Actual machine varification of autonomous coordinated control of legged model with redundant antagonist muscle groups

Abstract

McKibben pneumatic artificial muscle (MPA) is a kind of artificial muscle used in musculoskeletal robots. This actuator contracts and exerts tension like a muscle by applying compressed air. In musculoskeletal models with MPAs, the MPAs are antagonistically placed around joints, and the joints are driven by antagonistic drives, such as in real animals. Therefore, the antagonistic MPAs should cooperate with each other to generate various movements. However, in previous studies, the cooperative patterns for dynamic movements such as walking or jumping were typically found by trial and error. As a method to generate the cooperative patterns autonomously, we have proposed a tension feedback cooperative control law. In our previous study, we applied the control law to a legged model with four muscles (two pairs of antagonist’s muscles) around one joint. We confirmed the autonomous coordination of antagonistic muscle groups through simulations. In this paper, the legged robot was actually developed, and the validity of the simulation results was verified through experiments on the robot. The experimental results confirm that the control law changes the air inputs to the MPAs autonomously and that the coordination of the antagonist muscle groups and the periodic motion are generated autonomously.