Gait Data Generation Method for Humanoid Robot Training

Abstract

This paper presents an approach to generate the human gait data applied in the articulated humanoid robot training. Usually, the human body of gait sequence is expressed by a dynamical skeleton that are used to train humanoid robots by a tenacity machine intelligence algorithm. Such a machine intelligence algorithm needs massive data for training. To solve this issue, we formulate the skeletons' shapes as trajectories on the shape space of skeletons. Furthermore, we construct a trajectory manifold with stretching channels, where the gait data generation is formulated as an issue of elements interpolation. Several experiments demonstrate the performance of our method in gait data generation.