain to lift it, exposed all the while to a fierce onslaught. Finally,
after a number of the devoted band had succumbed, the bridge was
released, but before it could be drawn over the causeway the enemy
had borne it down at the other end so as effectually to wreck it.[829]
The loss of the bridge was a great calamity, and was so regarded by
the troops, hemmed in as they were between two deep channels, on
a causeway which in width would hold only twenty men in a line. On
all sides were enemies thirsting for blood. Presently a rush was
made for the second channel, where the soldiers had already begun,
in face of the foe, to cross on a single beam, which had been left
intact when the bridge was destroyed. As this was an exceedingly
slow process, many took to the water, only to receive their death-
blow at the hands of the watermen. Some were taken prisoners;
some sank beneath their burden of gold; the horses found a ford on
one side where the water was not above the saddle.
The canoes, however, were as numerous here as elsewhere,
and their occupants as determined; and the horsemen had the
greatest trouble to keep their seats while resisting them. The
general, being at the head, suffered most. At one time some Indians
seized him by the legs and tried to drag him off. The footing of the
horse being so insecure, the attempt would probably have
succeeded but for the prompt aid of Antonio de Quiñones, and
Texmaxahuitzin, a Tlascaltec, known afterward as Antonio. Olid, who
also came to the rescue, was almost overpowered, but managed to
free himself by means of backhanded blows from his muscular arm.
One of the cavalry, Juan de Salazar, the page of Cortés, then took
the lead to clear the way for the rest, only to fall a victim to his zeal.
The next moment his master had gained the bank, and thereupon
directed the troops by the ford.[830]
Thus in the darkness the wild roar of battle continued, the
commingling shouts and strokes of combatants falling on the distant
ear as one continuous moan. The canoes now pressed on the
fugitives in greater number at the ford than in the channel. Sandoval,
with his party, had swum the channel before the Mexicans
assembled there in great numbers, and was now leading the van
down the causeway, scattering the assailants right and left. Little
regular fighting was attempted, the Spaniards being intent on
escaping and the Mexicans quickly yielding before the cavalry, taking
refuge in and round the canoes. With greater hardihood and
success, however, they harassed those on foot. On reaching the
next channel, which was the last, the fugitives found with dismay that
it was wider and deeper than the others, and with bitter regret they
saw their mistake in not bringing three portable bridges. The enemy
was here also gathering in ever increasing force, to watch the death
trap. Every effort to clear a passage was stubbornly resisted, and,
the soldiers growing more irresolute, a rider was sent to bring
Cortés. Before he arrived, however, Sandoval had already plunged
in with a number of the cavalry, followed by foot-soldiers, who seized
the opportunity to fall into the wake, by either holding on to the
trappings of the horses or striking out for themselves. The passage
was extremely difficult, and more than one horseman reeled and fell,
from the united pressure of friends and foes. Those who followed
suffered yet more, being pushed down by comrades, struck by clubs
and stones, pierced by spears, or, most horrible of all, drawn in by
dusky boatmen, who carefully guarded them for the dread stone of
sacrifice.
With five horsemen Cortés led a body of one hundred infantry to
the mainland. Accompanying this force was a number of carriers with
treasures secured by the general and his friends. Leaving the gold in
charge of Jaramillo, with orders to hold the entrance of the causeway
against assailants from the shore, Cortés returned to the channel
where Sandoval had taken a stand to keep clear the bank and
protect the passage. Tidings coming that Alvarado was in danger,
Cortés proceeded to the rear, beyond the second channel, and found
it hotly contested. His opportune arrival infused fresh courage, as
with gallant charges he relieved the troops from the terrible pressure.
He looked in vain for many comrades who had been placed at this
post, and would have gone in search of them had not Alvarado
assured him that all the living were there. He was told that the guns
reserved for the rear had for a while been directed with sweeping
effect against the ever growing masses of warriors around them; but
finally a simultaneous attack from the canoe crews on either side,
and from the land forces to the rear, impelled by their own volume,
had overwhelmed the narrow columns nearest the city, together with
their cannon, killing and capturing a large number, and throwing the
rest into the panic-stricken condition from which he had just
extricated them.
Leaving Alvarado to cover the rear as best he could, Cortés
hastened to direct the passage of the middle channel. What a sight
was there! Of all the bloody terrors of that dark, sorrowful night, this
was the most terrible! A bridge had been wanting, and behold, the
bridge was there! With dead and living fugitives the chasm on either
side the slippery beam had been filled,[831] and now the soldiers and
allies were rushing, heedless of the groans beneath them, across
this gory support, still narrow and full of gaps, to be filled by the next
tripping fugitive. Scattered pell-mell on the bank lay the baggage and
artillery, abandoned by the fleeing carriers, which, proving only an
obstruction, Cortés ordered it thrown into the channel in order to
widen the crossing.
But the end was not yet. Great as had been the woe, it was yet
to be increased at the last and wider channel. Here was indeed a
yawning abyss, having likewise a single remaining beam, whose
narrow slippery surface served rather as a snare than a support.[832]
The necessarily slow motion of the train had enabled the Mexicans
to come up in swarms, and like sharks surround the chasm.
Harassed on every side, and with an avalanche rolling against the
rear, the retreating thought only of escaping the new danger, and at
once. They threw aside their arms and treasures and plunged in,
bearing one another down regardless of any claims of friendship or
humanity. And woful to hear were the heart-rending cries from that
pit of Acheron. Some begged help of Mary and Santiago; some
cursed their fate and him who had brought them to it, while many
sank with mute despair into the arms of death; and over all roared
the wild cries and insults of the Mexicans. In strong contrast to the
panic-stricken men appeared a woman, María de Estrada, who, with
