Learning and control for a compliant computer-controlled manipulator

Abstract

Precise automated manipulation and assembly by means of a computer-controlled manipulator with elastic drives require a model of the compliant structure. A calibration procedure based on a learning technique is used to identify the structural parameters of the model. The force reflecting performances of the cable drives make possible simple continuous active accommodations to adapt the apparent compliance in the task space during fine motions. The damping of oscillations induced by fast transient motions is carried out by means of measurements of cable deformations and their derivatives. Experimental results with a computer-controlled slave telemanipulator show that performance may be improved by adaptive control.