Analyzing human skill through control trajectories and motion capture data

Abstract

Extracting and transferring human skills is an important step towards creating an intelligent robot in a cooperative environment. Human skill contains human control skill and human motion skill. Many studies have been done to extract and transfer either human control skill or human motion skill successfully; however, the complete human-to-human skill transfer process (including both human control skill and human motion skill) has not been realized. In this study, the objective is to achieve the complete human-to-human skill transfer process from an expert to a novice. To realize the human control skill extract, synthesis and transfer process, a dynamic simulator is built up to record human operators’ control trajectories. Based on the simulation results, the best control strategy can be obtained. To analyze human motion skill, motion capture equipment is employed to record human operators’ motions, and BvhViewer software is utilized to compute human joint angle values. To make human novices understand human experts’ motions easily, it is absolutely necessary to reproduce and synthesize the experts’ motions. In this case, a kinematic human body simulator is utilized to synthesize the experts’ motions. A ball shooting task is employed to evaluate the effect of the proposed method. The results show that, after training by the synthesized human control skill model and the human motion skill model, the novice’s performance is significantly improved.