shield and sword, faced the enemy like a lioness, standing forth
among the men as a leader, and astonishing friend and foe with her
prowess.[833]
Cortés did all he could, as became an able commander and
valiant soldier, to save his men. He was indefatigable in his efforts,
being everywhere present, encouraging, guiding, and protecting. Yet
his position was most trying; there were that night so many brave
soldiers given over to despair, so many ears deaf to commands and
prudent counsel. Unable to do more at the channels, he hastened to
look to those who had crossed and were proceeding in straggling
bands to join Jaramillo. Heedless of companies or officers, the
soldiers had banded in parties of a score or two, and sword in hand,
where this had not been thrown away, they were hurrying down the
causeway.[834] The assailants fell off somewhat beyond the last
channel, and finding the advance comparatively safe, guided by his
soldierly impulses Cortés again returned with a few horsemen[835]
and foot-soldiers to cover the remnant of the army. The rear,
composed chiefly of the Narvaez party, were approaching the last
channel, but under the continued onslaught panic had seized them.
They made hardly an effort to defend themselves, and like the
Indians during the massacre by Alvarado they huddled one against
the other, offering their backs as a target for unsparing attack.
Among this number was the loyal and noble Velazquez de Leon,
who shared with the Tonatiuh the command of this section. How he
fell is not known, but he never crossed the last breach.[836]
Alvarado had been wounded and had lost his horse, in common
with most of his party. Finding it impossible to control the men, he
gathered a small band round him and sought the channel, leaving
the rest to look to themselves.[837] On reaching the spot he saw a
confused mass of struggling humanity in the water, but the solitary
beam which spanned it was vacant, and steadying himself with his
lance he sprang swiftly across. Narrow and slippery as was the
beam, it was no insignificant feat for a wounded man to cross upon
it, but time magnified the performance to something miraculous.
When Alvarado came to the channel, it is related, no friendly beam
spanned the wide, deep gap. His life turned on brief resolve and
instant action. Lithe, strong, and determined, even though wounded,
he was not yet ready to yield all. With a searching glance into the
troubled pool and across the awful chasm he stepped back for a
preparatory spring. Then, rushing forward, he planted the long pike
upon the yielding débris and vaulted across, to the wonder of all
witnesses. The Indians, says Camargo, prostrated themselves in
admiration, and tearing up grass, ate it, with the exclamation, “Truly,
this man is the Tonatiuh!” So runs the story, preserved by tradition,
and by the name yet given to the spot, ‘El Salto de Alvarado.’[838]
Cortés and his small band of rescuers came up as Alvarado
appeared, pike in hand and bleeding, accompanied by a few
stragglers.[839] Among these was Juan Tirado, who, in gratitude for
his deliverance, erected at this bridge after the conquest a hermitage
to San Acacio, known also as De los Mártires—martyrs to avarice,
as Torquemada intimates.[840] The badly wounded were now
mounted behind the horsemen,[841] and repelling the foes who still
pressed on them, Cortés in person covered the remnant of the army
in its retreat toward Tlacopan,[842] losing in this final struggle the
gallant Captain Morla.[843] The route lay through Popotla village or
suburb; and here, according to tradition, Cortés seated himself on a
stone to weep over the misfortunes of this Sorrowful Night.[844]
By a similar process of annealing, gold is made soft and iron
hard; so by misfortune the wise man is made wiser while the fool is
hardened in his folly.
FOOTNOTES
[808] According to the version of the rabid Duran, based on native paintings and
narratives, the bodies of the prisoners were found in the fort after its evacuation,
that of Montezuma with five stabs in the breast. Hist. Ind., MS., ii. 477-9. Acosta
seems rather to favor the story, improbable as several of its points are. Hist. Ind.,
524. To some extent it rests on the statement repeated by Ixtlilxochitl, which
assumes that Cacama, who had made himself particularly obnoxious to the
Spaniards, was killed with 47 stabs before the fort was evacuated. Hist. Chich.,
301. A more severe account is found in a manuscript fragment in Ramirez’
collection, written by a Tezcucan, wherein it is related that a sword was thrust into
the intestines, ‘por la parte baxa.’ The body was thereupon taken to the roof, as if
to address the people. A stone struck the head, and now the Spaniards
proclaimed that this had caused the death. Soc. Mex. Geog., Boletin, x. 362. This
is substantially repeated in Ixtlilxochitl, Rel., 457. A stronger testimony, however,
comes from Sahagun, who states that Cortés recommended to his followers the
murder of the prisoners in order to terrify the natives and to assume the mastery.
‘Y lo primero que hicieron, fue dàr Garrote à Motecuhçuma, y à Itzquauhtzin,
Señor de Tlatelolco, y à otros.’ Version in Torquemada, i. 498, and in Sahagun,
Hist. Conq. [ed. 1840], 113. The issue, modified by the censor, merely states that
the bodies were found near a stone, Teoaioc, outside the fort. Id. [ed. 1829], 31.
This account has received its chief support in the quasi admission of Torquemada;
and when he, the otherwise zealous champion of the conquerors, takes such a
view, others may be pardoned for accepting it. ‘Y que esto aia sido asi, puede ser
posible, pues para tenerse por seguros, le avian prendido; y viendo agora, que no
bastaba la prision, vsarian de este vltimo medio, para vèr si le aplacaban, y
atemoriçaban estos Mexicanos.’ i. 498-9. Brasseur de Bourbourg accepts the
version, and adds that Montezuma was told of his fate and urged to accept
baptism. ‘On répandit aussitôt la nouvelle de sa mort comme s’il eût expiré
naturellement.’ Hence even the soldiers did not know of the murder. Hist. Nat. Civ.,
iv. 330-1. Bustamante, of course, adopts anything Sahagun may say against the
Spaniards; and Carbajal takes Torquemada’s view, Hist. Mex., ii. 375, as does
Beltrami, eager for any sensation. Mex., ii. 145. Vetancurt seeks to reconcile
conflicting opinions by assuming that when Montezuma died Itzquauhtzin and
several other prisoners were murdered and cast out together with the emperor’s
body, in order to terrify the Mexicans and occupy their attention while the
Spaniards hurried away. Teatro Mex., pt. iii. 142. ‘Io non posso persuadermi,’ says
Clavigero, ‘che gli Spagnuoli si risolvessero a toglier la vita ad un Re, a cui
