Technologies for Autonomous Navigation in Unstructured Outdoor Environments A dissertation submitted to the Division of Research and Advanced Studies of the University of Cincinnati in partial fulfillment of the requirements for the degree of DOCTORATE OF PHILOSOPHY In the Department of Mechanical, Industrial and Nuclear Engineering of the College of Engineering 2003 by Souma M. Alhaj Ali B.S. in Industrial Engineering, University of Jordan, 1995 M.S. in Industrial Engineering, University of Jordan, 1998 Committee Chair: Dr. Ernest L. Hall ABSTRACT Robots have been used in manufacturing and service industries to improve productivity, quality, and flexibility. Robots are usually mounted on a fixed plate, or on rails, and can move in a limited manner. The success of robots in these environments encourages the use of mobile robots in other applications where the environments are not structured, such as in outdoor environments. This dissertation presents the development of an autonomous navigation and obstacle avoidance system for a wheeled mobile robot (WMR) operating in an unstructured outdoor environments. The algorithm produces the robot’s path positioned within the road boundaries and avoids any fixed obstacles along the path. The navigation algorithm was developed from a feedforward multilayer neural network. The network used a quasi-Newton backpropagation algorithm for training. Proportional derivative computed-torque, proportional, integral, derivative computed- torque, digital, and adaptive controllers were developed to select suitable control torques for the motors, which cause the robot to follow the desired path from the navigation algorithm. Simulation software permitting easy investigation of alternative architectures was developed by using Matlab and C++. The simulation software for the controllers was developed for two case studies. The first case study is the two-link robot manipulator, and the second is a navigation controller for the WMR. The simulation software for the WMR navigation controller used the Bearcat III dynamic model, developed in this dissertation. Simulation results verify the effectiveness of the navigation algorithm and the controllers. The navigation algorithm was able to produce a path with a small mean square error, compared to the targeted path, which was developed by using an experienced driver. The algorithm also produced acceptable results when tested with different kinds of roads and obstacles. The controllers found suitable control torques, permitting the robot to follow these paths. The digital controller produced the best results. The significance of this work is the development of a dynamic system model and controllers for WMR navigation, rather than robot manipulators, which is a new research area. In addition, the navigation system can be utilized in numerous applications, including various defense, industrial and medical robots. To: my deceased father, my mother, my husband, Ayman, and my sons, Mahmoud and Omar ACKNOWLEDGEMENTS The author wish to express her appreciation and gratitude towards her advisor: Dr. Ernest L. Hall for his support and guidance. His continued guidance has helped the author in directing this dissertation towards achieving valuable contributions. The author conveys her heartiest thanks to the other dissertation committee members: Dr. Richard L. Shell, Dr. Issam A. Minkarah, Dr. Ronald L. Huston, and Dr. Thomas R. Huston for their continued support that enabled the student to achieve the objectives of his research. My gratitude to all who provided me with continued support: my mother, husband, brothers and sisters. Table of Contents Chapter 1: Introduction….………………………………………………………….. 1.1 Introduction………………………………………………………………… 1.2 Research Goals……………………………………………………………... 1.3 Significance………………………………………………………………… 1.4 Contribution and relation to the present state of knowledge in the field…... 1.5 Methodology……………………………………………………………….. Chapter 2: Background….………………………………………………………….. 2.1 Introduction………………………………………………………………… 2.2 Navigation………………………………………………………………….. 2.2.1 Systems and Methods for mobile robot navigation……………………… 2.2.1.1 Odometry and Other Dead-Reckoning Methods………………………. 