Engineering Series Lyapunov-Based Control of Robotic Systems Aman Behal Warren Dixon Darren M. Dawson Bin Xian CRC Press i renMATLAB‘ is a trademark of The MathWorks, Ine. and is used with permission. The MathWorks does not warrant the accuracy of the text or exercises in this book. This book's use or discussion of MATLAB* soft- ware or related products does not constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular use of the MATLAB" software. CRE Press Taylor & Francis Group (6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 ©2010 by Taylor and Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original US. Government works Printed in the United States of America on acid-free paper 10987654321 International Standard Book Number: 978-0-8493-7025-0 (Hardback) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders ofall material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmit- ted, or utilized in any form by any electronic, mechanical, or other means, now knownor hereafter invented, including photocopying. microfilming. and recording. or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright. com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC Is « not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. ‘Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging: in-Publication Data Lyapunov-based control of robotic systems / Aman Behal ... [etal p-em.-- (Automation and control engineering) Includes bibliographical references and index. ISBN 978-0-8493-7025-0 (hardcover : alk. paper) 1, Robots--Controlsystems.2. Nonlinear contrel theory.3. Lyapunoy functions. I. Behal, Aman. ‘T}211.35.L83 1008 629.892--de22 2009040275, Visit the Taylor & Francis Web site at http:/www.taylorandfrancis.com and the CRC Press Web site at hitp:/www.ctepress.comContents oo ee Cee eee eee ee ee es 1 Introduction 1d 1.2 1.3 History of Robotics Lyapunoy-Based Control Philosophy. ‘The Real-Time Computer Revolution . References eee 2 Robot Control 2 2.2 23 24 Introduction Modeling and Control Objective 2.2.1 Robot Manipulator Model and Properties 22.2 Control Objective Computed Torque Control Approaches 23.1 PDControl..... 2.3.2 Robust Control. . . 2.3.3. Sliding Mode Control Adaptive Control Design . . cece ee 24.1 Direct Adaptive Control... 2.2... 2.4.2 Neural Network-Based Control Task-Space Control and Redundancy . . 25.1 Kinematic Model... 0.0.0.0. 0c 2.5.2 Control Objective and Error System Formulation . . aeor st 10 10 12 12 12 15 16, 17 18 24 28 29 30Contents 2.5.3 Computed Torque Control Development and Stabil- (PEMMAYBilics & acs o oo we as ew OH . 32 5.4 Adaptive Control BYtension. occ... . 33 References... ee eee 34 3 Vision-Based Systems 37 84, Wtioductionisn awe 4 wey % OER BEES 4 BES FS . 37 3.2 Monocular Image-Based Geometry . 2... 0.0. ss 2 41 3.2.1 Fixed-Camera Geometry... 2.0.0.0 00 55 . 41 3.2.2. Euclidean Reconstruction... 2... .....5 . 44 3.2.3. Camera-in-Hand Geometry... 0.00.0 0s 46 3.2.4 Homography Calculation»... .....0..5 . 4a 3.2.5 Virtual Parallax Method... 2.0.0.0 006s 50 3.3 Visual Servo Tracking ©... eee 51 3.3.1 Control Objective 2.2... ee eee 51 3.3.2. Control Formulation ©... 22. eee 54 3.3.3. Stability Analysis... 6... eee 56 3.3.4 Camera-in-Hand Extension... 00.0000. 37 3.3.5 Simulation Results... 220. 58 34 Continuum Robots 6... 6 ee 05 3.4.1 Continuum Robot Kinematics . Biime & .4.2 Joint Variables Extraction... 6.0.0.0... 72 3.4.3. Task-Space Kinematic Controller... 2... 5 . 74 3.4.4 Simulations and Discussion... 0... 2... . 76 3.5 Mobile Robot Regulation and Tracking... 0.2... . 78 3.5.1 Regulation Control... 02.2... - 2 eee . 79 3.5.2 Tracking Control... 2... eee . 98 3.6 Structure from Motion... 0.00... 0 ee eee . 107 3.6.1 Object Kinematics 2.0... 2. eee . 107 3.6.2 Identification of Velocity. 6... es 108 3.6.3 Camera-in-Hand Extension... 0.0.0.0. 55 U3 3.6.4 Simulations and Experimental Results oss. 119 BT Notes 2... cee eee ee . 125 References... eee . 129 4 Path Planning and Control 141 4.1 Introduction ql 4.2. Velocity Field and Navigation Function Control for Manip- ulators 144 ADA. Spat MOUs © nox x eens a os vee 2 . 145 4.2.2 Adaptive VFC Control Objective ea a Bee 1464.3 4.4 45 4.6 Contents: 4.2.3 Navigation Function Control Extension . 0... . . 4.2.4 Experimental Verification ©. .......0.005 Velocity Field and Navigation Function Control for WMRs 43.1 Kinematic Model 4.3.2. WMR Velocity Field Control»... . . 43.3 WMR Navigation Function Control Objective . Vision Navigation... 2.22... 4.4.1 Geometric Modeling... . . 44.2 Image-Based Path Planning . 44.3 Tracking Control Development . . 4A. Simulation Results . Optimal Navigation and Obstacle Avoidance 4.5.1 IMlustrative Example: Planar PBVS .. . . 4.5.2 6D Visual Servoing: Camera-in-Hand . . . Background and Notes References . . . Human Machine Interaction 5. 5.2 z 5.5 TinbrOdtiGtiGit. ae eos ce x eee ce wr atone & eet co we ore Exercise Machine ........ 5.2.1 Exercise Machine Dynamics . 5.2.2 Control Design with Measurable User Input Desired Trajectory Generator . Control Design without Measurable User Input Desired Trajectory Generator... .. Experimental Results and Discussion Steer-by-Wire . Control Problem Statement Dynamic Model Development Control Development Stability Analysis Elimination of Torque Measurements: Extension Numerical Simulation Results .7 Experimental Results Robot ‘Teleoperation ..... . . 5.4.1 System Model . . 54.2 MIF Control Development : UMIF Control Development . Rehabilitation Robot . 55.1 Robot Dynamics 2 Path Planning and Desired Trajectory Generator . - 184 191 233 241 246 7 260 265, 271 274 277 278 284 295, 296 297Contents 5.5.3. Control Problem Formulation . . . . 5.5.4 Simulation Results . 5.6 Background and Notes References... 0... eee ee ee Appendices Mathematical Background References o 0. 0 oo. eae ees Supplementary Lemmas and Expressions B1 Chapter 3Lemmas............-. B.L.1 Open-Loop Rotation Error System . B.L.3_ Persistence of Excitation Proof B.2 Chapter 4 Lemmas and Auxiliary Expression Experimental Velocity Field Selection GUBLemma...........05 Boundedness of By (t) . 0... Open-Loop Dynamies for ¥ (t) Measurable Expression for Ly, (t) Global Minimum . B.3. Chapter 5 Lemmas and Auxiliary Expression: B.3.1 Numerical Extremum Generation . . Proof of Lemma5.1 0. ......5 Definitions from Section 5.3.2... . Upperbound for Vai (t). 0... es Upper Bound Development for MIF An: ‘Teleoperator ~ Proof of MIF Controller ‘Teleoperator ~ Proof of MIF Passivity Boundedness 2... 2-0 00-5% Development of an Image Space NF and Its Gradient alysi sis. Stability Teleoperator — Proof of UMIF Desired Trajectory B.3.9 Teleoperator ~ Proof of UMIF Controller Stability . 1 Proof of Bound on } 2 Calculation of Region of Attraction References Index 302 307 317 . 318 326 327 334 335 335 335 337 337 339 339 340 342 344 344 345 347 347 347 349 350 350 351 304 358) 359 363 366 . 367 369 370 373aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.xii Preface cra) — interestingly, both these scenarios require extra preprocessing and computation to extract the required information. All the aforementioned factors can conspire to make for highly challenging control design scenarios. As the title suggests, our preferred framework for addressing these issues will be Lyapunov-based nonlinear control design. Robots are highly nonlin- ear systems and even though they can be linearized under some restrictive assumptions, most practical scenarios require the design of nonlinear con- trollers to work around uncertainty and measurement-related issues. It has been our experience over the years that Lyapunov’s direct method is an extremely effective tool to both design and analyze controllers for robotic systems. Chapter 1 begins by providing a brief history of robotics. It is followed by an introduction to the Lyapunoy-based design philosophy — pros and cons are discussed. The chapter ends on a practical note by describing the evolution of real-time control design systems and the associated operat- ing environments aud hardware platforms that they are based upon. In Chapter 2, we provide the reader with a quick introduction to a host of standard control design tools available for robotic systems. In order to pre pare for the chapters ahead, all these techniques are analyzed in a common Lyapunoy-based framework. The chapter begins by discussing computed torque methods, where the model nonlinearities are canceled through exact model knowledge, and the system robustness to unmodeled disturbances is discussed under PD control, continuous robust control, and sliding mode control. Next, adaptive control techniques are discussed when the model is uncertain. A variety of adaptive update laws are discussed including the de- sign of a NN-based strategy when the model cannot be expressed as linear in the unknown/uncertain parameters. The chapter closes by discussing the challenges of designing control laws for redundant link robot manipulators when the control objectives are stated in the task-space of the robot. When robots need to navigate through and/or interact with unstruc- tured or dynamically changing environments, camera-based vision systems are utilized to provide adequate sensing of the environment. Chapter 3 discusses some problems in visual servoing control. The first problem ad- dressed is that of a robot end-effector tracking a prerecorded time-varying reference trajectory under visual feedback from a monocular camera — the use of a single camera results in uncertainty in the depth parameter which needs to be adaptively corrected. The next problem that we address in this chapter is that of estimating the shape of a continuum robot. Tra~ ditional position sensing devices such as encoders cannot be used in this situation since it is not easy to define links and joints for such robotsPreface xiii instead, a vision based solution is proposed and then validated through the design of kinematic controllers that regulate the pose of the robot end- effector via feedback from a sequence of images from a fixed monocular camera. The third problem dealt with in this chapter is that of design- ing homography-based visual servo control methods for solving tracking and regulation problems in the context of wheeled mobile robots. The fi- nal problem addressed in