Behera L., Kumar S., Patchaikani P.K., Nair R.R., Dutta S. Intelligent Control of Robotic Systems

Boca Raton: CRC Press, 2020. — 675 p.This book illustrates basic principles, along with the development of the advanced algorithms, to realize smart robotic systems. It speaks to strategies by which a robot (manipulators, mobile robot, quadrotor) can learn its own kinematics and dynamics from data. In this context, two major issues have been dealt with; namely, stability of the systems and experimental validations. Learning algorithms and techniques as covered in this book easily extend to other robotic systems as well. The book contains MatLAB- based examples and c-codes under robot operating systems (ROS) for experimental validation so that readers can replicate these algorithms in robotics platforms.Half Title.
Title Page.
Copyright Page.Authors.Vision-Based Control.
Kinematic Control of a Redundant Manipulator.
Redundancy Resolution using Null Space of the Pseudo-inverse.
Extended Jacobian Method.
Optimization Based Redundancy Resolution.
Redundancy Resolution with Global Optimization.
Neural Network Based Methods.
Visual Servoing.
Image Based Visual Servoing (IBVS).
Position Based Visual Servoing (PBVS).
-/-D Visual Servoing.
Visual Control of a Redundant Manipulator: Research Issues.
Learning by Demonstration.
DS-Based Motion Learning.
Stability of Nonlinear Systems.
Optimization Techniques.
Genetic Algorithm.
Expectation Maximization for Gaussian Mixture Model.
Composition of the Book.
Part I: Manipulators.
Kinematic and Dynamic Models of Robot Manipulators.
PowerCube Manipulator.
Kinematic Configuration of the Manipulator.
Estimating the Vision Space Motion with Camera Model.
Transformation from Cartesian Space to Vision Space.
The Camera Model.
Computation of Image Feature Velocity in the Vision Space.
Learning-Based Controller Architecture.
Universal Robot (UR ).
Mechatronic Design.
Platform.
End-Effector.
Perception Apparatus.
Kinematic Model.
Barrett Wam Manipulator.
Overview of the System.
Experimental Setup.
Dynamic Modeling.
System Description and Modeling.
State Space Representation.Hand-eye Coordination of a Robotic Arm using KSOM Network.
Kohonen Self Organizing Map.
Competitive Process.
Cooperative Process.
Adaptive Process.
System Identification using KSOM.
Introduction to Learning-Based Inverse Kinematic Control.
The Network.
The Learning Problem.
The Approach.
The Formulation of Cost Function.
Weight Update Laws.
Visual Motor Control of a Redundant Manipulator using KSOM Network.
The Problem.
KSOM with Sub-Clustering in Joint Angle Space.
Network Architecture.
Training Algorithm.
Testing Phase.
Redundancy Resolution.
Tracking a Continuous Trajectory.
Simulation and Results.
Network Architecture and Workspace Dimensions.
Training.
Testing.
Reaching Isolated Target Positions in the Workspace.
Tracking a Straight Line Trajectory.
Tracking an Elliptical Trajectory.
Real-Time Experiment.
Redundant Solutions.
Tracking a Circular and a Straight Line Trajectory.
Multi-Step Movement.Model-based Visual Servoing of a DOF Manipulator.Kinematic Control of a Manipulator.
Kinematic Control of Redundant Manipulator.
Visual Servoing.
Estimating the Vision Space Motion with Camera Model.
Transformation from Cartesian Space to Vision Space.
The Camera Model.
Computation of Image Feature Velocity in the Vision Space.
Kinematic Control of a Manipulator Directly from Vision Space.
Image Moments.
Image Moment Velocity.
A Pinhole Camera Projection.
Image Moment Interaction Matrix.
Experimental Results using a DOF Manipulator.Learning-Based Visual Servoing.Kinematic Control using KSOM.
KSOM Architecture.
KSOM: Weight Update.
Comments on Existing KSOM Based Kinematic Control Schemes.
Problem Definition.
Analysis of Solution Learned Using KSOM.
KSOM: An Estimate of Inverse Jacobian.
Empirical Verification.
Inverse Jacobian Evolution in Learning Phase.
Testing Phase: Inverse Jacobian Estimation at each Operating Zone.
Inference.
KSOM in Closed Loop Visual Servoing.
Stability Analysis.
Redundancy Resolution.
Results.
Learning Inverse Kinematic Relationship using KSOM.
Visual Servoing.
Redundancy Resolution.
Tracking a Straight Line.
Tracking an Elliptical Trajectory.Reinforcement Learning-Based Optimal Redundancy Resolution Directly from the Vision Space.Redundancy Resolution Problem from the Vision Space.
SNAC Based Optimal Redundancy Resolution from Vision Space.
Selection of Cost Function.
Control Challenges.
T-S Fuzzy Model-Based Critic Neural Network for Redundancy Resolution from Vision Space.
