ADAPTIVE RECURRENT CEREBELLAR ERROR OBSERVER FOR ROBUST DYNAMIC BIPED WALKING

Helin Wang, Qijun Chen, and Hao Zhang

Keywords

Biped walking, external disturbance uncertainty, sliding mode control, recurrent cerebellar error observer

Abstract

This paper is concerned with the robust and efficient dynamic walking of biped robots under disturbances. The walking system is controlled based on the sliding mode control over a recurrent cerebellar model neural network. Due to instantaneous change of two legs and complex dynamics during the walking process, the robot can be regarded as a non-linear system, which is strong coupling and hybrid. The robot’s dynamic robust walking is turned into the stability analysis problem of a multi-input/multi-output non-linear system with bounded uncertainties. A robust sliding mode controller is designed to make the robot walking system asymptotically stable on the sliding surface in finite time. To estimate and compensate the upper bound of the error, recurrent cerebellar error observer is applied in the proposed control scheme, in which the learning factors of network weight are adjusted adaptively. Finally, the effectiveness and advantages of the proposed schemes are illustrated by some simulations and comparisons.

Important Links:



Go Back