2.2.1.2 Inertial Navigation………………………………………………............ 2.2.1.3 Active Beacon Navigation Systems…………………………………….. 2.2.1.4 Landmark Navigation…………………………………………………... 2.2.1.5 Map-based Positioning………………………………………………….. 2.2.1.6 GPS……………………………..………………………………………. 2.3 Literature Review...………………………………………………………… 2.3.1 Vision and sensor based navigation..…………………………………….. 2.3.2 Use of fuzzy logic..………………………………………………………. 2.3.3 Use of artificial neural networks………………………………………..... 2.3.4 Use of neural integrated fuzzy controller………………………………… 2.3.5 Map-based navigation..…………………………………………………... 2.3.6 Biological navigation…………………………………………………….. 2.3.7 Some new methods are proposed………………………………………… 2.3.8 Navigation in unstructured environment………………………………..... 2.4 Controllers for mobile robots autonomous navigation…………………….. 2.5 Defense Robot……………………………………………………………… 2.5.1 Remote-controlled robots………………………………………………… 2.5.2 Autonomous Robots……………………………………………………… 2.5.3 Other research trends in defense robot …………………………………... Chapter 3: Robot modeling….……………………………………………………… 3.1 Introduction….……………………………………………………………... 3.2 Robot kinematics….………………………………………………………... 3.2.1 Robot description..……………………………………………………….. 3.2.2 Posture kinematic model…………………………………………………. 3.2.3 Configuration kinematic model..………………………………………… 3.3 Robot dynamics using Lagrange formulation……………………………… 3.3.1 Configuration dynamic model..………………………………………….. 3.3.2 Posture dynamic model..…………………………………………………. 3.4 Robot dynamics using Newton-Euler method……………………………... 3.4.1 Dynamic analysis………………………………………………………… 3.4.2 Calculation of the Pseudo-inverse matrix (Moore-Penrose generalized inverse)…………………………………………………………………… 3.4.3 Properties of the dynamic model………………………………………… 3.4.4 Bearcat III dynamic model……………………………………………….. 3.4.4.1 Calculation of the moment of inertia…………………………………… 3.4.4.2 Calculation of the resultant normal force f …………………………… n Chapter 4: ANN…………………………………………………………………… 4.1 Introduction...………………………………………………………………. 4.2 ANN Mathematical model…………………………………………………. 4.3 ANN learning rules………………………………………………………… 4.3.1 Supervised learning………………………………………………………. 4.3.2 Reinforcement learning…………………………………………………... 4.3.3 Unsupervised learning……………………………………………………. Chapter 5: The navigation system…………………………………………………... 5.1 Introduction………………………………………………………………… 5.2 The sensory devices………………………………………………………... 5.3 Navigation algorithm………………………………………………………. 5.3.1 ANN architecture………………………………………………………… 5.3.1.1 ANN input………………………………………………………………. 5.3.1.2 ANN output……………………………………………………………... 5.3.1.3 ANN layers architecture………………………………………………… 5.3.2 ANN training……………………………………………………………... 5.3.3 Results……………………………………………………………………. 5.3.4 Conclusion……………………………………………………………….. Chapter 6: Dynamic simulation for real time motion control of the robot…………. 6.1 Introduction………………………………………………………………… 6.2 CT controllers……………………………………………………………… 6.2.1 PD CT controller…………………………………………………………. 6.2.2 PID CT controller………………………………………………………… 6.2.3 Simulation for the PD CT controller for the two-link manipulator……… 6.2.3.1 Simulation architecture…………………………………………………. 6.2.3.2 Simulation results……………………………………………………….. 6.2.4 Simulation of the PD CT controller for the WMR navigation…………… 6.2.4.1 Simulation architecture…………………………………………………. 6.2.4.2 Simulation results……………………………………………………….. 6.2.4.3 Conclusion……………………………………………………………… 6.2.5 Simulation for the PID CT controller for the two-link manipulator……... 6.2.5.1 Simulation architecture…………………………………………………. 6.2.5.2 Simulation results……………………………………………………….. 6.2.6 Simulation for the PID CT controller for the WMR navigation…………. 6.2.6.1 Simulation architecture…………………………………………………. 6.2.6.2 Simulation results……………………………………………………….. 6.2.6.3 Conclusion……………………………………………………………… 6.3 Digital controller…………………………………………………………… 6.3.1 Simulation of the digital controller for the two-link manipulator………... 6.3.2 Simulation of the digital controller for the WMR motion……………….. 6.3.2.1 Conclusion……………………………………………………………… 6.4 Approximation-based adaptive controller………………………………….. 6.4.1 The tracking problem…………………………………………………….. 6.4.2 Error Dynamics………………………………………………………….