Chapter 3 is the classic Structure from Motion (SFM) problem, specifically the development of an adaptive nonlinear es- timator in order to identify the Euclidean coordinates of feature points on an object based upon relative movement between that object and a single fixed camera. Chapter 4 deals with the problems of path planning and control for manipulator arms and wheeled mobile robots, both when the obstacle lo- cations are known a priori and when they need to be determined in real time using fixed or in-hand vision as an active feedback element. The first problem addressed is path following using velocity field control (VFC) this technique can be applied when it is more critical to follow a contour exactly than it is to track a desired time-varying trajectory (which inciden- tally is the standard problem solved in most robotics literature). Another application of path following is when a navigation function (NF) approach is utilized to create a path around obstacles to a desired goal location. As an extension, we also show how VFC- and NF-based techniques can be uti- lized to solve the obstacle avoidance problem for mobile robots. We then shift gears and address the problem of hybrid servoing control under visual feedback, which may be required to manipulate a robot in unstructured cavironments — the challenge here is the design of a desired trajectory in the image space based on an image space NF that ensures that the features on the object stay in the camera’s field-of-view through the course of the chapter deals with the design of an image space extremum seeking path planner such that a singularity free PBVS controller can be designed that works on visual feedback and is able to reject lens distortion and uncertainties in the camera calibration In Chapter 5, we deal with the emerging research area of human-machine interaction. While the primary control objective during human-machine robot’s motion. The final portion of this interaction is application specific, the secondary objective invariably is to ensure the safety of the user — to this end, we illustrate the design of control schemes based on passivity such that the machine is a net energy sink. The chapter begins hy exploring smart exercise machines that provide optimal physical training for the user by altering the machine's resistance based on user performance. Steer-by-wire control of vehicles is discussedxiv Preface next, with the focus being on locking the steering response of the vehicle to the user input as well as ensuring that the road feel experienced by the user can be appropriately adjusted. ‘The third problem addressed in this chapter is that of teleoperator systems where the focus is on both facilitating the application of desired motion and desired force in the remote environment by the user, as well as ensuring that the system is safely able to reflect desired forces back to the user. The final topic addressed in Chapter 5 is that of a rehabilitation robot which is safely able to direct user limb motions along selectable trajectories in space that optimize their rehabilitation after disease or injury. The material in this book (unless noted otherwise) has been derived from the authors’ research work during the past several years in the area of con- trols and robotics. This book is aimed at graduate students and researchers who would like to understand the application of Lyapunoy-based control design techniques to emerging problems in robotics. This book assumes a background in undergraduate level linear controls theory. Some knowledge of nonlinear systems and Lyapunoy-bas niques for such sys- ed design tems may be desirable ~ however, the book contain adequate background material in Chapter 2 and Appendix A as well as references to textbooks that deal with these subjects should they be of interest to the reader. ‘The authors would like to thank our colleagues and collaborators for their valuable assistance and counsel without which this work would not have been possible. We are especially grateful to Dr. Jian Chen, whose work early on as a graduate student at Clemson University and later as a collaborator of the authors, has influenced many of the issues examined in this book. We would also like to acknowledge the unselfish support of the following past and present members of the Controls and Roboties group in the Department of Electrical Engincering at Clemson University: Dr. Vilas Chitrakaran, Dr. Michael McIntyre, Dr. Pradeep Sctlur, Dr. Ximing Zhang, Dr. David Braganza, Dr. Enver Tatlicioglu, and Mr. Abhijit Baviskar. Finally, we would like to thank Mr. Jonathan Plant, Senior Editor at CRC Press for his patience and counsel as well as Marsha Pronin, Project Coordinator, and Karen Simon, Project Editor, for their help with reviewing samples and assistance with preparation of the camera-ready manuscript.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2 1, Introduction functioning robot named HAL that turns rogue and is eventually discon- nected. A movie that received a great deal of acclaim was Ridley Scott’s Blade Runner that was released in 1982 and which featured Harrison Ford asa hunter of illegal mutinous androids known as Replicants. Other movies like The Terminator and the Matrix series have dealt with sophisticated, high-functioning humanoids. Most recently, Pixar produced the smash hit animated robotics film WALL-E which features a sentimental robot of the same name that is designed to clean up the pollution created by mankind. Space exploration has advantageously employed manipulator arms and mobile robots over the years, Lunokhod 1 and 2 were the first robotic ex- ploration vehicles (rovers) to be launched to an extraterrestrial body, the moon, by the Soviets in 1970. After a long gap, the rover Sojourner landed on Mars in 1997 as part of the Pathfinder Mission; it had vision assisted autonomous navigation and was successful at obtaining and analyzing rock and soil samples. This was followed in 2004 by the rovers Spirit and Op- portunity that are still active and continue to analyze the Martian geology, as well as its environment, to assess the possibility that life may have been supported on Mars in the past. Most recently, the Phoeniz lander executed the first successful polar landing on the Martian surface and is currently exploring the possibility of water existing or having existed on the red planet. ‘The first commercial robotics company, named Unimation, was started in 1956 by George Devol and J. Engelberger. As a result of this venture, the first industrial robot was manufactured and marketed in the United States. Unimate began work in a General Motors automobile plant in New Jersey in 1961. This manipulator arm performed spot welding operations as well as unloading of die casts. This was followed in 1978 by the Programmable Universal Machine for Assembly, a.k.a. PUMA. Since that time, quite a few other robot manufacturers have come and gone with only a few achieving commercial success or longevity in the market. In recent years, personal and professional service robots have picked up steam. For example, Lego has achieved success with its Mindstorms Robotics Invention System as has Sony with its AIBO robot pets. Most recently, Honda’s humanoid robot ASIMO has hogged media limelight with its ability to perform a wide variety of service and human interaction tasks. Robotics in research settings has steadily continued to experience an up- ward spiral since its inception. Robotics research got its academic start in 1959 with the inauguration of the Artificial Intelligence Laboratory at the Massachusetts Institute of Technology by John McCarthy and Marvin Min- sky. Other inaugurations of note were the establishment of the Artificial1.2 Lyapunov-Based Control Philosophy 3 Intelligence Laboratory at Stanford University in 1963, and the Robotics In- stitute at Carnegie Mellon University in 1979. The first computer-controlled mechanical hand was developed at MIT in 1961 followed by the creation of the Stanford Arm in the Stanford Artificial Intelligence Laboratory by Vie- tor Scheinman in 1969. Early flexible robots of note were Minsky’s octopus- like Tentacle Arm (MIT, 1968) and Shigeo Hirose’s Soft Gripper (Tokyo Institute of Technology, 1976). An eight-legged walking robot named Dante was built at Carnegie Mellon University which was followed by a more ro- bust Dante II that descended into the crater of the yoleano Mt. Spurr in Alaska in 1994. Demonstrations of planning algorithms for robots began in the late 1960s when Stanford Research Institute's Shakey was able to navigate structured indoor environments. A decade later, the Stanford cart attempted navigation of natural outdoor scenes as well as cluttered indoor environments. Modern robotics research is focused on higher dimensional robots, modular robots, and the planning issues associated with these types of devices. Simultaneously, robotics is making great strides in medicine and surgery as well as assistance for individuals with disabilities. 