Reducing Redundancy of Musculoskeletal Robot With Convex Hull Vertexes Selection

Abstract

It has been a long-term dream of roboticist to create a robot that has similar behavior and appearance to human. With highly redundant joints and actuators, the musculoskeletal system of human can fulfill movement and operation with high precision, flexibility, and reliability. However, the redundancy of the musculoskeletal system also brings great difficulties to the control and fabrication of a human-like robot. Therefore, we propose an algorithm based on the convex hull theory to effectively reduce the number of redundant muscles. Specifically, taking the strength limitation and state-dependent characteristics of muscles into consideration, by analyzing the torque contribution of muscles for the discrete motion of joints, each muscle can be represented by a feature vector, and the convex hulls are formed by these vectors in high-dimensional space. By applying the proposed algorithm, the muscles corresponding to the vertexes of the convex hull are reserved while the redundant muscles within the convex hull are deleted, and a simplified model can be obtained. Sets of motion experiments demonstrate that the simplified model can achieve high-precision movement as a complete model. Based on the simplified model, a hardware platform of a robotic arm is constructed, which consists of six degrees of freedom and 11 pneumatic muscles.