Biologically Inspired Control of Humanoid Robot Arms (eBook)

Dr. Adam (‘Ad’) Spiers is an Associate Research Scientist at the GRAB lab in Yale University (Connecticut, USA). The majority of this book represents research he carried out for his PhD (2007-2011) at the Bristol Robotics Laboratory (BRL) and University of Bristol (UK) under the supervision of Dr. Guido Herrmann and Prof. Chris Melhuish. In addition to a PhD, he holds an MSc in Engineering and Information Sciences (2006) and BSc in Cybernetics and Control Engineering (2004), both from the University of Reading (UK). He currently conducts research into human movement, upper limb prosthetics, underactuated robot hands and shape-changing haptic navigation interfaces for pedestrian guidance. His other research work, outside that noted in this book, has included surgical robotics, tele-haptics, medical diagnostic simulators, robot gripper / sensor development, tactile displays, neural interfacing and enactive interfaces. He has also worked on remote handling robots at UK nuclear facilities and collaborated with several artists to create unique technology driven theater and art installations. He was an invited resident of the Pervasive Media Studio (Bristol, UK) from 2012-2014.Dr Said G Khan is an Assistant Professor in Robotics and Control in the Department of Mechanical Engineering, College of Engineering Yanbu, Taibah University Al Madinah, Saudi Arabia. He has previously worked as a postdoctoral researcher at the Department of Mechanical Engineering, Queen's School of Engineering, University of Bristol. Dr. S G Khan is a member of the Nonlinear Robotics Control Group (NRCG) at the Bristol Robotics Laboratory. He has completed his PhD at the Bristol Robotics Laboratory, University of the West of England (and University of Bristol) in 2012 in Robotics and Control, specializing in “safe human-robot interaction via adaptive compliance control”. He did his M.Sc. Robotics with Distinction from the University Of Plymouth, UK in 2006. He received his B.Sc. in Mechanical Engineering, First Division with Honours, from University of Engineering Technology (UET) Peshawar, Pakistan in October, 2003. He was a part-time lecturer in University of the West of England, Bristol, from September 2008 to October, 2011, and a Research Associate at Department of Mechanical Engineering, GIK Institute of Engineering Sciences and Technology, Pakistan, from March 2007 to July 2008.Dr. Guido Herrmann is an accomplished control engineer and a Reader in Control and Dynamics at the University of Bristol with extensive experience in the development and implementation of novel control schemes for industrially relevant systems. He received the German degree 'Diplom-Ingenieur der Elektrotechnik' (with highest honours) from the Technische Universität zu Berlin. In 2001, he obtained a PhD from the University of Leicester. From 2001 to 2003, he worked in the mechatronics and micro-systems group of the A*Star Data Storage Institute (Singapore) doing research and consultancy for the data storage industry. From 2003 until February 2007, he was a Research Associate, Research Fellow and a Lecturer in the Department of Engineering at Leicester University. In March 2007, Dr Herrmann took up a permanent lecturing position at the Department of Mechanical Engineering of the University of Bristol. He was at several occasions a Visiting Lecturer and a Visiting Professor in Malaysia and Singapore; in September 2005, he was a Visiting Professor at the Data Storage Institute. He is leading the Nonlinear Robotics Control Group (NRCG) at the Bristol Robotics Laboratory. He is a Senior Member of the IEEE and a Fellow of the IET. He has been an Associate Editor of the International Journal of Social Robotics (since foundation in 2009) and a Technical Editor of the IEEE/ASME Transactions on Mechatronics (2009/2010-2014/2015 editorial cohort).