1.2 Lyapunov-Based Control Philosophy ‘The requirements for increasing levels of autonomy and precision in robots » Mul- tiple link robots have presented complex, coupled nonlinear dynamics that have inspired the design of numerous output and state feedback control designs, especially the global output feedback problem for robots has been very challenging. Wheeled mobile robots have inspired the design of set- have necessitated the development of sophisticated control strategi point and tracking controllers for nonholonomic systems. Other appliea- gid link flexible joints in the late nineties as well as higher dimensional and continuum tions that have challenged control d igners have been robots of late. Linear control design is often inadequate outside narrow operating regimes where linearized system models are valid. Nonlinear control strategies can take advantage of full or partial knowledge of the structure and/or pa- rameters of the em in order to craft techniques that are robust to ex- Re- have utilized a variety of tools for analyzing nonlinear ogenons disturbances, measurement noise, and unmodeled dynamics search investigators systems arising from nonlinear controllers, nonlinear models, or a combina- tion thereof — singular perturbation, deseribing functions, and phase plane analysis are some of the popular tools. However, Lyapunov-based tech- niques (in particular, the so-called direct method of Lyapunov) offer the4 1, Introduction distinct advantage that they allow both design and analysis under a com- mon framework with one stage motivating the other in an iterative fashion. Lyapunov theory and its derivatives are named after the Russian mathe- matician and engineer Aleksander Mikhailovich Lyapunov (1857-1918) Lyapunov stability theory has two main directions — the linearization method and the aforementioned direct method of Lyapunov. The method of linearization provides the fundamental basis for the use of linear con- trol methods [1]. It states that a nonlinear system is locally stable if all the eigenvalues of its linear approximation (via a Taylor series expansion about a nominal operating point) are in the open left half plane and is unstable if at least one is in the open right half plane. Furthermore, the stability not be determined without further analysis if the linearization is marginally stable. The direct method of Lyapunov relies on the physical property that a system whose total energy is continuously being dissipated must even- tually end up at an equilibrium point [1, 2]. Given a scalar, non-negative energy (or energy-like) function V(t) for a system, it can be shown that if its time derivative V (¢) < 0, the system is stable in the sense of Lyapunov ‘an in that the system states (energy) can be constrained for all future time to lie inside a ball that is directly related to the size of the initial states of the system, While a lot of results have been derived in the last fifty years in order to deduce stability properties based on the structure of the Lyapunov function V(t) and its time derivative, we are no closer to understanding how one may choose an appropriate V (t), i it is not clear how closely the scalar function V (¢) should mimic the physical (kinetic and potential) energy of the system, What is clear is that the objectives of the control design and the constraints on the measurements lead to the definition of system states that often guide the development of the Lyapunov function. Furthermore, the control design itself is impacted by the need to constrain the time derivative of the Lyapunov function to be negative definite or semi-definite ‘Thus, the control design and along the closed loop system trajectories the development of the Lyapunov function are intertwined, even though the presentation may tend to indicate a monotonic trajectory from control design toward stability analysis. In the ensuing chapters, one will be able to gain an insight into the variety of choices for Lyapunov functions as well as the appearance of non-intuitive terms in the control input signals that will likely indicate an influence of the Lyapunov-based analysis method on the control design. While Lyapunov’s direct method is good at characterizing the stability of equilibrium points for autonomous and nonantonomous systems alike, it1.3 The Real-Time Computer Revolution 5 works equally well in showing the boundedness and ultimate boundedness of solutions when no equilibrium points exist [3]. Furthermore, the analysis not only provides a guarantee of stability and the type of stability result (uniform asymptotic, exponential, semi-global ultimately bounded, etc.), it is also able to point out bounds on the regions where the results are guaran- teed to be valid. This is in sharp contrast to linearization based approaches where regions of convergence are not easily obtained. Finally, Lyapunov- based design leads to faster identifiers and stronger controllers that are able to prevent catastrophic instabilities associated with traditional esti- mation based methods such as certainty equivalence [4]. While traditional methods work well with linear systems, they can lead to troubling results such as finite escape times in the case of nonlinear systems. A shortcoming of the Lyapunoy-based analysis techniques is that the chosen parameters (while guaranteed to produced closed-loop stability) may be too conserva- tive, thereby compromising the transient response of the sysiem. Moreover, Lyapunov stability theorems only provide sufficient conditions for stability, ie., without further work, it is not possible to say which of those conditions are also necessary [3]. 1.3. The Real-Time Computer Revolution As we will see in the ensuing chapters, the nonlinear control, estimation, and observation schemes emanating from an application of Lyapunov’s di- rect method tend to have a complex structure and are generally computa- tionally intense compared to their linear counterparts [2]. Thus, there is a requirement for the use of microprocessors, microcontrollers, and/or com- puters to crunch the numbers. Furthermore, there is a requirement for fast interface hardware for allowing the physical (generally analog) world to in- teract bidirectionally with the digital domain without creating instabilities and uncertainties due to factors such as phase lags from slow computation, quantization noise due to finite precision, aliasing due to slow sampling, uncertain order of execution of various computation modules, ete. In the last decade, a multitude of control environments have been created in the academic and industrial research and development. communities to serve this emerging need for reliable real-time computation A real-time implementation is different from a traditional implementa- tion in that the worst-case performance of the hardware and the software is, the most important consideration rather than the average performance. In real-time operation, the processing of external data arriving in the com- puter must be completed within a predetermined time window, failingaa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.References 7 have made it possible for GPCs to be able to execute complex control strategies at rates in excess of tens of KHz. Today’s desktop computers can not only run complex control algorithms in the background, they can simultaneously render high-bandwidth data in one or more GUI windows. Furthermore, the other advantage offered by multiprocessor architectures, namely, deterministic response, has been eliminated by the emergence of hard real-time operating systems such as RTLinux and QNX [7]. Examples of systems that hamess the PC’s computational ability include Opal-RT ‘Technologies’ RT-LAB as well as QMotor 2.0 and its successor QMotor 3.0 both of which were developed by the Controls and Robotics group at Clem- son University. All of these systems are based on QNX which is a real-time microkernel operating system [9]. In particular, QMotor 3.0 [8] allows easy incorporation of new hardware by employing a client-server architecture where data from hardware can be accessed by the control program (client) by talking to the server for that hardware — this communication is done via message passing or by using shared memory. Nearly all of the robot path planning and control algorithms derived in the ensuing chapters have been validated through simulation or experi- ments. While MATLAB™ and SIMULINK™ have been the primary envi- ronments for running computer simulations, the experimental work shown, here has been performed via control code written in C and compiled and executed in the QMotor environment running on QNX-based desktop com- puters. References E, Slotine and W. Li, Applied Nonlinear Control, Englewood Cliff, Prentice Hall, Inc., 1991. ti [2] M. S. de Queiroz, D.M. Dawson, S. Nagarkatti, and F. Zhang, Lyapunov-Based Control of Mechanical Systems, Cambridge, MA: Birkhauser, 2000. [3] H. K. Khalil, Nonlinear Systems, 3*¢ edition, Prentice Hall, 2002 [4] M. Krstic, I. Kanellakopoulos, and P. Kokotovic, Nonlinear and Adap- tive Control Design, New York: John Wiley & Sons, 1995. [5] M. Colnaric, “State of the Art Review Paper: Advances in Embedded Hard Real-Time Systems Design,” Proceedings of the IEEE Interna- tional Symposium on Industrial Electronics, Vol. 1, pp. 37-42, 1999.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.10 2. Robot Control then designed and analyzed to show that high frequency feedback can be utilized in lieu of high gain feedback to damp out the disturbance uncer- tainty. Finally, a discontinuous controller is also presented to work around the issue of steady-state tracking error. Discontinuous controllers (¢.g., slid- ing mode control) provide a method to obtain an asymptotic/exponential stability result in the presence of uncertainties in the dynamics, provided the actuator is able to provide infinite bandwidth. ‘The problem with high gain and high frequency approaches is that they can lead to reduced stability margins and are susceptible to noise. Adap- tive controllers are utilized in conjunction with or as an alternative to robust control methods. Adaptive controllers are feedforward controlle! with a self-adjusting mechanism to compensate for uncertainties in the sys- tem parameters ~ this online adjustment of the system weights allows for asymptotic tracking without needing high gain or high frequency actuation. However, the price to be paid for this improved performance is an increase in the order of the overall control design. The Adaptive Control Design Section describe: rect adaptive -veral types of adaptive control methods including: a di- ntroller, a desired compensation extension that allows for off-line computation of the regression matrix, and a neural network based. controller when a linear in the parameters model is not available for the uncertainty. Finally, there are unique challenges that emerge when describing the control objectives in the task-space of the robot. Given the fact that the usefulness of robots is generally derived from establishing a desirable pose (ie., position and orientation) of the robot end-effector with respect to the environment, it is often best to formulate control objectives for the robotic system in the task-space. A potential exists for an over- or underdetermined problem based on the task-space objective and the number of actuated joints in the robot. A transformation is also required because the control objective is formulated in the task-space while the control is implemented in the int-space. The last section in this chapter highlights these issues and provides some typical solutions. 