doveano tanti beni, e dalla cui morte non potevano aspettarsi, se non molti mali.’
Storia Mess., iii. 131. Solis cannot believe Cortés guilty of an act so bad and
reckless. Hist. Mex., ii. 150-1. Prescott disposes of the charge as an ‘absurdity’
and a ‘monstrous imputation.’ Hist. Mex., ii. 321. It must be considered, however,
that the Spaniards did seek to profit by the death; and scruples about ‘killing a dog
of an Indian,’ as they had so lately termed him, could not have weighed with such
men when their interests were concerned; how much less when their lives were at
stake? The whole argument, then, may be said to depend on the question whether
Montezuma was more valuable as captive or as corpse. If the people manifested
little respect for the living ruler, the Spaniards could have had no reason to expect
more for the dead. His death would only have loosened the bond which still
restrained a vast number, whether of kindred or of mere subjects, and given the
hostile leaders fresh motives and strength for their operations. Besides,
Montezuma must have retained a great influence outside the city, which a fugitive
army would have found of service. The recognition of this influence is shown by
the efforts made to save the imperial children, as noticed even by the most rabid
accusers of the Spaniards. It may be mentioned that no charge is brought forward
in the residencias either against Cortés or Alvarado.
[809] Apanecatl, according to Brasseur de Bourbourg, Hist. Nat. Civ., iv. 332.
Bernal Diaz sends these men to carry the news of the death, and following them
were six high personages and most of the captive priests, carrying the body. Hist.
Verdad., 105.
[810] Stones were thrown upon the cortege, and it was driven from quarter to
quarter. Finally Apanecatl took refuge in the palace where Cuitlahuatzin held forth,
and appealed to him, only to be repulsed by his courtiers. The body was
nevertheless secured by a friendly party. Manuscrit Nahuatl, 1576, in Brasseur de
Bourbourg, Hist. Nat. Civ., iv. 333.
[811] ‘Hicieron todas las solemnidades que solian hacer ... Mocthecuzoma lo
enterraron en México ... algunos decian mal de Mocthecuzoma porque habia sido
muy cruel.’ Hist. Conq., 31. ‘Vimos q̄ hizierõ muy gran llanto, q̄ biẽ oimos las
gritas, y aullidos q̄ por èl dauan.’ Bernal Diaz, Hist. Verdad., 105. ‘Hizieron muy
gran llanto, para enterrar al rey en Chapultepec.’ Gomara, Hist. Mex., 154.
Herrera combines these two authorities in saying ‘le deuieron de enterrar en el
monte de Chapultepèque, porque alli se oyò vn gran llanto.’ dec. ii. lib. x. cap. x.
He forgets that Chapultepec lay three miles off. Torquemada corrects Herrera, and
insists that the ‘Copalco’ was the place. He gives specimens of the insults offered
during the cremation, i. 499. ‘Estaban indignados contra él.’ Ixtlilxochitl, Hist.
Chich., 301. It has been asserted by some, says Duran, that the ashes were
scattered to the winds, as unworthy of preservation. Hist. Ind., MS., ii. 479. Acosta
attempts a modification by stating that the body being contemptuously rejected, a
servant burned it, ‘y puso sus cenizas dõde pudo en lugar harto desechado.’ Hist.
Ind., 524. The burial-place has certainly not been pointed out to posterity.
According to Sahagun, the body of Itzquauhtzin was ‘cast forth’ from the quarters,
together with that of Montezuma, and was taken in charge by his subjects of
Tlatelulco, by whom he was greatly beloved and mourned. ubi sup. To ‘cast forth’
the bodies could have been only a needless insult, which Cortés was too prudent
to inflict on the people.
[812] ‘Que alçassen a su primo del Monteçuma, que con nosotros estaua, por
Rey.’ Bernal Diaz, Hist. Verdad., 105. ‘Dixo Cortes ... el se queria hallar a sus
honras.’ Herrera, dec. ii. lib. x. cap. x.
[813] Prescott, following Clavigero, assumes that the whole was an artifice to
liberate the two captive priests, one of whom was indispensable in the event of a
coronation. Mex., ii. 338. Brasseur de Bourbourg supposes that the parley was
conducted by the party favorable to the Spaniards, and duped by the stronger
faction, which never intended to adhere to the arrangement. Hist. Nat. Civ., 321.
[814] ‘Como los peones estaban cansados y heridos y atemorizados; ... ninguno
me siguió. A cuya causa, después de pasadas yo las puentes ... las hallé
tomadas.’ Cortés, Cartas, 134. Where had he left his prudence?
[815] ‘Hallé á todos los de caballo que conmigo iban, caidos en ella, y un caballo
suelto.’ Id.
[816] With a loss of over twenty men. Bernal Diaz, Hist. Verdad., 105. This author
places all the fighting on the causeway on one day, a Thursday, the day of
evacuating Mexico, and the day following the surrender of Montezuma’s body.
Herrera, who is far more confused, has a sally on this day in three directions, one
being the Tlacopan road; but the operations on the latter route are only partially
told, and the rest referred to the third day of the siege. There are also several
contradictions to aid in confusing the many who follow him. Herrera, dec. ii. lib. x.
cap. xi. Prescott abandons himself to the guidance of Clavigero for the
occurrences of these days, but embellishes the narrative with some incidents
belonging to the siege of Alvarado.
[817] ‘Botello ... afirmò que ... supiessen que moriria el o su hermano, y algunos
de la cõpañia, y qui se saluaria el Capitan, y otros muchos, y ninguno si salian de
dia.’ Herrera, dec. ii. lib. x. cap. xi. ‘Hora lo creyesen, hora no.’ Gomara, Hist.
Mex., 159. ‘Anteponendo le vane osservazioni di quel meschino Soldato alla luce
della prudenza militare,’ is the indignant comment of Clavigero, Storia Mess., iii.
135. But there is no doubt that ‘military prudence’ had more weight in the matter
than Botello’s words, and that the result was not due to his advice. Solis casts the
blame of crediting the ‘ignorant charlatan’ Botello chiefly on the majority of the
council, to whom Cortés yielded. Hist. Mex., ii. 171-2. In order to lull any
suspicions among the Mexicans, says Bernal Diaz, a leading priest and some
other captives were sent to the Mexican camp with a proposal to surrender all the
gold if the Spaniards were allowed to leave in peace eight days later. Hist.
Verdad., 105.
[818] Lejalde, Segunda Prob., in Icazbalceta, Col. Doc., ii. 424. For carrying the
royal treasures ‘les dió siete cauallos heridos, y cojos, y vna yegua, y muchos
Indios Tlascaltecas, que segun dixeron, fueron mas de ochenta.’ Bernal Diaz,
Hist. Verdad., 106.
[819] Bernal Diaz, for instance, states that he had hardly taken the four
chalchiuites for his share, from a collection in a mat, when Cortés gave orders to
his mayordomo to secure it. loc. cit. His penchant for appropriating large shares to
