Biologically Inspired Self-Stabilizing Control for Bipedal Robots

Abstract

Despite recent major advances in computational power and control algorithms, the stable and robust control of a bipedal robot is still a challenging issue due to the complexity and high nonlinearity of robot dynamics. To address the issue an efficient and powerful alternative based on a biologically inspired control framework employing neural oscillators is proposed and tested. In a numerical test the virtual force controller combined with the neural oscillator of a humanoid robot generated rhythmic control signals and stable bipedal locomotion when coupled with proper impedance components. The entrainment nature inherent to neural oscillators also achieved stable and robust walking even in the presence of unexpected disturbances, in that the centre of mass (COM) was successfully kept in phase with the zero moment point (ZMP) input trajectory. The efficiency of the proposed control scheme is discussed alongside simulation results.