Fuzzy Critic Model.
Weight Update Law.
Selection of Fuzzy Zones.
Initialization of the Fuzzy Network Control.
Remark.
KSOM Based Critic Network for Redundancy Resolution from Vision Space.
KSOM Critic Model.
KSOM: Weight Update.
Initialization of KSOM Network Control.
Simulation Results.
T-S Fuzzy Model.
Kohonen’s Self-organizing Map.
Real-Time Experiment.
Tracking Elliptical Trajectory.
T-S Fuzzy Model.
KSOM.
Grasping a Ball with Hand-manipulator Setup.Visual Servoing using an Adaptive Distributed Takagi-Sugeno (T-S) Fuzzy Model.
T-S Fuzzy Model.
Adaptive Distributed T-S Fuzzy PD Controller.
Offline Learning Algorithm.
Online Adaptation Algorithm.
Stability Analysis.
Experimental Results.
Visual Servoing for a Static Target.
Compensation of Model Uncertainties.
Visual Servoing for a Moving Target.
Computational Complexity.Kinematic Control using Single Network Adaptive Critic.Discrete-Time Optimal Control Problem.
Adaptive Critic Based Control.
Training of Action and Critic Network.
Single Network Adaptive Critic (DT-SNAC).
Choice of Critic Network Model.
Costate Vector Modeling with MLN Critic Network.
Costate Vector Modeling with T-S Fuzzy Model-Based Critic Network.
Adaptive Critic Based Optimal Controller Design for Continuous-time Systems.
Continuous-time Single Network Adaptive Critic (CT-SNAC).
Critic Network: Weight Update Law.
Choice of Critic Network.
Critic Network using MLN.
T-S Fuzzy Model-Based Critic Network with Cluster of Local Quadratic Cost Functions.
CT-SNAC.
Discrete-Time Input Affine System Representation of Forward Kinematics.
Modeling the Primary and Additional Tasks as an Integral Cost Function.
Quadratic Cost Minimization (Global Minimum Norm Motion).
Joint Limit Avoidance.
Single Network Adaptive Critic Based Optimal Redundancy Resolution.
T-S Fuzzy Model-Based Critic Network for Closed Loop Positioning Task.
Training Algorithm.
Computational Complexity.
Simulation Results.
Global Minimum Norm Motion.
Joint Limit Avoidance.
Experimental Results.
Global Minimum Norm Motion.
Joint Limit Avoidance.Dynamic Control using Single Network Adaptive Critic.Optimal Control Problem of Continuous Time Nonlinear System.
Linear Quadratic Regulator.
Hamilton-Jacobi-Bellman Equation.
Optimal Control Law for Input Affine System.
Adaptive Critic Concept.
Policy Iteration and SNAC for Unknown Continuous Time Nonlinear Systems.
Policy Iteration Scheme.
Optimal Control Problem of an Unknown Dynamic.
Model Representation and Learning Scheme.
TSK Fuzzy Representation of Nonlinear Dynamics.
Learning Scheme for the TSK Fuzzy Model.
Critic Design and Policy Update.
Construction of Initial Critic Network using Lyapunov Based LMI.
Lyapunov Function.
Conditions for Stabilization.
Design of Fitness Function.
Learning Near-Optimal Controller.
Update of Critic Network.
Fitness Function for PI Based Training.
Examples.
Simulated Model.
Example using Real Robot.Imitation Learning.Dynamic Movement Primitives.
Mathematical Formulations.
Choice of Mean and Variance.
Spatial and Temporal Scaling.
Example.
Motion Encoding using Gaussian Mixture Regression.
SED: Stable Estimator of Dynamical Systems.
Learning Model Parameters.
Log-likelihood Cost.
FuzzStaMP: Fuzzy Controller Regulated Stable Movement Primitives.
Motion Modeling with C-FuzzStaMP.
Fuzzy Lyapunov Function.
Learning Fuzzy Controller Gains.
Design of Fitness Function.
Example.
Motion Modeling with R-FuzzStaMP.
Stability Analysis of the Motion System.
Design of the Fuzzy Controller.
Global Validity and Spatial Scaling.
Examples.
Learning Skills from Heterogeneous Demonstrations.
Stability Analysis.
Asymptotic Stability in the Demonstrated Region.
Ensuring Asymptotic Stability outside Demonstrated Region.
Learning Model Parameters from Demonstrations.
Motion Modeling using GMR.
Motion Modeling using LWPR.
Motion Modeling using e-SVR.
Complete Pipeline.
Spatial Error Calculation.
Examples.
Example of Monotonic and Non-monotonic State Energy.
Example of Multitasking with Single and Multiple Task-equilibrium.Visual Perception.Deep Neural Networks and Artificial Neural Networks.
Neural Networks.
Multi-layer Perceptron.
MLP Implementation using Tensorflow.
Deep Learning Techniques: An Overview.
