Task space control considering passive muscle stiffness for redundant robotic arms

Abstract

In service robotics, control systems allowing for skillful manipulation and dexterity constitute one of the most valuable technologies. Recently, control approaches inspired by humans or animals have attracted widespread attention, due to their merit of allowing various tasks to be performed naturally without precisely calculating their behaviors. This work, thus, focuses on the embodiment of a notable control method for a multi-DOF robotic system considering a human physical activity. In contrast to the traditional approaches, in the proposed control, the linear superposition of four control terms is exploited. These consist of joint spring-damping and virtual spring-damper terms in the joint and Cartesian spaces, respectively. Remarkably, the joint spring term is newly designed for the consideration of the simple passive muscle stiffness effect under gravity to guarantee motion repeatability and avoid the problem of ill-posedness. In the experiment, various abilities with respect to position control and compliant behavior are exposed through a real robot. Additional experiments are performed for the verification of the motion repeatability and energy-efficient motion under DOF redundancy.