Abstract
This paper describes a human-like musculoskeletal leg robot which achieves pedaling movement. We introduce the concepts of A-A ratio (Agonist-antagonist muscle-pairs ratio) and A-A activity (Agonist-antagonist muscle-pairs activity) to decompose the EMG signals in human movement into several patterns of muscle coordination and to clarify the physical functions of the extracted patterns. Transferring human muscle coordination into a musculoskeletal robot enables us to simply control the human-like robot with highly redundant multi-degrees of freedom. In this paper, we focus on a pedaling task and explain the process of skill transfer including the analysis of human movement and its implementation to a robot.