himself is well known. Greed of gold was not now his motive, however, but rather a
prudential care to secure means for his plans, and he could hardly neglect them
when taking so great care of the royal portion. Martyr, Gomara, and Herrera
estimate the treasure at 700,000 ducats, chiefly in bulky jewels according to
Gomara. Bernal Diaz reckons in pesos, which may mean pesos de oro. Peter
Martyr assumes it to have been the general fund, from which the royal fifth had
been set apart only at the last moment, but not apportioned. dec. v. cap. vi. Solis
assumes that 700,000 pesos remained after the king’s portion had been deducted.
Hist. Mex., 174-5. One witness estimates that over 2,000,000 pesos were lost
during that night. Cortés, Residencia, ii. 414. The Carta del Ejército reduces the
loss to 400,000 pesos de oro. Another witness states that 300,000 castellanos
remained when the soldiers were told to help themselves; afterward the general
compelled them to surrender what had been thus given, only to keep it for himself.
Cortés, Residencia, i. 241-2. ‘Lo demás ... lo dimos y repartimos por los españoles
para que lo sacasen,’ says Cortés, Cartas, 135, which may be interpreted as
either giving or intrusting. Whatever may have been left after the Spaniards had
taken their loads was gleaned by the allies. Gomara, Hist. Mex., 159.
[820] Bernal Diaz gives Sandoval 100 young unmarried soldiers, with Francisco de
Acevedo, the dandy, Ordaz, Tapia, and eight or nine of Narvaez’ men, captains on
his staff. To Cortés he gives 50 men, and adds B. V. de Tapia to his staff. Hist.
Verdad., 105. This author is contradictory, however. Herrera places Antonio de
Quiñones as Sandoval’s chief aid, and Olid and Ordaz in the rear. dec. ii. lib. x.
cap. xi.; Gomara, Hist. Mex., 160; Cortés, Cartas, 134; Ramirez, Proceso contra
Alvarado, 30 et seq.
[821] Herrera adds a brother of Montezuma, and Sahagun names two sons. Hist.
Conq., 33. So does Vetancurt, although he assumes that one was saved. Teatro
Mex., pt. iii. 142-3. Ixtlilxochitl gives a longer list, including two sons of
Montezuma, and two sons and four daughters of Nezahualpilli, of Tezcuco. One of
the daughters escaped, but it was not the beloved of Cortés, who had been
baptized and named Juana. Cacama is not included in the list, because he is
assumed to have been stabbed to death before the fort was evacuated. Hist.
Chich., 302; Relaciones, 390. With the prisoners’ division went Marina, the
interpreter, the Tlascaltec princesses Luisa and Elvira, and some other women,
protected, says Bernal Diaz, by 30 soldiers and 300 Tlascaltecs.
[822] This date is based on Cortés’ letter, wherein he places the arrival on
Tlascala’s border on Sunday, July 8th, after giving a clear account of the
intermediate days. Any doubt about this date is removed by the testimony in
Lejalde, Segunda Probanza, in Icazbalceta, Col. Doc., i. 423, wherein the leading
captains state that the siege lasted six days. This testimony also clears up the only
doubtful point in Cortés’ account of the siege operations, where he disposes of the
wounding and death of Montezuma in one sentence, and then resumes the
description of the fighting in a manner that has assisted to mislead Gomara and
many others into extending the stay in Mexico till July 10th. Ixtlilxochitl adopts this
date, yet in the Relaciones, 390, 412-13, he states that the siege lasted only
seven days. Bernal Diaz places the eve of the departure on a Thursday, July 10th
[with Cortés it is Saturday], yet he dates the battle of Otumba just one week later
than Cortés. Hist. Verdad., 105, 108. This latter date induces Zamacois to change
the date of flight to July 8th. Hist. Méj., iii. 406-7. ‘La notte del 1 Luglio,’ says
Clavigero, Storia Mess., iii. 135, but his reasons for the date are wrong, and the
term he uses may apply also to the night following that adopted in the text.
[823] The Spaniards recognized this as a favoring shield direct from God, says
Duran, Hist. Ind., MS., ii. 473-4.
[824] Ojeda was instructed to see that no somnolent or sick person was left. He
found one man asleep on the roof and roused him. Herrera, dec. ii. lib. x. cap. xi.
Gomara takes the trouble to deny the statement of Cano that 270 men, ignorant of
Cortés’ departure, were left behind to perish. Oviedo, iii. 551. A later note will
explain the cause of this rumor.
[825] The ravaging sallies of the preceding days, which had involved the
destruction of houses in the vicinity and along the approaches to Tlacopan, had
evidently obliged the enemy to retire from these streets and seek shelter
elsewhere for the night. Oviedo assumes that Cortés led the way, but Diaz and
Herrera let him advance only when the first troops are guided into Tlacopan.
[826] ‘Pasaron cuatro acequias, y antes que pasasen las demas salió vna muger
á tomar agua y viólos.’ Sahagun, Hist. Conq., 32. ‘A devil, without doubt,’ adds
Camargo, who describes her as a keeper of an eating-house, and indicates
exactly the location of her house. Hist. Tlax., 167. Cortés states that the alarm was
given by the guard at the first breach held by the Mexicans, where the portable
bridge was laid down. Cartas, 136.
[827] Sahagun names this Mictlantonco, and the next two Tlantecayocan and
Petlacalco. In another place he names the first Tecpantzinco, and the second or
third Tolteacali. Hist. Conq., loc. cit., and [ed. 1840] 121-2. Torquemada gives the
second breach of the causeway the latter name. The names should probably be
written Tecpantzinco, Tolteca-Acalulco, and Petlacalco.
[828] This native rumor, as recorded in the manuscripts used by Duran, Hist. Ind.,
MS., ii. 476-7, is probably the foundation for Cano’s statement, that Cortés
abandoned 270 men in the fort. Herrera reduces them to 100. ‘Que se boluieron a
la torre del templo, adonde se hizieron fuertes tres dias.’ dec. ii. lib. x. cap. xii.
[829] Bernal Diaz, Hist. Verdad., 106, assumes that the enemy bore it down before
the baggage train had crossed, and that the channel was filled in consequence
with artillery, baggage, and dead bodies. Gomara gets the bridge across the
second breach. Both must be mistaken, however.
[830] Camargo relates the incidents of the passage in detail, and says that Cortés
fell into a hole as the enemy pounced upon him. The two deliverers disputed the
honor of having rescued the general. Hist. Tlax., 169.
[831] ‘El foso se hinchó hasta arriba; ... y los de la retroguardia pasaron sobre los
muertos. Los españoles que aquí quedaron muertos fueron trescientos, y de los
tlaxcaltecas y otros indios amigos fueron mas de dos mil.’ Sahagun, Hist. Conq.
(ed. 1840), 122.
[832] Ramirez, Proceso contra Alvarado, 4, 53, 68.