2.2 Modeling and Control Objective 2.2.1 Robot Manipulator Model and Properties The system model for an n- link, revolute, direct-drive robot can be written as M(q)G + Vin(9, 44 + G(q) + Fad + Ta = 7 (2.1)2.2 Modeling and Control Objective 11 where q(t), G(t), G(t) € R" denote the link position, velocity, and accel- eration vectors, respectively, M(q) € R"*" represents the inertia matrix, Ving.) € R"*" represents the centripetal-Coriolis matrix, G(q) € R rep- resents the gravity vector, Fy € R"*" is the constant, diagonal, positive- definite, dynamic friction coefficient matrix, rq € R” is a bounded distur- bance vector that represents other unmodeled dynamics (¢.g., static fric- tion), and 7(t) € R" represents the torque input vector. The dynamic equation of (2.1) is assumed to exhibit the following prop- erties which are employed during the control development and stability analysis in the subsequent sections. Property 2.1: The inertia matrix M(q) is a symmetric, positive-definite matrix that satisfies the following inequality mi |Igll? <€7M@E a. ‘The result- ing closed-loop error system can be determined by substituting (2.11) into (2.10) as é+kaé +kpe=M~*ra. (2.13) Based on the closed-loop system of (2.13) and the bounds prescribed in (2.5), it is easy to see the UUB property of the tracking error. ‘ontrol Development and Analysis ‘The PD controller developed in (2.11) is written in a traditional manner gains for the proportional and derivative feedback. An al- ternative approach is provided in this section that makes greater use of the filtered tracking error formulation and illustrates the advantages of us- ing Property 2.2. To facilitate the alternative design, the open-loop error system for r(t) is formulated by taking the time derivative of (2.8), pre- multiplying the resulting expression by M(q), and substituting (2.1) to obtain Mr Méa + Vn(a,4)4 + G(q) + Fad + ta — 7 + ME. (2.14) Motivated by the desire to use Property 2.2, the expression in (2.14) is rewritten as Mi = —-Vi(q.@)r — Far + Ta -T +N. (2.15) where N(q,q,t) € R" is now defined as N = Mdat Vinla.4) (da + ae) + G(q) + Fa (da + ae) + aMe.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2.3 Computed Torque Control Approaches 15 Based on (2.19), the inequality in (2.21) can be written as 2 km Ir@r? < - t) Ga < ie (2.22) = miki ko (2.22) The inequality in (2.18) can be obtained from (2.22) by invoking Lemma. A.21 in Appendix A. lll 2.3.2 Robust Control The previous section illustrated how a PD controller can be used to yield a UUB stability result in the presence of a bounded disturbance torque. From the expression in (2.22), it is clear that by increasing the proportional and derivative gains arbitrarily large (i.¢., high gain feedback), the residual ate error ¢, can be decreased arbitr: small. Yet, high gain feedback can be problematic if noise is present in the system (e.g., feedback from a vision system, use of backwards differencing to obtain velocities from position measurements, etc.). Motivated by the desire to reduce the steady- state error without increasing the gains arbitrarily large, an alternative ‘ontroller can be developed that relies on high bandwidth from the actuator. robust Control Development Consider the following computed torque controller Tokr+ugt+N. a robustifying feedback component up € RB". The high frequency (v structure) robustifying feedback term in (2.23) is defined as where ¢ € R is an arbitrarily small positive design constant. After substi- tuting (2.23) and (2.24) into (2.15), the following closed-loop error system is obtained M? = —-Vin(q@r - Far + ta rtd — kr. (2.25)16 2. Robot Control Stability Analysis Theorem 2.2 Given the open-loop error system in (2.15). the robust con- troller given in (2.23) and (2.24) ensures that the tracking error is uni- formly ultimately bounded in the sense that lle(O)Il S VGaexp(—Cit) + Ca (2.26) for some positive constants Co, 61,¢2- Proof: By using (2.25) and Properties 2.1 and 2.2, the time derivative of (2.19) can be expressed as 2 @ = Litsmiet (Yor — Fur+ rq — Pott ir) 2 lirll Gea + € 7 rea = r'(-Far+ta- TSS - kr) ‘ rll + € < ~rit +e( let) I Cea + € < —h|rP +e. (2.27) ‘The inequality in (2.27) can be used along with the same stability analysis developed for the PD controllers to conclude that mM: km 2 km \ir(t)|? <= ir) |? exp(—S+48) + —(1 -—exp(-4)) m 2 km, 2 2.28) 2 1) — 28 (may, 2c (2. Sa” Fm ) POD ny The inequality in (2.26) can be obtained from (2.28) by invoking Lemma A.21 in Appendix A, Comparing the results in (2.22) and (2.28) highlights the benefits of the robust control design. Specifically, the steady state error ¢y in (2.28) does not depend on the upper bound on the disturbance term Gia and the residual error can be made arbitrarily small by either increasing the control gains (as in the PD controller) or by decreasing the design parameter ¢, Decreasing the design parameter ¢ increases the bandwidth requirements of the actuator. 2.3.3 Sliding Mode Control Control Development The last section illustrates how the increasing the frequency of the con- troller can be used to improve the steady state error of the system. Takenaa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.18 2. Robot Control controllers to compensate for uncertainty that does not satisfy the linear in the parameters assumption (Property 2.3). ‘The controllers in the previous sections were based on the assumption of exact model knowledge (ie., N(-) was used in the control designs to (partially) feedback linearize the system) with the exception of the added disturbance. In this section, the assumption of exact model knowledge is relaxed, but the added disturbance is neglected for simplicity and without loss of generality in the sense that the previous robust and sliding mode feedback methods can be used in conjunction with the methods in this section to also compensate for added disturbances. 2.4.1 Direct Adaptive Control Control Development The open-loop error system in (2.14) (with ry = 0) can be written as Mi =—Vin(9.)r — Far + YO-7 (2.33) where ¥(q.q,t) € R”*? is a nonlinear regression matrix, and @ € RP is a vector of uncertain constant parameters (i.e. linear in the parameters assumption) defined as YO = Maa + Vin(q,4) (Ga + a¢) + G(q) + Fa da + 0€) + aMé. Based on (2.33), an adaptive feedforward controller can be designed as =YO+hr (2.34) where 6(t) € RP denotes a time varying estimate of @. There is significant variation in how the adaptive update law is developed to generate 0(t). Typically, a tracking error based gradient update law is designed motivated by the desire to cancel common terms in the Lyapunov analysis. Based on the subsequent stability analy error system in (2.33) is is, a gradient update law for the open-loop @=TyTr (2.35) where [ € IR’*? is a diagonal matrix of adaptation gains. In practice, the initial condition 6(0) are best-guess estimates of the parameters (e.g., information obtained from a manufacturer's specification sheet, or results from some off-line parameter estimation method, etc.). Substituting (2.34) into (2.33) yields the closed-loop error system Mt = -V,,(q,4)r — Far + YO — kr (2.36)2.4 Adaptive Control Design 19 where @(t) € IR? denotes the mismatch between the unknown parameters and the estimate vector as 0=0-6. (2.37) Stability Analysis Theorem 2.4 Given the open-loop error system in (2.93), the adaptive controller given in (2.34) and (2.35) ensures global asymptotic tracking in the sense that eit) +0 as t— 00. (2.38) Proof: Let V(t) € R denote the non-negative function LBs + joe. (2.39) v 2 By using (2.36), the time derivative of (2.37), and Properties 2.1 and 2.2, the time derivative of (2.39) can be expressed as Vine? (VO — Far br) - 0 T%. (2.40) Substituting (2.35) into (2.40) and canceling common terms yields V=-r? (Far + kr). (2.41) Since the expression in (2.41) is always negative semi-definite, (2.39) can be used to conclude that V(¢),r(t),6(t) € Loo. Since r(t) € Loo, linear analysis methods [9] can be applied to (2.8) to prove that e(¢),€(t) € Lo, and since the desired trajectory ga(t) and its time derivatives are assumed to be bounded, q(é), q(t) € Loo. Given that q(é), g(t) € Loc, then Proper- ties 2.2 and 2.4 can be used to conclude that ¥(q,4,t) € Loo. The fact that 0(t) € Loo can be used with (2.37) to conclude that 0(t) € Lao. Since ¥ (44,1), 0(0), r(t) € Loo, the control is bounded from (2.34) and the adap- tation law 0(t) € Lac from (2.35). The closed-loop error dynamics in (2.36) can be used to conclude that F(t) € £0; hence, r(t) is uniformly continu- ous from Lemma A.12 in Appendix A. Lemma A.14 in Appendix A can be applied to (2.39) and (2.41) to conclude that r(t) is square integrable (ie., r(t) € La). Since r(t),F(t) € Loo and r(t) € Lo, a corollary to Barbalat’s Lemma given in Lemma A.1 in Appendix A can be used to conclude that r(t) 30 as too. (2.42) Based on (2.42), Lemma A.18 in Appendix A can be invoked to conclude the result in (2.38). I20. 2. Robot Control ‘The motivation for the tracking error-based gradient update given in (2.35) is clear from (2.40). Through some control design and analysis mod- ifications, additional update laws can also be used to obtain the result in (2.38) including least squares update laws and composite adaptive update laws based on both tracking and prediction (of the output) error, An ex- ample of a least squares adaptation law based on the tracking error is given by é=py', P=-pytyP (2.43) where P(t) € R?