Time Varying Controller Design for Robot Manipulator Control

摘要

This report studies the design of time-varying controller for robot manipulator control. Piece-wise constant time-varying controller is shown to be capable of producing desired fast response without causing overshoot. Such response can not be achieved with time-invariant controller. Nonlinear feedback linearization method is used to transform nonlinear robot equation into decoupled sets of second-order linear equations so that time-varying controller can be designed. Stability of time-varying system and time-varying control based on learning algorithm are also presented. This report comprise of the following four parts: 1) Robot Manipulator Controller Design: It was demonstrated that time-varying controllers can produce more desirable performances than those of time-invariant ones; 2) Stability of Time-Varying System: An important observation was made regarding stability in the design of time-varying systems. Implications of magnitude of change, and rate of change of parameter variations on stability are noted; 3) Learning Controller Design: One particular time-varying controller design approach that is of most interest to robot control is the learning control method. This part presents some preliminary results on learning control theory; and 4) A collection of papers on work which are in part supported by this ONR contract.