Integrated structure/control design of mechatronic systems using a recursive experimental optimization method

Abstract

A recursive experimental design method for simultaneously optimizing both mechanical structure and control is presented in this paper. Control gains are optimally tuned for a given prototype of mechatronic system, and its mechanical structure is physically modified so that control performance can be further improved. The entire control tuning, evaluation, and structure alteration process is repeated until the desired performance goals are achieved. In each iteration, incremental design changes are determined based on a sensitivity Jacobian relating structural changes to performance improvements. The sensitivity Jacobian is updated during the recursive process using the actual data of the design changes. To estimate the Jacobian despite few data points, a pseudoinverse method and a parameter perturbation algorithm, termed singular-value excitation, are developed. The mechanical structure is modified recursively and quickly by using structure reinforcement and rapid prototyping techniques. The feasibility of mechanical structure changes is taken into account in determining the design changes. The proposed methodology is verified through simulation and applied to the design of a robot positioning system. The robot arm structure is modified with regard to stiffness and damping by using steel reinforced epoxy. PD control gains are optimized every time the structure is modified. Through the recursive procedure, an optimal combination of the arm structure and control gains is obtained.