Convolutional Neural Network (Flow and Training with Back-propogation).
Different Architectures of Convolutional Neural Networks (CNNs).
Examples of Vision-Based Object Detection Techniques.
Automatic Annotation of Object ROI.
Image Acquisition.
Manual Annotation.
Augmentation and Clutter Generation.
Two-class Classification Model using Deep Networks.
Experimental Results and Discussions.
Automatic Segmentation of Objects for Warehouse Automation.
Network Architecture.
Base Network.
Single Shot Detection.
Automatic Generation of Artificial Clutter.
Multi-Class Segmentation using Proposed Network.
Experimental Results.
System Description.
Server.
Ground Truth Generation.
Image Segmentation.Vision-Based Grasping.Model-Based Grasping.
Problem Statement.
Hardware Setup.
Dataset.
Data Augmentation.
Network Architecture and Training.
Axis Assignment.
Grasp Decide Index (GDI).
Final Pose Selection.
Overall Pipeline and Result.
Grasping without Object Models.
Problem Definition.
Proposed Method.
Creating Continuous Surfaces in D Point Cloud.
Finding Graspable Affordances.
Experimental Results.
Performance Measure.
Grasping of Individual Objects.
Grasping Objects in a Clutter.
Computation Time.Warehouse Automation: An Example.Problem Definition.
System Architecture.
The Methods.
System Calibration.
Rack Detection.
Object Recognition.
Grasping.
Motion Planning.
End-Effector Design.
Suction-based End-effector.
Combining Gripping with Suction.
Robot Manipulator Model.
Null Space Optimization.
Inverse Kinematics as a Control Problem.
Damped Least Square Method.
Experimental Results.
Response Time.
Grasping and Suction.
Object Recognition.
Direction for Future Research.Part II: Mobile Robotics.
Introduction to Mobile Robotics and Control.System Model: Nonholonomic Mobile Robots.
Robot Attitude.
Rotation about Roll Axis.
Rotation about Pitch Axis.
Rotation About Yaw Axis.
Composite Rotation.
Coordinate System.
Earth-Centered Earth-Fixed (ECEF) Co-ordinate System.
Control Approaches.
Feedback Linearization.
Backstepping.
Sliding Mode Control.
Conventional SMC.
Terminal SMC.
Nonsingular TSMC (NTSMC).
Fast Nonsingular TSMC (FNTSMC).
Fractional Order SMC (FOSMC).
Higher Order SMC (HOSMC).Multi-robot Formation.Path Planning Schemes.
Multi-Agent Formation Control.
Fast Adaptive Gain NTSMC.
Fast Adaptive Fuzzy NTSMC (FAFNTSMC).
Fault Detection, Isolation and Collision Avoidance Scheme.
Experiments.Event Triggered Multi-Robot Consensus.
Introduction to Event Triggered Control.
Event Triggered Consensus.
Preliminaries.
Sliding Mode-Based Finite Time Consensus.
Event Triggered Sliding Mode-based Consensus Algorithm.
Consensus-based Tracking Control of Nonholonomic Multi-robot Systems.
Experiments.Vision-Based Tracking for a Human Following Mobile Robot.
Visual Tracking: Introduction.
Difficulties in Visual Tracking.
Required Features of Visual Tracking.
Feature Descriptors for Visual Tracking.
Human Tracking Algorithm using SURF Based Dynamic Object Model.
Problem Definition.
Object Model Description.
Maintaining a Template Pool of Descriptors.
The Tracking Algorithm.
Step : Target Initialization.
Step : Object Recognition and Template Pool Update.
Step : Occlusion Detection, Target Window Prediction.
SURF-Based Mean-Shift Algorithm.
Modified Object Model Description.
Modified Tracking Algorithm.
Human Tracking Algorithm with the Detection of Pose Change due to Out-of-plane Rotations.
Problem Definition.
Tracking Algorithm.
Template Initialization.
Tracking.
Scaling and Re-positioning the Tracking Window.
Template Update Module.
Error Recovery Module.
KD-tree Classifier.
Construction of KD-Tree.
Dealing with Pose Change.
Tracker Recovery from Full Occlusions.
Human Tracking Algorithm Based on Optical Flow.
The Template Pool and its Online Update.
Selection of New Templates.
Re-Initialization of Optical Flow Tracker.
Detection of Partial and Full Occlusion.
Visual Servo Controller.
Kinematic Model of the Mobile Robot.
Pinhole Camera Model.
Problem Formulation.
Visual Servo Control Design.
Simulation Results.
Example: Tracking an Object which Moves in a Circular Trajectory.
Experimental Results.
Experimental Results for the Human Tracking Algorithm Based on SURF-based Dynamic Object Model.
Tracking Results.
Human Following Robot.
Discussion on Performance Comparison.
Experimental Evaluation of Human Tracking Algorithm Based on Optical Flow.Exercises.