Robust and model-free link position tracking control for humanoid COMAN with multiple compliant joints

This paper focuses on developing a robust tracking control scheme for a compliant humanoid, COMAN. Passive compliance in COMAN offers many advantages in terms of safe interaction with the environment and users. In the case of light tasks and soft interaction, classical PD motor control is sufficient. However, the compliance poses some challenges in particular tasks when the robot's joints are under large gravitational forces or external disturbances. These scenarios include carrying a heavy load, manipulating heavy objects or going up stairs where the joints need high torques. But low passive stiffness can cause some issues for the robot balance. Hence, an accurate and robust link position tracking controller is proposed for the compliant robot to achieve precise task-requirements and to reject the effect of external disturbances. Thanks to the use of time-delay estimation, the proposed controller is model-independent and fully decentralized. Thus its implementation to high degrees-of-freedom system becomes particularly simple. The proposed control scheme is validated and compared in a full body dynamic simulation of COMAN with 23 DoFs against the existing proportional-derivative controller. Two application studies are presented in this paper. First, the controller performance is assessed on the torso of COMAN. Second, a whole body control is shown for carrying heavy loads during squatting motion.