| In recent years, design philosophies based on a mechatronics approach have been explored to create high performance robotic systems. In a mechatronics approach, the mechanical, electrical, and controller designs for the robotic system cannot be separated but must be performed simultaneously. There are significant interactions and intricate relationships between the dynamic behaviors of these components which make up the overall system. Thus, to achieve high performance robotic systems, a comprehensive design methodology, considering these interactions and tradeoffs, is necessary. This thesis focuses on the development of such a design methodology for the simultaneous plant/controller design optimization of a class of serial rigid manipulators and a class of nonlinear feedback controllers for improving various performance measures of the closed-loop system. The development of the design optimization methodology is based on multiobjective evolutionary algorithms due to their potential for handling optimization problems involving mixed continuous/discrete design variables, and multiple dynamic objectives and constraints. Multiobjective optimization problem formulations are incorporated within the proposed methodology to generate multiple Pareto-optimal design solutions for the simultaneous plant/controller design optimization problems. Enhancements and extensions are made to the evolutionary algorithms used in the methodology for solving the design optimization problems that involve multiple design objectives and design variables including topological (or structural) changes to the plant and/or the controller. A class of nonlinear feedback controllers are introduced for the position control of rigid manipulators along with their stability and robustness analysis. A unified and efficient computational scheme for solving the feedback controller optimization problem for rigid manipulators is presented by combining a controller parameterization technique and a global numerical optimization method. The members of a class of stable nonlinear feed-back controllers are parameterized using B-splines and the optimization method to search for the optimal controller parameters is implemented using an evolutionary algorithm. Finally, design optimization examples involving rigid serial manipulators and nonlinear feedback controllers are presented to illustrate the methodology for handling multiple design objectives and constraints along with mixed continuous/discrete design variables. In one example, a fixed topology non-redundant manipulator and a nonlinear feedback controller both consisting of only continuous design variables, are optimally designed for executing point-to-point motions. In another two examples, redundant manipulators and nonlinear controllers are optimized for performing end-effector trajectory tracking tasks. For the optimal design of redundant manipulators, the manipulator topology is included as one of the design variables. The effective use of multiobjective evolutionary algorithms for obtaining Pareto-optimal design solutions is demonstrated by these design examples. In this thesis, numerical simulation results are also presented to show the superior performance of the simultaneous design optimization method over the sequential design optimization method. |
