Data-driven Motion-force Control for Acceleration Minimization of Robots

It is tricky to accurately solve the redundancy solutions of redundant robots with uncertain-structure information. Besides, position-force control as a challenging technical problem is significant for redundant robots with impact forces generated by the end-effector especially. Noteworthily, without considering posture maintenance, the end-effector of redundant robots may experience jittery movements and potentially fail in accurately tracking the target. A data-driven motion-force scheme, considering constraints including position-force control, posture maintaining, and physical joint limits, solved by neural dynamics in the acceleration level, is proposed for redundant robots with unknown structure information. Comparisons and simulation experiments are supplied to substantiate the availability and superiority of the proposed date-driven motion-force scheme.