Modal Control of Fast Large-Scale Robot Motions

Abstract

This study concentrates on the following topics in linear state-feedback robotic control: an algorithm for the generation of linearized robot models, a control law providing a desired eigenstructure for the linearized models, and the eigenvalue sensitivity to changes of the linearized model parameters. The algorithm allows the computer generation of linearized dynamic models for any articulated mechanism with revolute or prismatic joints. It does not include numerical differentiation and is based on a compound-vector technique and Newton-Euler dynamics. The control law allows the arbitrary assignment of all eigenvalues and certain entries of the closed-loop eigenvectors. The general structure of the closed-loop modal matrix and the flexibility available in eigenvector assignment are considered. A sensitivity analysis is given for the decoupled control law resulting from a particular eigenvector assignment. An experimental vertion of the developed modal controller was implemented on a multiprocessor system based on Motorola 68020 microprocessors. Details of the implementation and results of robot motion simulation are also included.