Motion Control of a Snake Robot via Cerebellum-inspired Learning Control

This paper explores the body motion control of a biomimetic snake robot via a cerebellum-inspired control scheme. In the scheme, two corrective actions are conducted, namely adaptive compensation and feedback correction. The former adopts a cerebellar model articulation controller (C- MAC) to compute the inverse dynamics of the robot via online learning and the latter employs a proportional-derivative (PD) controller to provide correction terms to the desired motion. By virtue of the universal nonlinear approximation ability and the rapid learning response rate, the proposed CMAC is able to compensate the model uncertainties and thus enhance the adaptability of the control system. Moreover, the PD controller is utilized to expedite the converge speed through a regular feedback mechanism. In the end of paper, simulation studies in different friction cases are offered to validate the effectiveness of the proposed motion control approach.