Human elbow motor learning skills of varying loads: Proof of internal model generation using joint stiffness estimation

Abstract

This study presents a human elbow motor learning strategy responding to varying loads. Inspired by Kawato’s internal model theory, we suggest hypothesis that human minimize the internal model error by updating the joint stiffness to generate stable and robust motion during repetitive voluntary action with varying weight of load condition. We designed experimental robotics device to verify our hypothesis and the device is capable of precisely measuring human elbow joint stiffness very accurately. The subject was instructed to perform the prescribed elbow motion without notifying the weight of the load for neutral experimental condition and we recorded joint position, perturbation torque of actuator, reaction torque from torque sensor, and mean absolute value (MAV) of the surface EMG (sEMG) in forearm muscles and upper arm muscles as a reference criterion for elbow joint impedance modulation during motor learning. Modified ensemble-based system identification was applied to characterize the dynamic elbow mechanical impedance in transient state of moving loads. Experimental results show that subjects utilized high joint stiffness initially, but it decreases gradually and saturated to the level of 20%~60% of initial value after repetitive motion tests. The degree of saturation of motor learning varied with the weight of loads, this result supports the hypothesis that motor learning reduces joint stiffness by providing accurate internal model.