Central Pattern Generators based adaptive control for a quadruped robot

Abstract

This paper presents an adaptive control system of central pattern generators (CPG) made of a network of mutually coupled Van Del Pol nonlinear oscillators. The CPG based controller is used to generate stable rhythmical locomotion trajectories. With the sensorial information feedback, a novel adaptive control strategy is implemented to help a quadruped robot with only two rotary Degrees Of Freedom (DOF)s per leg in the sagittal plane to adapt various terrain environments and response external instructions such as up-slope, down-slope, direction turning and preserving. Moreover, it provides an easy mean to manipulate the robot to move forward, backward and hold steps. The effectiveness of the proposed approach is examined with computer simulations.