Learning Approach to the Active Compliance Control of Multi-Arm Robots Coupled through a Flexible Object

Abstract

This paper presents a quasi-static model and a control strategy for N robot arms cooperating through a concerning its compliant behaviour partly unknown flexible object. The control strategy is based on the position/force decomposition of an extended 6N-dimensional space. The strategy includes feedforward and feedback levels. The Feedback level is organized in the form of an active compliance control law. An AMS-based learning approach is used to accommodate the compliance behaviour of the system and utilized as an additional feedforward loop in the control system. The applicability of the control strategy is verified by simulation.