*? is a time-varying symmetric matrix, where P(0) is selected to be a positive definite, symmetric matrix. Composite adaptive update laws are updates based on combining information from both the tracking and the prediction error 6 = PYfe+ PY’ (2.44) where Y;(-) denotes the regression matrix Y(-) after it has been convolved with a low-pass filter, while <(t) denotes the prediction error [12], [28]. DCAL Extension The adaptive update laws in (2.35), (2.43), and (2.44) depend on q(t) and q(t). This dependency means that the regression matrix must be computed or-line and requires velocity feedback. For applications with demanding sampling times or limited computational resources, the need to compute the regression matrix on-line can be problematic, Moreover, velocity measure- ments may only be available through numerical differentiation (and hence, will likely contain noise), and some control designs require the derivative of the regression matrix, thus requiring acceleration measurements. To ad- dress these issues, this section describes the desired compensation adapta- tion law (DCAL) first developed in [24]. DCAL control designs are based on the idea of formulating a regression matrix that is composed of the desired position and velocity feedback rather than the actual states. Therefore, the regression matrix can be computed off-line (for an a priori given trajec- tory) and does not require velocity measurements. Furthermore, even if an adaptation law uses velocity feedback outside of the regression matrix (eg., such as r(t) in (2.35)), the DCAL-based desired regression matrix can be integrated by parts so that the actual estimate @(t) is only a fune- tion of position feedback. The following development illustrates how the DCAL strategy can be applied along with a filter mechanism to develop an adaptive output feedback controller (ie., only q(t) and hence e(t) are measurable)aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.22 2. Robot Control where 0(t) is generated from the following gradient-based DCAL adaptive update law 6=TY}r. (2.54) By integrating (2.54) by parts, the estimate 0(t) can be expressed as at) = (2.55) From (2.53) and (2.55), the controller does not depend on velocity, and the regression matrix can be computed off-line. Substituting (2.53) into (2.49) yields the closed-loop error system =—Vin(q. d)r + Yu0 — kMr +key —€ +x. (2.56) itate the subsequent stability analysis, let k be defined as the con- 1 b= — (Gila + C3kna + 1) (2.57) m where ¢, and ¢, are the constants defined in (2.51), and ki and kn2 € R ive constants selected (large enough) to satisfy the following initial condition dependent sufficient condition ye! mnax(1,r2) l2(0)[? + Amos) [@rua(O)| 4a min(1, ma, min )) (2.58) 1 kno (1- a ( Than where Amax() denotes the maximum eigenvalue of the argument, and Bynax(0) denotes a known upperbound on the parameter estimate mismatch. Stability Analysis Theorem 2.5 Given the open-loop error system in (2.49). the adaptive controller given in (2.53) and (2.54) ensures semi-global asymptotic track- ing in the sense that e(t) +0 as t-+00 (2.59) provided k is selected according to (2.58). Proof: Let V(t) € R denote the non-negative function 1 1 1 1-7. > Via prTMr + see + sees + 3 rp (2.60)2.4 Adaptive Control Design 23 that can be upper and lower bounded as F2sin(d me Amin Po) IP 0 (2.66) for some positive constant c. The second inequality in (2.66) illustrates a semi-global stability result. That is, if the condition in (2.66) is satisfied, then V (t) will always be smaller or equal to V(0). Therefore, if the condition in (2.66) is satisfied for V (0) then it will be satisfied for all time. Asa result, provided the sufficient condition given in (2.58) is satisfied (i.e., fis selected large enough based on the initial conditions in the system), then (2.60) and (2.66) can be used to conclude that V(t), 2(t),2x(t),¢(t), es(t).r(t), 6(t) € £2. Similar boundedness arguments can now be used to conclude that all closed-loop signals are bounded and that x(t) € £2, and Barbalat’s Lemma given in Lemma A.1 in Appendix A can be used to conchide that 4kpy— min(1,m4,Amin(P—)) Fn: e(t) +0 as t— co. aaa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2.4 Adaptive Control Design 27 Based on Properties 2.4, 2.6-2.10, x(t) can be bounded as IIxll $e + e2 [ell + es Ilr. (2.81) where c; € R, (i = 1, 2,3) are known positive bounding constants. In (2.79), k ER is a positive constant control gain defined, based on the subsequent stability analysis, as hE hy + hpict + knoe (2.82) where fi, kui, kng € R are positive constant gains. Stability Analysis Theorem 2.6 Given the open-loop error system in (2.74). the NN-based controller given in (2.76) ensures global uniformly ultimately bounded sta- bility in the sense that lle $ coexp(-e18) +2 (2.83) where €o,€1,€2 are some positive constants, provided k is selected according to the following sufficient conditions a (2.84) a hn > y Tne Tee 1 is defined in (2.81). Proof: Let V(t) € R denote a non-negative, radially unbounded function defined as lp Lyd mae orig i Via gr Mr + see + str PpW) + serv Tz1V). (2.85) It follows directly from the bounds given in Properties 2.1, 2.7, and 2.8, that V(t) can be upper and lower bounded as Aa llzll? < Vie) < do Ilzll? +6, (2.86) where 1, Az, ¢ € Rare known positive bounding constants, and z(t) € R2" is defined as 5 z=[7? et J". The time derivative of V(t) in (2.85) can be determined as V = rh (gM — Vina. dr +WTe —WG VI + WIG VI +x —kr — e) + eT (r — ae) — tr(WTTy!W) —tr(V7T ZV) (2.87)aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2.5 Task-Space Control and Redundancy 29 A kinematically redundant manipulator is a robotic arm that has more degrees of freedom (DOF) than required to perform an operation in the task-space; hence, these extra degrees of freedom allow the robot manip- ulator to perform more dextrous manipulation and/or provide the robot manipulator system with increased flexibility for the execution of sophis- ticated tasks. Since the dimension (i.e., n) of the link position variables is greater than the dimension (i.e., m) of the task-space variables, the null space of Jacobian matrix has a minimum dimension of n— m. That is, any link velocity in the null space of the manipulator Jacobian will not affect the task-space velocity. This motion of the joints is referred to as self-motion, since it is not observed in the task-space. As stated in [7], [18], [19], and [27], there are generally an infinite number of solutions for the inverse kinematics of a redundant manipulator. Thus, given a desired task-space trajectory, it can be difficult to select a reasonable joint-space trajectory to ensure stability and boundedness of all signals along with satisfying the mechanical constraints such as singularities and obstacle avoidance. ‘This section considers the nonlinear control of kinematically redundant robot manipulators through the development of a computed torque expo- nential link position and sub-task tracking controller. An adaptive full-state feedback controller is also developed that achieves asymptotic link position and sub-task tracking despite parametric uncertainty associated with the dynamic model. The developed controllers do not require the computa- tion of the inverse kinematics and do not place any restrictions on the self-motion of the manipulator; hence, the extra degrees of freedom are available for subtasks (i.c., maintaining manipulability, avoidance of joint limits and obstacle avoidance). The reader is referred to [32] and [33] for more details. 2.5.1 Kinematic Model ‘The end-effector position and orientation in the task-space, denoted by a(t) € R", is defined as follows SQ (2.93) where f(q) € R™ denotes the forward kinematics, and q(t) € R" denote the link position. Based on (2.93), the differential relationships between the end-effector position and the link position variables can be calculated as & =J(q)¢ # = Jair saa ee)aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2.5 Task-Space Control and Redundancy 31 where [, € R"*" denotes the n x n identity matrix, and g(t) € R" is an auxiliary signal that is constructed according to the sub-task control objective (¢.g., joint-limit avoidance, or obstacle avoidance). ‘The subse- quent stability analysis mandates that the sub-task control objective be formulated in such a manner that both g(t) and g(t) are bounded signals. ‘To provide motivation for the definition of the sub-task control objective given by (2.101), take the time derivative of (2.100) and then substitute (2.94) for £(¢) to obtain é = by +a0— a0 — Jq (2.102) where the term ae has been added and subtracted to right-hand side of (2.102) to facilitate the control formulation, and a € R™*™ denotes a diagonal, positive definite gain matrix. Using the properties of the pseudo- inverse of the manipulator Jacobian defined in (2.97), the relationship in (2.102) can be rewritten as —ae + J (J* (ta + ae) + (In — J*J)g — 4)- (2.103) Based on the structure of (2.103) and the subsequent analysis, the filtered tracking error signal, denoted by r(t) € R", is defined as r=I* (éq +0) +(In— Jt Dg = 45 (2.104) hence, the closed-loop task-space position tracking error system can now be written as -ae + Jr. (2.105) In the following control development, the structure of (2.105) is used to ensure that the task-space error and the filtered tracking error defined by 100) and (2.104), respectively, are both regulated. To illustrate how the regulation of the filtered tracking error also ensures regulation of the sub- task tracking error defined by (2.101), the filtered tracking error in (2.104) is pre-multiplied by (I, —J+J) and then the properties given in (2.99) are applied to obtain én =(In-JtJ)r (2.106) where (2.101) was used. From (2.106), if r(t) is regulated then ex(t) is regulated, and hence, the sub-task control can also be achieved. ‘The structure of (2.105) also provides motivation to regulate r(t) in order to regulate e(t). Taking the time derivative of (2.104), pre-multiplying by the inertia matrix M(q), and then substituting (2.1) yields the open loop dynamics Mi =—Var+Yo—r (2.107)aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.2.5 Task-Space Control and Redundancy 33 motion limits the ability to show that q(t) remains bounded; however, all signals in the manipulator kinematics/dynamics and the control remain bounded independent of the boundedness of q(t) because q(2) only appears as the argument of trigonometric functions. The structure of (2.111) and (2.112) indicates that e(t) and r(t) are bounded by an exponential enve- lope, and hence, due to the boundedness of J* (q) and J(q), (2.106) can be used to conclude that y(t) is also bounded by an exponential envelope. . 