Humanoid robot walking pattern generation based on five-mass with angular momentum model

Abstract

In this paper, we propose an effective walking pattern generator based on the five-mass with angular momentum model for humanoid robot. The proposed model is composed of conventional three-mass model and conservation law of angular momentum, which describes the relationship between two swing arms and legs, to achieve a quasi human-like walking. We further design a walking pattern compensator to revise the previous pattern that is not efficient for specific walking time period. Through increasing or reducing the responsibility of joints, the compensator works for sharing and distributing redundant time to other period that need more time. The results are confirmed by simulation and experiments on RENBO, the humanoid robot developed in our lab, respectively.