An LMI approach to robust multimodel controller design for a four-link lightweight manipulator arm

Abstract

An alternative to a complex nonlinear state feedback controller design for robot manipulators is to use a linear state feedback controller based on linearized manipulator models. For a given range of operation, the nonlinear dynamic model of the manipulator is linearized in a number of operating points (OP). New linear matrics inequalities (LMI)-methods can be used to design a linear state feedback controller which achieves good performance simultaneously in all operating points. This approach is applied to a laboratory four-link flexible manipulator. Linear models have been identified experimentally in suitable operating points, and the performance of a robust state feedback controller is illustrated by experimental results and comparison with a pole placement controller.