Decoupling control of robot manipulator with perturbation feedforward compensation

Abstract

A simple decoupling technique which is based on feedforward compensation of dynamic perturbation is proposed. Each joint servo consists of two separate control loops. In the inner loop, a user specified reference model is used to deduce the perturbation torque from the joint acceleration signal. By feeding forward the estimated disturbance torque, the servo system in each joint become linearized and decouple. As the result, the acceleration torque can be controlled directly and the system is robust against variations of parameters. For the outer loop, an error based servo controller enables asymptotic tracking or trajectory. When perfect decoupling is achieved, the input-output relation for each joint becomes linearized and hence the design of the servo controller is made simpler. >