[833] ‘Casò esta Señora, con Pedro Sanchez Farfan [who seized Narvaez], y
dieronle en Encomienda el Pueblo de Tetela.’ She married a second time, and
died in Puebla. Torquemada, i. 504.
[834] Bernal Diaz formed one of a band of 50, who were repeatedly attacked with
arms and midst insults. He quotes some of the low expressions used. Hist.
Verdad., 106.
[835] One authority states that Cortés was nearing Tlacopan, when Olid and
others called out to him that the fugitives were accusing the captains of
abandoning them, and urged that they should turn back. ‘It is a miracle to have
escaped,’ was his reply, ‘and fewer will be left if we return.’ Saying this he headed
a dozen horsemen and a few foot-soldiers and galloped back. Bernal Diaz, Hist.
Verdad., 106. But Cortés was not the man to wait in such a case till entreaty came.
‘Yo con tres ó cuatro de caballo,’ he says, ‘y hasta veinte peones, que osaron
quedar conmigo, me fuí en la rezaga.’ Cartas, 135. He takes the palm from all
American conquerors, exclaims Oviedo, iii. 326.
[836] Zamacois makes atonement for a lack of research by inventing doughty

deeds for this hero. Hist. Méj., iii. 417-18.
[837] Among the soldiers contributed in later times by Garay’s expedition was one
Ocampo, who, fond of scandal and pasquinades, libelled many of the captains,
among them Alvarado, declaring that he had left Velazquez with over 200 men to
die. Bernal Diaz, Hist. Verdad., 106-7. The charge came forward in the residencia,
but Alvarado brought witnesses to prove that he had lost all control over the men,
and could do nothing else than to save himself, wounded and unhorsed as he
was. There were other witnesses who did all they could to blacken his fame, and
to attribute to his neglect of duty a great portion of the loss sustained during that
sad night. Ramirez, Proceso, 4, 38, 53, 68, and 288. Ramirez decides against the
accused. But Alvarado was admittedly brave, recklessly so, and it must be
regarded rather as his misfortune that a panic seized the men. Perhaps, as
commander intrusted with this section, he should have remained longer at his
post. This signified death, and such men as then comprised his command he
regarded as hardly worth dying for. He chose to save life at the expense of a
blemish on his honor. More it never amounted to, for the court absolved him. He
redeemed the fault afterward by brave achievements.
[838] Camargo intimates that several Tlascalan chiefs of the expedition testified to
the feat. Hist. Tlax., 168; and Gomara adds that several followers tried to imitate it,
but failed, and were drowned. Hist. Mex., 160. Contradictory as Bernal Diaz is
about the incidents of the night, he strenuously insists that the channel was
examined during the following siege and found to be too wide and too deep to
allow of such a leap. Hist. Verdad., 107. This solitary denial of a story which has
been adopted by almost every writer, from Oviedo to Prescott, finds support in
testimony during the hero’s residencia, wherein it is distinctly stated that he
crossed the channel on a fixed beam. His own testimony gives assent to the
charge so formulated, although hitherto he had no doubt allowed the other version
to be believed. Ramirez, Proceso, 4, 53, 68 et seq.
[839] Seven Spaniards and eight Tlascaltecs, all badly wounded. Bernal Diaz,
Hist. Verdad., 106.
[840] ‘Durò poco este nombre, pues tampoco les convenia à los muertos, que iban
cargados de Oro.’ Monarq. Ind., i. 504. Zamacois describes the site as he found it
not long ago. Hist. Méj., iii. 421-3. Bernal Diaz implies that the ‘martyr’ name was
given in honor of those captured and sacrificed during the siege, a year later. Hist.
Verdad., 153.
[841] Alvarado was taken on Gamboa’s horse, Laso on Sandoval’s. Ramirez,

