A Real-time Control Method for Humanoid Robot to Walk Stably on Uneven Ground

Abstract

To solve the problem that a humanoid robot is prone to tip over while walking on uneven ground, this paper presents an online regulating control algorithm based on the Kane collision theory. The algorithm, which introduces a new physical quantity named "generalized speed", can reduce second-order differential equation to first-order ordinary differential equation, and shorten the time of calculation. We adopt the offline gait pattern generated by Lagrange dynamic model to control the humanoid walking on level ground. When the robot lands on uneven ground surfaces, we first calculate the foot impact force using the Kane regulating algorithm and then use this information to adjust the gait of the legs and the torso in order to ensure a stable walk on the uneven ground. The effectiveness of the proposed algorithm was verified by walking simulations on a 5-link humanoid model under the Matlab/Simulink environment