Neuro-Inspired Motion Control of a Soft Myriapod Robot

Abstract

Centipedes exploit soft structures to move efficiently in complex environments. These abilities have promoted scientists to develop centipede-like robots with soft robot technologies. In this paper, a centipede-like robot constructed by multiple body segments connected in series is designed. Each body segment features a pair of antagonistic artificial muscles and four electro-adhesive feet, which plays the role of the basic motion unit of the centipede-like robot and can be regarded as a soft myriapod robot. To accomplish the desired movement tasks for the soft myriapod robot, this paper proposes a neuro-inspired hierarchical motion control scheme according to the functions of the centipede brain. In the proposed control scheme, a mushroom-body-inspired controller with an error-based learning mechanism is adopted for the soft actuator motion control, and a central-complex-inspired deep reinforcement learning algorithm is designed for the robot motion selection. At last, the effectiveness of the proposed motion control scheme is verified via both numerical studies and experiments.