Proceso contra Alvarado, 69, 119.
[842] Alvarado said that only the dead remained behind, but Olid insisted that a
number were still fighting the enemy. Cortés accordingly went back again and
rescued several more. Castañeda, in Id., 44. Bernal Diaz also states that Cortés
returned as far as the bridges. Hist. Verdad., 106.
[843] The same who sprang into the sea, off Yucatan, to replace the rudder of his
vessel, unshipped during the storm.
[844] La Noche Triste, as it has ever since been called. Amid so much that is
romantic the tendency to further romance is often uncontrollable. The truth of this
statement is open to grave doubts. ‘Llegó [Alvarado] á Cortés, que estaba ençima
de unas gradas de un qü, sentado diçiendo muchas lástimas.’ Oviedo, iii. 514.
Zamacois describes the enormous tree, yet standing, which shaded the stone and
bears the name of ‘Arbol de la Noche Triste.’ Hist. Méj., iii. 424. Prescott improves
the occasion by allowing the army to file past in sad dilapidation, regardless of the
fact that the army was already gathered in Tlacopan. Mex., ii. 371-2. Testimony
confirms the statement of Cortés that ‘fuí en la rezaga, peleando con los indios
hasta llegar á una ciudad que se dice Tacuba’ [Tlacopan]. Cartas, 135-6.
CHAPTER XXVII.
RETREAT TO TLASCALA.
July, 1520.
Fatal Mistake of the Mexicans—A Brief Respite Allowed the Spaniards—

