Muscle Synergy Model-Based Human Force Estimation and Disturbance Rejection for Assistive Manipulator Control

Abstract One main challenge in developing robot-assisted control systems lies in accurately identifying and estimating the human intention forces during physical interactions with robots, particularly in the presence of unknown disturbances. This study introduces a robotic control design for manipulator-assisted, human-in-the-loop applications under unknown disturbance forces applied to the robotic end-effector. The proposed control design incorporates a muscle synergy model-based neural network approach to predict human forces and motion intentions during human-robot interactions. To mitigate the effects of unknown force disturbances and enable human operators to perform tasks with minimal additional effort, a disturbance observer-based control strategy is implemented. With predicted motion intentions and reaction forces, the robot control system offers effective assistance in human-manipulator tasks. Experimental evaluations with human subjects demonstrate the system's robustness and effectiveness in disturbance rejection design. Furthermore, the comparative analysis with the benchmark controller confirms that the proposed control design significantly improves manipulator-assisted capabilities while effectively reducing human effort.