A Cerebellum-Inspired Control Scheme for Kinematic Control of Redundant Manipulators

Abstract

Due to the redundancy, the kinematic control of redundant manipulators is a knotty issue in the field of robotics. The cerebellar computation sheds a new light on controlling redundant manipulators by simulating the motor learning and coordination in the human brain. This article makes progress along this direction by introducing an echo state network-based cerebellum network to achieve the efficient control of redundant manipulators. First, a Woodbury matrix identity-based cerebellum network (WMICN) is proposed with the online learning ability. Then, a novel control scheme of redundant manipulators is designed on the basis of the proposed WMICN, where the error feedback information of the joint space, as a teaching signal, is leveraged to achieve the real-time and effective control of redundant manipulators. In the end, simulations, experiments, and comparisons with the existing control methods are conducted to verify the effectiveness and superiority of the proposed WMICN.