A new method on foot rolling trajectory generation for human-like bipedal walking

Abstract

Acquiring human-like locomotion capability with stable, flexible and natural characteristics for humanoid robot is still a great challenge in the area of robotics, and attracts plenty of researchers in past decades. An immediate standpoint is to learn from human behavior, which leads the utilizing of Human Motion Capture Data (HMCD) to be a hot research topic. In this paper, among many most expressive characteristics of human locomotion behavior, foot rolling regarding to ankle movement is focused. Unlike most previous researches where robot ankle trajectory is directly obtained from HMCD, in this research, it is achieved through a constructed transformation function based on human Center of Pressure (COP) movement. The contribution of this research lies in following two aspects. First, through the analysis on the plantar pressure data collected from 32 healthy college students by footscan® system, it is revealed that human COP movement is fairly stable for human natural walking behavior. Second, to achieve a human-like robot ankle movement, a proper transformation function based on robust COP trajectory is introduced through the analysis on entire human foot rolling motion during walking. It is believed that a robust characteristic indicating the invariance of human motion should be the first choice in building human-like robot motion controller. With the obtained ankle trajectory, a human-like robot walking is learned and demonstrated on a humanoid robot PKU-HR5. The performance illustrates that the proposed new method is effective and promising.