Preface 8
Contents 12
Nomenclature 18
1 Introduction 21
1.1 Prologue 21
1.1.1 Industrial Robots 22
1.1.2 Humanoid Robots 24
1.1.3 The Importance of Human-Like Motion 25
1.1.4 Biologically Inspired Design 26
1.1.5 Physical Safety and Active Compliance for Safety 26
1.1.6 Robust and Adaptive Control 27
1.2 Objective of the Book 27
1.3 Guidance for the Reader 28
1.3.1 Recommended Reading Routes 29
References 30
Part I Background on Humanoid Robots and Human Motion 33
2 Humanoid Robots and Control 34
2.1 Humanoid Robots 34
2.1.1 Functional Tools 35
2.1.2 Models of Humans 36
2.1.3 Human–Robot Interaction 36
2.2 Goals of Human-Like Motion 37
2.3 Robot Motion Control Overview 38
2.3.1 Kinematics-Based Robot Motion Control 40
2.3.1.1 Forward Kinematics 41
2.3.1.2 Inverse Kinematics 41
2.3.1.3 Basic Inverse Kinematics Example 42
2.3.1.4 Kinematics Control Discussion 43
2.3.2 Dynamic-Based Robot Motion Control 44
2.3.2.1 Dynamic Modelling 45
2.3.2.2 Model Specification 46
2.3.3 Optimal Control 47
2.3.4 Operational Space Control 48
2.3.5 Dual Robot Arm Control 49
2.3.6 Hand Grasping Control 50
2.4 Sensing and Robot Arm Motion 51
2.5 Robot and Control Hardware 52
2.5.1 Elumotion Robotic Platform 53
2.5.2 Robot Structure 56
2.5.3 Actuators 58
2.5.4 Motor Drivers 58
2.5.5 EPOS Interface method 59
2.6 Summary 60
References 60
3 Human Motion 67
3.1 Introduction 67
3.2 Motion Studies 68
3.3 Motion Models 69
3.3.1 Kinematic Models 69
3.3.2 Dynamic Models 71
3.4 Physiological Modelling 73
3.4.1 Muscle Models 73
3.4.2 Physiological Complexity 74
3.4.3 Neural Models 75
3.4.4 Simplified Models 77
3.5 Motion Capture Methods and Technology 77
3.6 Human Motion Reproduction and Synthesis 81
3.6.1 Direct Reproduction of Human Motion 81
3.6.2 Learning Techniques 84
3.6.3 Dynamic Movement Primitives 85
3.6.4 Operational Space Control 86
3.7 Summary 86
References 88
Part II Robot Control: Implementation 93
4 Basic Operational Space Controller 94
4.1 Introduction 94
4.1.1 Human Verification 95
4.1.2 Robot Specification 96
4.1.3 Robot Goal Modification 98
4.2 The Operational Space Mathematical Formulation 98
4.3 Task Control 99
4.3.1 Jacobian Pseudo Inverse 99
4.3.2 Task-Space Dynamic Projection 100
4.3.3 Feedback Linearisation 102
4.4 Posture Control 103
4.4.1 `Effort' Cost Function 103
4.4.2 Task/Posture Isolation 106
4.5 Simulation and Implementation 107
4.5.1 Controller Realisation 107
4.5.2 Simulation Results 109
4.5.2.1 Task Only Control 109
4.5.2.2 Posture-Dependent Trajectories 109
4.5.3 Robot Implementation 114
4.5.3.1 Increased Gains 115
4.6 Summary 116
References 117
5 Sliding Mode Task Controller Modification 118
5.1 Introduction 118
5.2 Sliding Mode Control Overview 119
5.3 Controller Design 120
5.3.1 Switching Function 120
5.3.2 Variable Structure Law 122
5.4 Lyapunov Stability Analysis 124
5.5 Results 125
5.5.1 Simulation 125
5.5.2 Physical Robot 128
5.5.3 PID Results 128
5.5.4 Sliding Mode Results 129
5.5.5 Demand Filter 130
5.6 Compliance 131
5.7 Summary 131
References 132
6 Implementing `Discomfort' for Smooth Joint Limits 133
6.1 Introduction 133
6.1.1 Dynamic Model Simplicity 134
6.2 Visualisation Technique 135
6.2.1 Motion Analysis 137
6.3 Joint Limit Function Design 138
6.3.1 Integration with the Effort Function 139
6.4 Results 140
6.4.1 Simulated Results 140
6.4.2 Practical Results 142
6.5 Summary 144
References 146
7 Sliding Mode Optimal Controller 147
7.1 Introduction 147