2.5.4 Adaptive Control Extension The computed torque controller in (2.109) can also be developed as an adaptive controller as r=Yo+Krt+JTe (2.113) where 6(t) € R” denotes the parameter estimate vector that is generated based on the following update law 6=TyYTr (2.114) where Ty € R™” is a constant, positive definite, diagonal gain matrix. After substituting (2.113) into (2.107), the closed loop dynamics for r(t) can be obtained as Mi = -Var+Y6-J'e- Kr. (2.115) where 4(t) = — 6(t) € R” denotes the parameter estimation error. Theorem 2.8 The control law given by (2.113) and (2.114) guarantees global asymptotic task-space and sub-task tracking in the sense that (2.116) lim e(t), en(t Proof: Let V(t) € R denote a non-negative, radially unbounded function defined as letestui s be psig Va = 5e7e+ Sr? Mr+ 50°T5'4. (2.117) After taking the time derivative of (2.117), substituting (2.105), (2.114), and 2.2, and then canceling common and (2.115), using Properties 2. terms, yields Va = el ae — rT Kr, (2.118) The structure of (2.117) and (2.118) indicates that e(t), r(t), A(t) € Loo. All signals remain bounded by noting that 4(t) = —d(t) (i.e., @ is a constant34 References vector) and employing similar arguments to those used in the previous proof. Since all signals are bounded, (2.105) and (2.115) that é(t) and #(t) are bounded (i.e., e(t) and r(t) are uniformly continuous). From the structure of (2.118), standard arguments can be utilized to show that e(1), r(t) € Lo. Since e(t), r(t) € £2 and uniformly continuous, Lemma A.1 in Appendix A can be invoked to conclude that lim, |le(¢)|), |Ir(0)|| = hence, (2.116) follows directly from (2.106). Ill References 1] R. Colbaugh, and K. Glass, “Robust Adaptive Control of Redundant Manipulators,” Journal of Intelligent and Robotic Systems, Vol. 14, pp. 68-88, 1995. Craig, P. Hsu, and 8. Sastry, “Adaptive Control of Mechanical Ma- nipulators,” Proceedings of the IEEE Conference on Robotics and Au- tomation, pp. 190-195, San Francisco, CA, Mar. 1986. 3] W. Dayawansa, W. M. Boothby, and D. L. Elliot, “Global State and Feedback Equivalence of Nonlinear Systems,” Systems and Control Letters, Vol. 6, pp. 229-234, 1985. 4] W. E. Dixon, D. M. Dawson, E. Zergeroglu and A. Behal, Nonlinear Control of Wheeled Mobile Robots, Springer-Verlag London Limited, 2001. W. E. Dixon, A. Behal, D. M. Dawson, and S. Nagarkatti, Nonlin- ear Control of Engineering Systems: A Lyapunov-Based Approach, Birkhauser Boston, 2003. [6] E. G. Gilbert and I. J. Ha, “An Approach to Nonlinear Feedback Con- trol with Applications to Robotics,” IEEE Transactions on Systems, Man, and Cybernetics, Vol. SMC-14, No. 6, pp. 879-884, 1984. 7] P. Hsu, J. Hauser, and S, Sastry, “Dynamic Control of Redundant Manipulators,” Journal of Robotic Systems, Vol. 6, pp. 133-148, 1989. 8] L. R. Hunt, R. Su, and G. Meyer, “Global Transformations of Nonlin- car Systems,” IEEE Transactions on Automatic Control, Vol. 28, pp. 24-31, 1983. 9] T. Kailath, Linear Systems, Englewood Cliffs, NJ: Prentice Hall, 1980. 0] H. K. Khalil, Nonlinear Systems, 3° edition, Prentice Hall, 2002. I20) References 35 . Khatib, “Dynamic Control of Manipulators in Operational Space,” Proceedings of the 6th IFTOMM Congress on Theory of Machines and Mechanisms, pp. 1-10, New Delhi, Dec. 1983. M. Krstié, I. Kanellakopoulos, and P. Kokotovié, Nonlinear and Adap- tive Control Design, New York, NY: John Wiley and Sons, 1995. F. L. Lewis, C. T. Abdallah, and D. M. Dawson, Control of Robot Manipulators, New York: MacMillan Publishing Co., 1993. F. L. Lewis, “Nonlinear Network Structures for Feedback Control,” Asian Journal of Control, Vol. 1, No. 4, pp. 205-228, 1999. F. L. Lewis, J. Campos, and R. Selmic, Neuro-Fuzzy Control of Indus- trial Systems with Actuator Nonlinearities, SIAM, PA, 2002. S. Y. Lim, D. M. Dawson, and K. Anderson, “Re-examining the Nicosia-Tomei Robot Observer-Controller from a Backstepping Per- spective,” IEEE Transactions on Control Systems Technology, Vol. 4, No. 3, pp. 304-310, May 1996 R. Middleton and C. Goodwin, “Adaptive Computed Torque Control for Rigid Link Manipulators,” Systems Control Letters, Vol. 10, pp. 9-16, 1988. Y. Nakamura, Advanced Robotics Redundancy and Optimization, Addison-Wesley Pub. Co., Inc., 1991. D.N. Nenchev, “Redundancy Resolution through Local Optimization: A Review,” Journal of Robotic Systems, Vol. 6, pp. 769-798, 1989. R. Ortega and M. Spong, “Adaptive Motion Control of Rigid Robots: A Tutorial,” Automatica, Vol. 25, No. 6, pp. 877-888, 1989. P. M. Patre, W. MacKunis, C. Makkar, W. E. Dixon, “Asymptotic ‘Tracking for Uncertain Dynamic Systems via a Multilayer NN Feed- forward and RISE Feedback Control Structure,” IEEE Transactions on Control Systems Technology, Vol. 16, No. 2, pp. 373-379, 2008. Z. X. Peng and N. Adachi, “Compliant Motion Control of Kinemati- cally Redundant Manipulators,” [EEE Transactions on Robotics and Automation, Vol. 9, No. 6, pp. 831-837, Dec. 1993. C. R. Rao, and S. K. Mitra, Generalized Inverse of Matrices and Its Applications, New York: Wiley, 1971.36 26] 27] 28) 29) (30 3h [32] 33) References N. Sadegh and R. Horowitz, “Stability and Robustness Analysis of a Class of Adaptive Controllers for Robotic Manipulators,” Interna- tional Journal of Robotic Research, Vol. 9, No. 3, pp. 74-92, June 1990. S. Sastry, and M. Bodson, Adaptive Control: Stability, Convergence, and Robustness, Prentice Hall, Inc: Englewood Cliffs, NJ, 1989. H. Seraji, “Configuration Control of Redundant Manipulators: Theory and Implementation,” IEEE Transactions on Robotics and Automa- tion, Vol. 5, No. 4, pp. 472-490, August 1989. B, Siciliano, “Kinematic Control of Redundant Robot Manipulators: A Tutorial,” Journal of Intelligent and Robotic Systems, Vol. 3, pp. 201-212, 1990. J. J. B. Slotine and W. Li, Applied Nonlinear Control, Prentice Hall, Inc: Englewood Cliffs, NJ, 1991. M. W. Spong and M. Vidyasagar, Robot Dynamics and Control, John Wiley and Sons, Inc: New York, NY, 1989. M. W. Spong and M. Vidyasagar, Robot Dynamics and Control, New York: John Wiley and Sons, Inc., 1989. 'T. Yoshikawa, “Analysis and Control of Robot Manipulators with Re- dundancy,” in Roboties Research — The First International Sympo- sium, MIT Press, Cambridge, MA, pp. 735-747, 1984. E. Zergeroglu, D. M. Dawson, I. Walker, and A. Behal, “Nonlinear Tracking Control of Kinematically Redundant Robot Manipulators,” Proceedings of the American Control Conference, Chicago, IL, pp. 2513-2517, June 2000. E, Zergeroglu, D. M. Dawson, I. Walker, and P, Setlur, “Nonlinear Tracking Control of Kinematically Redundant Robot Manipulators,” IEEE/ASME Transactions on Mechatronics, Vol. 9, No. 1, pp. 129 132, March 2004.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.38 3. Vision-Based Systems The terminology visual servo control refers to the use of information from a camera directly in the feedback loop of a controller. The typical ob- jective of most visual servo controllers is to force a hand-held camera to a Euclidean position defined by a static reference image. Yet, many practical applications require a robotic system to move along a predefined or dynam- ically changing trajectory. For example, a human operator may predefine an image trajectory through a high-level interface, and this trajectory may need to be modified on-the-fly to respond to obstacles moving in and out of the environment. It is also well known that a regulating controller may produce erratic behavior and require excessive initial control torques if the initial error is large. The controllers in Section 3.3 focus on the more gen- eral tracking problem, where a robot end-effector is required to track a prerecorded time-varying reference trajectory. To develop the controller a homography-based visual servoing approach is utilized. The motivation for using this approach is that the visual servo control problem can be incorporated with a Lyapunoy-based control design strategy to overcome many practical and thet al obstacles associated with more traditional, es. Speci problem is that the translation error system is corrupted by an unknown purely image-based approac ically, one of the challenges of this depth-related parameter. By formulating a Lyapunov-based argument, an adaptive update law is developed to actively compensate for the unknown depth parameter. In addition, the presented approach facilitates: i) transla~ tion/rotational control in the full six degree-of-freedom task-space without the requirement of an object model, ii) partial servoing on pixel data that yields improved robustness and increases the likelihood that the centroid jew [85], and iii) the use of an image Jacobian that is only singular for multiples of 27, in contrast to the state-dependent singularities present in the image Jacobians associated with many of the purely image-based controllers. The controllers target of the object remains in the camera field-of- both the fixed camera and the camera-in-hand configurations. The control development for the fixed camera problem is presented in detail, and the camera-in-hand problem is included as an extension. Conventional robotic manipulators are designed as a kinematic chain of rigid links that bend at discrete joints to achieve a desired motion at its end-effector. Continuum robots [92] are robotic manipulators that draw inspiration from biological appendages like elephant trunks and squid ten- tacles, and can bend anywhere along the length of their body. In theory, they have infinite mechanical degrees-of-freedom so that their end-effector can be positioned at a desired location while concurrently satisfying work- space constraints such as tight spaces and the presence of obstacles. How-aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.40 3. Vision-Based Systems Homography-based visual servo control methods can also be used to de- velop relative translation and rotation error systems for the mobile robot regulation and tracking control problems. By using a similar approach as in the previous applications in this chapter, projective geometric relationships are exploited to enable the reconstruction of the Euclidean coordinates of feature points with respect to the mobile robot coordinate frame. By de- composing the homography into separate translation and rotation com- ponents, measurable signals for the orientation and the scaled Euclidean position can be obtained. Full Euclidean reconstruction is not possible due to the lack of an object model and the lack of depth information from the on-board camera to the target; hence, the resulting translation error system is unmeasurable. To accommodate for the lack of depth information, the unknown time-varying