Control methodology for reaction force at impact based only on the viscoelasticity of muscles

This paper suggests a novel reaction force control in a robot manipulator by position feedback that is motivated by the analysis of human musculoskeletal system. First, we suggest an analogy between a muscle force characteristic and two-degree-of-freedom control input structure so that the complicated human muscle structure can be reflected in robot manipulator control. In order to clarify functions of complicated muscle mechanism, the complicated muscle structure is simplified as a three-pair six-muscle model including the biarticular muscle. Furthermore, pairs of muscles that work always in an antagonistic way are identified with two-degree-of-freedom control input and the equivalence between the force characteristic of each muscle and PD position control is adopted. Based on these suggestions, a robot manipulator which has the biarticular muscle is controlled to simulate the muscle characteristic. By this simulation, the characteristics and roles of the biarticular muscle are clarified; an optimal viscoelasticity design of the biarticular muscle and its effect at the endeffector. This optimal viscoelasticity design can control the reaction force at the endeffector so that it can improve the balance in a certain direction against the external force. This optimal viscoelasticity that can be realized by position feedback control is also analyzed theoretically based on the novel kinematics. The suggested kinematics also reveals the relationship between the direction of reaction force at the foot and the length ratio of the upper link and the lower link.