The Remnant of the Army at Tlacopan—They Set out for Tlascala—An
ever increasing Force at their Heels—Rest at the Tepzolac Temple—
Cortés Reviews his Disasters—The March Continued amidst Great
Tribulation—Encounter of the Grand Army—Important Battle and
Remarkable Victory—Arrival at Tlascala—The Friendly Reception
Accorded them There.
What would Emperor Charles have said to Hernan Cortés had

they met on the morning after the Sorrowful Night! It is related of
Xerxes that with a golden crown he rewarded a pilot who had saved
his life, and thereupon ordered him beheaded for having sacrificed in
the operation the lives of so many of his Persian subjects. Now
Cortés had not saved the emperor’s life, nor yet the emperor’s gold;
he had sacrificed many lives, and had little to show for them. Had
Charles been there, and had he valued Spaniards as did Xerxes
Persians, he might have cut off the Estremaduran’s head; but Cortés
was yet worth to Charles more than all that had been thus far lost in
New Spain.
Prosperity implies ability; adversity, weakness of mind and
character. In the high-souled and chivalrous, prosperity tends to yet
loftier heights, while adversity sinks the unfortunate still lower;

Textbook Intelligent Optimal Adaptive Control For Mechatronic Systems 1St Edition Marcin Szuster Ebook All Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Textbook Intelligent Optimal Adaptive Control For Mechatronic Systems 1St Edition Marcin Szuster Ebook All Chapter PDF