depth information is related to a constant depth- related parameter. The closed loop error systems are then constructed us- ing Lyapunoy-based methods including the development of an adaptive estimate for the constant depth related parameter. Both the setpoint and tracking controllers are implemented on an experimental testbed. Details of the testbed are provided along with experimental results that illustrate the performance of the presented controllers. In addition to visual servo control, the recovery of Euclidean coordinates of feature points of a moving object from a sequence of images (i.e., imag) based motion estimation) is a mainstream research problem with signifi cant potential impact for applications such as autonomous vehicle/robotic guidance, navigation, and path planning. Motion estimation bears a close resemblance to the classical problem in computer vision, known as “Struc ture from Motion (SFM),” scone from its 2D projections on a moving camera. In the motion estimation section of this chapter, a unique nonlinear estimation strategy is presented that simultancously estimates the velocity and structure of a moving object which is the determination of 3D structure of a using a single camera. By imposing a persistent excitation condition, the inertial coordinates for all the feature points on an object are determined. A homography-based approach is then utilized to develop the object kine- matics in terms of reconstructed Euclidean information and image-space information for the fixed camera system. The development of object kine- matics requires a priori knowledge of a single geometric length between two feature points on the object. A novel nonlinear integral feedback e: mation method i: then employed to identify the linear and angular velocity of the moving object. Identifying the velocities of the object facilitates the development of a measurable error system that can be used to formulate a nonlinear least squares adaptive update law. AL yapunov-based analy-3.2 Monocular Image-Based Geometry 41 sis is then presented that indicates if a persistent excitation condition is satisfied then the time-varying Euclidean coordinates of each feature point can be determined. While the problem of estimating the motion and Eu- clidean position of features on a moving object is addressed in this chapter by using a fixed camera system, the development can also be recast for the camera-in-hand problem where a moving camera observes stationary objects. That is, by recasting the problem for the camera-in-hand, the de- velopment in this chapter can also be used to address the Simultaneous Localization and Mapping (SLAM) problem [36], where the information gathered from a moving camera is utilized to estimate both the motion of the camera (and hence, the relative position of the vehicle/robot) as well as position of static features in the environment. 3.2 Monocular Image-Based Geometry ‘This section focuses on the image geometry obtained between images taken by a single camera at different points in time and space. The geometry ob- tained from observing four coplanar and non-colinear feature points from a fixed camera is initially described. Euclidean reconstruction of the feature points from image coordinates is then described. The geometry is then ex- tended to the more popular camera-in-hand scenario, where the observed object is stationary and the camera is moving. For both the fixed camera and the camera-in-hand, the relative motion between the viewed object and the camera is encoded through the construction of a homography that re~ lates two spatiotemporal images. Development is provided that summarizes a well known decomposition algorithm that can be used to extract sealed translation and rotation information from the homography construction. Details are also provided that illustrate how a virtual parallax method en- ables the homography-based techniques to be applied for problems where the observed feature points do not lie in a plane. 3.2.1 Fited-Camera Geometry To make the subsequent development more tractable, four feature points located on an object (ie., the end-effector of a robot: manipulator) denoted hy O; Vi = 1, 2, 3, 4 are considered to be coplanar and not colinear (It should be noted that if four coplanar target points are not available, then the subsequent: development can exploit the classic eight-points algorithm [84] with no four of the eight target points being coplanar or the subse- quently described Virtual Parallax method). Based on this assumption,aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.46 3. Vision-Based Systems the camera. After substituting (3.14) into (3.10) and (3.11), the following relationships can be developed pi=0;(AH AM) p} pai = ou; (AHaAM') pt, on —— (3.16) G Ga where G(t) = [9iy(t)], Ga(t) = [gay(t)] Vi,g = 12,8 € RX? denote projective homographies. From the first relationship in (3.16), a set of 12 linearly independent equations given by the 4 target point pairs (pf, p, (¢)) with 3 independent equations per target pair can be used to determine the projective homography up to a scalar multiple (i.e., the product a, (t)G(t) can be determined). From the definition of G(t) given in (3.16), various techniques can then be used (e.g. see [47, 116]) to decompose the Euclidean homography, to obtain a;(t), G(t), H(t), and the rotation and translation signals R(t) and &,(t), and n*. Likewise, by using the target point pairs (pi. pai (t)), the desired Euclidean homography can be decomposed to ob- tain ay;(t), Gu(t), Hy(t), and the desired rotation and translation signals R,(t) and &,,(t). The rotation matrices R(t) and R,(t) can be computed from R(t) and Ra(t) by using (3.5) and the fact that R* is assumed to be known. Hence, R(t), R(t), Ra(t), Ra(t), Za(t), Fna(t), and the depth ratios a; (t) and ag;(t) are all known signals that can be used for control synthesis. 3.2.3 Camera-in-Hand Geometry Based on the development provided for the fixed camera problem in the previous sections, the geometry for the camera-in-hand problem can be de- veloped in a similar manner. Consider the geometric relationships depicted in Figure 3.3, where the camera is held by a robot end-effector (not shown). The coordinate frames F, Fa, and F* depicted in Figure 3.3 are attached to the camera and denote the actual, desired, and reference locations for the camera, respectively. From the geometry between the coordinate frames, m can be related to m,(t) and ma;(t) as mp=rp+Rm~ mai =cya+ Ram}, (3.17) where m,(t), myi(t), and mf now denote the Euclidean coordinates of O; expressed in F, Fy, and F*, respectively. In (3.17), R(t), Ru (t) € $O(3) denote the rotation between F and F* and between Fy and F*, respec- tively, and ry (t), yy (t) € R3 denote translation vectors from F to F* and F, to F* expressed in the coordinates of F and F,, respectively. By utiliz- ing (3.2), (3.7)-(3.9), and a relationship similar to (3.12), the expressions3.2 Monocular Image-Based Geometry 47 in (3.17) can be written as mj = a4 (R+ ayn?) mz (3.18) se 7 mi = agi (Ra + enan*? my (3.19) — He In (3.18) and (3.19), aa (t), ena (t) € R® denote the following scaled trans- lation vectors m= one =, a;(t) and agi(t) are introduced in (3.10) and (3.11), and mj(¢), mai(E), and m’ now denote the normalized Euclidean coordinates of O; expressed in F, Fa, and F*, respectively. Based on the development in (3.17)-(3.19), the Euclidean reconstruction and control formulation can be developed in the same manner as for the fixed camera problem. Specifically, the signals R(t), Ra(t), en(t), era(t), and the depth ratios a(t) and agi(t) can be computed. The error systems for the camera-in-hand problem are defined the same as for the fixed camera problem (i.e., see (3.36)-(3.39)); however, u(t), ug(t), 6(¢), and O4(t) are defined as in (3.45) in terms of R(t) and Ra(t), respectively, for the camera-in-hand problem. 3.2.4 Homography Calculation The previous development was based on the assumption that four cor- responding coplanar but non-collinear feature points could be determined and tracked between images. This section presents methods to estimate the collineation G(¢) and the scaled Euclidean homography by solving a set of linear equations (3.16) obtained from the four corresponding coplanar and non-collinear feature points. Based on the arguments in [60], a transformation is applied to the projec- tive coordinates of the corresponding feature points to improve the accuracy in the estimation of G(t). The transformation matrices, denoted by P(t), P* € R°*%, are defined in terms of the projective coordinates of three of the coplanar non-collinear feature points as Palm m ps] P Ale me wm). (3.20) From (3.16) and (3.20), it is easy to show that. PG=GP*, (3.21)aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.50 3. Vision-Based Systems 3.2.5 Virtual Parallax Method In general, all feature points of interest on the moving object may not be coplanar. In such a case, the virtual parallax method may be used to develop a virtual plane from noncoplanar points. Based on the development in [84], any three feature points on an object may be selected to define the plane x* shown in Figure 3.4. All feature points O; on a plane satisfy (3.16). Consider a feature point O; on the object that is not on the plane 7*. Let us define a virtual feature point Of, on x*, defined at the point of intersection of the vector from the optical center of the camera to O; and the plane 7*. Let p} be the projective image coordinates of the point O; (and O') on the image plane when the object is at the reference position denoted by F*. As shown in Figure 3.4, when the object is viewed from a different pose, resulting from either a motion of the object or a motion of the camera, the actual feature point Oj and the virtual feature point O', projects to p,(t) and p(t), respectively, on the image plane of the camera. For any feature point Oj, both p;(t) and p/,(t) lie on the same epipolar line J; [84] that is given by =i x Phy (3.30) where x denotes the cross product of the two vectors. Since the projective image coordinates of corresponding coplanar feature points satisfy (3.16), then Lj = pj X Gp}. (3.31) Based on the constraint that all epipolar lines meet at the epipole [84], a set of any three non-coplanar feature points can be selected such that the epipolar lines satisfy the constraint |G bbl = 0 (3.32) | py x Gp} pe x Gp, mx Gri | = 0. (3.33) The transformation matrices, denoted by P(i), P* € R°**, and defined in (3.20), are constructed using the image coordinates of the three coplanar feature points selected to define the plane m*. After coordinate transfor mations defined in (3.24) and (3.25), the epipolar constraint of (3.33) now becomes | axGat ax Gay a x Gay 0, (3.34) where G(t) € R3*3 js defined in (3.22). As shown in [84], the set of homo- geneous equations in (3.34) can be written in the form Cia = 0, (3.