A Human-Like Control for Sensorimotor Multi- Tasking in Humanoid Robots

This paper introduces a human-derived humanoid control concept that allows voluntary movements with automatic posture control. The concept consists of a modular control architecture, where each DOF is served by one control module. The control modules are based on the DEC model (DEC, Disturbance Estimation and Compensation). The DEC model combines sensory-derived disturbance estimates for re-active disturbance compensation with predicted-sensory disturbance estimates for compensation of self-produced disturbances. The control modules are interconnected, forming a network whose functionality dynamically reflects the robot’s current state of kinematic activity and its interaction with the external forces that are acting on it. By this, the DEC control allows superposition of several postural and voluntary activities. Keywords—Humanoid; Multi-Task Control; DEC model