Humanoid walking pattern generation based on the ground reaction force features of human walking

This paper presents a humanoid pattern generation method based on the ground reaction force features of human walking. A human can walk with high power efficiency and compliant landing, which have a close relationship to the ground reaction force. By investigating the ground reaction force of human walking, some features of the ground reaction force are obtained. According to these features, a novel method to design humanoid ZMP trajectories is proposed to mimic human ZMP trajectories. This method can vary ZMP trajectories easily to generate waking patterns with less fluctuations of walking velocity. In addition, a humanoid robot model is presented for alternating support leg smoothly to avoid the contact impact. This model covers the dynamics of both single support phase and double support phase. Finally, the reliability of the proposed methods is verified by dynamic simulation and walk experiment on a real humanoid robot.