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.3.3 Visual Servo Tracking 53 can be used to conclude that if ||e,(¢)|| > 0 then p,(¢) > par(t) and the ratio a (t)/oai(t) > 1; hence, (3.14) and the definition of the depth ratios in (3.10) and (3.11) can be used to show that m(t) —* mai(t) and 21(t) > zai(t). Given that m1(t) + mai(t) and z(t] > zai(t), (3.7) and (3.8) can be used to prove that mi(t) + mui(t). To examine if R(t) + Ra(t), the difference between the expressions defined in (3.5) can be determined as (see Lemma B.1 in Appendix B) R— Ra = sin 0 (u],, — sing [ug], + 2sin? = (uj? — 2sin? Saul . (3.41) If |Ic.,(#)|| > 0, then (3.38) and (3.39) can be used to show that u(t)0(t) > ua(t)éa(t) as t > 00, (3.42) which implies that ||u(t)(€) ||? — |lua(t)@a(e)||? as t > 00, and 6? (t) |lu(t)||? — 63(¢) |lua(t)||? as t + ov. (3.43) Since ||u(#)|] = ||ua(t)|| = 1, (3-43) can be used to conclude that O(t) + £04(t) as t + 00. ‘The result in (3.42) indicates that Case 1) u(t) > ua(#) when 6(¢) —> 04(#) (3.44) Case 2) u(t) 4 —ua(t) when 6(t) + —6a(0) « After substituting each case given in (3.44) into (3.41) and then passing the limit, it is clear that R(t) > R,(t). Based on the results that ma(t) > iar(t) and that R(t) > Ra(t), it is clear that m;(t) > jia:(#). A particular solution for 0(¢) and u(t) can be determined as [105] R— RT 0 = cos”? (Gorm - ») (ul, = ia (3.45) where the notation tr(-) denotes the trace of a matrix, and [u],, denotes the 3x3 skew-symmetric expansion of u(t) To develop a tracking control design, it is typical that the desired tra- jectory is used as a feedforward component in the control design. Hence, for a kinematic controller the desired trajectory is required to be at least first order differentiable and at least second order differentiable for a dy- namic level controller. To this end, a sufficiently smooth function (e.g., a54 3. Vision-Based Systems spline function) is used to fit the sequence of target points to generate the desired trajectory pyi(t); hence, it is assumed that pea(t) and Pea(t) are bounded functions of time. From the projective homography introduced in (3.16), pai(t) can be expressed in terms of the a priori known functions cui(t), Halt), Ra(t)and Zpa(t). Since these signals can be obtained from the prerecorded sequence of images, sufficiently smooth functions can also be generated for these signals by fitting a sufficiently smooth spline function to the signals. In practice, the @ priori developed smooth functions ayi(t), R(t}, and £pa(t) can be constructed as bounded functions with bounded time derivatives. Based on the assumption that Rq(t) is a bounded first order differentiable function with a bounded derivative, (3.45) can be used to conclude that ug(t) and @4(t) are bounded first order differentiable fune- tions with a bounded derivative; hence, @4(t) and @,4(t) can be assumed to be bounded. In the subsequent tracking control development, the desired signals jicq(t) and ©4(t) will be used as a feedforward control term. 3.3.2 Control Formulation To develop the open-loop error system for e,(t), the time derivative of (3.38) is determined as (see Lemma B.1 in Appendix B) = LsRive ~ ae (3.46) In (3.46), the Jacobian-like matrix L,,(t) € R°** is defined as te~ She + 1- (3.47) where sine (0(t)) © and we(t) € R¥ denotes the angular velocity of the object expressed in F. By exploiting the fact that u(¢) is a unit vector (i.c., |lu||? = 1), the determinant of L, (¢) can be calculated as [83] 1 we (8) where det (-) signifies the determinant operator. From (3.48), it is clear that L.,(t) is only singular for multiples of 27 (i workspace); therefore, L,, (t) is invertible in the assumed workspace. det (L..) = (3.48) out of the assumedaa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.3.3 Visual Servo Tracking 59 0.4098 —0.8660 —0.1710 R(O) = | —0.6477 0.4098 —0.5997 (3.74) 0.5997 —0.1710 —0.7817 0.9568 = —0.2555 —0.1386 Ra(0) = | —0.2700 -0.9578 —0.0984 (3.75) —0.1077 0.1316 = —0.9854 0.9865 0.0872 —0.1386 0.0738 —0.9924 _—0.0984 | . (3.76) —0.1462 0.0868 —0.9854 The initial pixel coordinates can be computed from (3.72)-(3.76) as ri(0)=[170 182 1]” p2(0)=[ 110 213 1)” p3(0)=[ 138 257 1]” p,(0)=[ 199 224 1] pa (0) = [ 359 307 : pes (0) = { 282 263 1 |" pas(0)=[ 298 324 1)” Tr T pi=[349 319 1] pp=[ 355 256 1] r T pj=[292 251 1] py=[286 314 1]. ‘The time-varying desired image trajectory was then generated by the kine- matics of the target plane where the desired linear and angular velocities were selected as vea(t) = [ O.2sin(t) O.8sin(t) 0 ] [mm/sec] (3.77) wealt) = [0 0 0.52sin(t) ] [rad/ sec). The desired translational trajector rotational trajectory given in Figure 3.5, and the desired depicted in Figure 3.6. The generated desired image trajectory is a continuous function; however, in practice, the image trajec- tory would be discretely represented by a sequence of prerecorded images and would require a data interpolation scheme; hence, a spline function (i.e., the MATLAB spline routine) was utilized to generate a contimious curve to fit the desired image trajectory. For the top two subplots in Fig- ure 3.5, the pixel values obtained from the prerecorded image sequence are denoted by an asterisk (only select data points were included for clarity of illustration), and a cubic pline interpolation that was used to fit the data points is illustrated by a solid line. For the bottom subplot in Figure 3.5 and all the subplots in Figure 3.6, a plus sign denotes reconstructed Eu- clidean values computed using the prerecorded pixel data, and the spline function is illustrated by a solid line.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.3.3 Visual Servo Tracking 61 pe eh Toe} FIGURE 3.6. Desired Rotational Trajectory of the Manipulator End-Effector Generated by a Spline Function to Fit Prerecorded Image Data. “ah ge sts FIGURE 3.7. Error between the Actual Translation Trajectory and the Desired ‘Translation Trajectory given in Figure 3.5 for the Noise-Free Case.62 3. Vision-Based Systems 6 ca Ey 7s %0 mes) FIGURE 3.8. Error between the Actual Rotation Trajectory and the Desired Rotation Trajectory given in Figure 3.6 for the Noise-Free C: i met FIGURE 3.9. Parameter Estimate for 2} for the Noise-Free Case.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.3.4 Continuum Robots 65 we(t) and ve(t) defined in (J Figure 3.18. Another simulation was also performed to test the robustness of the controller with respect to the constant rotation matrix A*. The constant rotation matrix R* in (3.5) is coarsely calibrated as diag{1,—1,—1}. The resulting translational and rotational errors of the target are depicted in Figure 3.19 and Figure 3.20, respectively. )) and (3.53) are depicted in Figure 3.17 and elem FIGURE 3.13. Error between the Actual Translation Trajectory and the Desired ‘Translation Trajectory given in Figure 3.5 for the Noise-Injected Case. 3.4 Continuum Robots This section explores the problem of measuring the shape of a continuum robot manipulator using visual information from a fixed camera. The mo- tion of a set of fictitious planes can be captured by an image of four or more feature points defined at various strategic locations along the body s and the decomposition of a homography relating a reference image of the robot to the actual robot image, the three dimen of the robot. Using expressions for the robot forward kinemati ional shape information can be continuously determined. This information can be used to demonstrate the development of a kinematic controller to regulate the manipulator end- effector to a constant desired position and orientation.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.74 3, Vision-Based Systems Alter utilizing (3.2), the relationship in (3.92) can be expressed as in (3.10) as (3.94) Osi Ay where a;(t) € R is the depth ratio, and H(t) € R'*3 denotes the En- clidean homography between the coordinate systems F, and Ff. The rela~ tionships in (3.14) can be used to rewrite (3.94) in terms of the projective homography as in (3.16). By utilizing the methods described in Section 3.2 (see also [47, 116}), H, (t) can be decomposed into rotational and translational components as in (3.94). Specifically, the rotation matrix R, (t) can be computed from the decomposition of H(t). The rotation matrix R.(t), defining the orientation of the end of the s“” section of the robot relative to the camera fixed frame I, can then be computed from R.(t) by using (3.93) and the fact that Ry is known a priori. Since R.(t) is a rotation matrix between Z and F., it can be viewed as a composition of two rotational transformations; a rotational transformation from frame I to F,_; followed by a second rotational transformation from F,_; to F.. Hence, R3_,(t) in (3.84) can be progressively computed (i.e., the rotation matrix from one section of the robot to the next) as [105] Ri =(Reu)7R, Vs = 1,2.3,4. (3.95) th From (3.83), the joint space variables for the s“” section can hence be determined as ky = % = (3.96) where 1, € R is the known are length of the section and the notation [],,, denotes a matrix clement at row 2 and column y. With the knowledge int variables q(t) as computed from (3.96), Ti of (3.86), and consequently, the Jacobian J(q) of (3.89) can be calculated online. 3.4.3 Task-Space Kinematic Controller ‘The control objective is the regulation of the end-effector of the manipula- tor to a desired position and orientation denoted by x, € R®. This desired configuration of the robot may be available as an image, and the techniqueaa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.aa You have either reached a page that is unavailable for viewing or reached your viewing limit for this book.