7.2 Controller Design 148
7.2.1 Optimal Sliding Surface 148
7.2.1.1 Modified Model for Analysis and Controller Design 149
7.2.1.2 Lyapunov Analysis of Surface 150
7.2.2 Control Method 151
7.2.2.1 Controller Design 151
7.2.2.2 Controller Design with Estimates 152
7.2.3 Velocity Decoupling 154
7.2.3.1 Cost Function Revision for Decoupling 157
7.2.3.2 Observation on the Cost Function Dynamics 158
7.2.3.3 Sliding Mode Analysis for Posture Control Only 159
7.2.4 Overall Controller 160
7.3 Implementation Issues: Viscous Friction Identification and Compensation 162
7.4 Simulated Implementation 163
7.4.1 Controller Effort 165
7.4.2 Friction Model 167
7.4.3 Simulated Results 167
7.5 Practical Implementation 172
7.6 Summary 174
References 175
8 Adaptive Compliance Control with Anti-windup Compensation and Posture Control 176
8.1 Introduction 176
8.2 Adaptive Compliance Control for Task Motion 178
8.2.1 Impedance Reference Model 180
8.2.2 Principle of the Model Reference Scheme 181
8.3 Effort-Minimising Posture Torque Controller 182
8.4 Anti-windup Compensator 183
8.5 Implementation 186
8.6 One-Dimensional Adaptive Compliance Control of a Robot Arm 186
8.6.1 Tracking 186
8.6.2 Compliance Results 188
8.6.3 Anti-windup Compensator Results 189
8.7 Multidimensional Adaptive Compliance Control of a Robot Arm 196
8.7.1 Joint Torque Sensors and Body Torque Estimates 197
8.7.2 Tracking and Compliance Results 200
8.7.3 Anti-windup Compensator Results 200
8.8 Summary 205
References 205
Part III Human Motion Recording for Task Motion Modelling and Robot Arm Control 207
9 Human Motion Recording and Analysis 208
9.1 Initial Motion Capture Objective 208
9.1.1 The Vicon System 212
9.1.2 Experimental Set-up 213
9.1.3 Results 214
9.1.4 Summary of Initial Motion Capture Experiments 216
9.2 Motion Capture for Robotic Implementation 219
9.2.1 Human–Robot Kinematic Mismatch 219
9.2.2 Motion Capture Process for Inconsistent Kinematic Models 221
9.2.3 Extended Motion Capture Method 223
9.2.4 Vicon Skeleton Model 225
9.2.5 Incompatible Kinematics Removal 226
9.2.6 Inverse Kinematics 228
9.2.7 Trajectory Discrepancy 230
9.3 Four Degrees of Freedom Comparative Trials 231
9.3.1 Results 232
9.4 Summary 234
References 235
10 Neural Network Motion Learning by Observation for Task Modelling and Control 237
10.1 Introduction 237
10.1.1 Learning by Observation 238
10.2 Learning by Observation Method 240
10.3 Minimal Trajectory Encoding 241
10.3.1 Polynomial Encoding Issues 241
10.3.2 Scaling and Fitting of Generated Trajectories 243
10.4 Network Structure 245
10.5 Experimental Procedure 246
10.5.1 Sub-motion Splitting 247
10.5.2 Training Data 247
10.5.3 Neural Network Results 249
10.6 Integration into the Robot Controller 252
10.7 Summary 257
References 257
Appendix A Kinematics: Introduction 259
A.1 Kinematics Notation 259
A.1.1 Position Vector 260
A.1.2 Rotation Matrix 261
A.1.3 Transformation Matrix 263
A.2 Denavit–Hartenberg Notation 263
A.2.1 Frame Assignment Convention 264
A.2.2 DH Parameters 264
A.3 Applied Kinematics 265
A.3.1 Forward Kinematics 265
A.3.2 Inverse Kinematics 266
A.4 Robot Jacobian 266
References 267
Appendix B Inverse Kinematics for BERUL2 268
B.1 Denavit–Hartenberg Parameters 268
B.2 Forward Kinematics 268
B.3 Algebraic Solution 269
Reference 275
Appendix C Theoretical Summary of Adaptive Compliant Controller 276
C.1 Proof of Theorem 1 276
References 280
Appendix D List of Videos 282
Index 283