Uploaded by

Copyright:

Available Formats

Intelligent Optimal Adaptive Control for

Mechatronic Systems 1st Edition

Adaptive Dynamic Programming with Applications in

Adaptive Critic Control with Robust Stabilization for

Intelligent Control of Robotic Systems 1st Edition

Reinforcement Learning for Optimal Feedback Control

Intelligent Algorithms for Analysis and Control of

Adaptive and Fault-Tolerant Control of Underactuated

Force and Position Control of Mechatronic Systems:

Optimal Control in Thermal Engineering 1st Edition

More information about this series at http://www.springer.com/series/13304

Intelligent Optimal Adaptive

ISSN 2198-4182 ISSN 2198-4190 (electronic)

© Springer International Publishing AG 2018

Printed on acid-free paper

This Springer imprint is published by Springer Nature

From Zenon Hendzel

5 Learning Methods for Intelligent Systems . . . . . . . . . . . . . . . . . . . . 85

7.5.1 Synthesis of DHP-Type Control . . . . . . . . . . . . . . . . 175

8.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286

10.7.1 Analysis of the WMR Motion Control . . . . . . . . . . . . 358

aforementioned controllers and additional algorithms for the generation of control

1.1 Artificial Intelligence and Neural Networks

field linked to automatic control, biocybernetics, electronics, applied mechanics and

1.2 Learning with a Critic

1.3 Scope of Study

The extension of this subject is the differential game theory, construed as an

In most cases, the process of experimental research on designed algorithms

Descriptions of individual control algorithms are supplemented with numerical test

2.1 Two-Wheeled Mobile Robot

© Springer International Publishing AG 2018 11

2.1.1 Description of the Kinematics of a Mobile Robot

ẏ A = ẋ A tan (β) , (2.1)

where β – instantaneous angle of rotation of the WMR’s frame.

x H = x A + l3 cos (β) , (2.2)

y H = y A + l3 sin (β) . (2.3)

Differentiating the above relations with respect to time, we get

ẋ H = ẋ A + l3 β̇ sin (β) , (2.4)

ẏ H = ẏ A + l3 β̇ cos (β) . (2.5)

ẋ A = v A cos (β) , (2.6)

ẏ A = v A sin (β) . (2.7)

ẋ H = v A cos (β) + l3 β̇ sin (β) , (2.8)

ẏ H = v A sin (β) + l3 β̇ cos (β) . (2.9)

It is assumed that point H moves in the x y plane, on a path that is analytically

where by differentiating (2.10) with respect to time, we get

and in the scalar form

where r1 = r2 = r – radiuses of the WMR drive wheels.

The inverse form of relation (2.17) takes the form

2.1.1.1 Simulation of Loop-Shaped Path Inverse Kinematics

where th – braking start-time, ς – transition curves approximation constant, α̇ p1 , α̇ p2

where c1 = 12 [1/s] – sigmoid functions slope coefficients, t1 – motion mean start-

2.1.1.2 Simulation of 8-Shaped Path Inverse Kinematics

It is the evening of the 30th of June.[822] Fiery copper has been

About midnight the order is given to march.[824] Stealthily they

[832] Ramirez, Proceso contra Alvarado, 4, 53, 68.

[836] Zamacois makes atonement for a lack of research by inventing doughty

[841] Alvarado was taken on Gamboa’s horse, Laso on Sandoval’s. Ramirez,

Fatal Mistake of the Mexicans—A Brief Respite Allowed the Spaniards—

What would Emperor Charles have said to Hernan Cortés had

You might also like