A Cerebellar Approach to Adaptive Locomotion for Legged Robots

This paper describes a neural learning architecture for control of legged robots inspired by mammalian neurophysiology. Biological studies indicate that the cerebellum is a key part of an adaptive control system which enables mammals to display remarkable limb coordination during locomotion. We present a distributed control system using reinforcement learning methods and mechanisms inspired by the cerebellum. Embedded within a framework of base locomotion controllers, the system is tasked with learning modulatory control signals which optimize gait performance measures. We briefly describe simulation studies in progress for a four-legged robot. 1 Introduction Designing a robust locomotion controller for a legged robot is a difficult engineering problem. Such a gait control system should satisfy multiple criteria spanning foot placement planning, limb motor control, and posture and stability maintenance. Our interest lies in developing learning systems for walking control which optimi..