Generation of human-like gait adapted to environment based on a kinematic model

Abstract

Human-like gait generation in 3D environment is one of major concerns for virtual characters and humanoid robot studies. Methods used to generate human-like gait include either combining the common components extracted from real human gait or mapping the particular points’ trajectories to the humanoid robots. However, the common components themselves do not represent the gait adopted by human, which is not comprehensible. Moreover, the complex stability problem of robots needs to be considered of mapping the particular points’ trajectories to the humanoid robots. This paper proposed a succinct and fast method to generate human-like gait by establishing a kinematic model with gait features (Forward distance, Lateral distance, Step height and Turning angle). First, a kinematic model of gait and gait features was established by Fourier Analysis and Principal Component Analysis. Then, for any given environment, the human-like gait was generated with the kinematic model. Finally, in order to remove the accumulated error of the generation gait, the gait features were adjusted at each step to get the final human-like gait. The result showed that the generated gait matched the environment well. The kinematic model proposed can be used for gait design and generation of virtual characters and humanoid robots.