Variable Impedance Control Based on Impedance Estimation Model with EMG Signals during Extension and Flexion Tasks for a Lower Limb Rehabilitation Robotic System

Abstract

Rehabilitation robotic devices could be used as an effective tool to restore impaired motion functionality. Due to the human-robot cooperative nature of rehabilitation, these systems are expected to be user-oriented i.e., they should be controlled considering users dynamic characteristics. In this article, we proposed a variable impedance control, in which desired impedance of a system was setup to match human joint stiffness estimated from Electromyogram (EMG) signals recorded. Two experiments were performed in this work. The objective of the first experiment (Experiment-1) was to study the relationship between EMG and changing impedance in knee joint extension and flexion tasks. Based on the recorded data, a nonlinear model was proposed to express the relationship between EMG and changing impedance. The results show that Root Mean Square of the EMG signals (RMS-EMG) of target muscles increases, as specified elastic modulus increases for both tasks, but there is a significant difference (p<0.01, t-test) between the extension and flexion task. The second experiment (Experiment-2) was to confirm the effectiveness of the variable impedance control with the motion-dependent models acquired in Experiment-1. Four different control policies were tested, i.e., NA: No Assist; FO: using EMG-impedance model from the Flexion Task; EO: using EMG-impedance graph from the Extension Task; FE: using two EMG-impedance models from Flexion and Extension task correspondently. Results indicate that the proposed control model (FE) achieved a smaller discrepancy (p<0.01, t-test) between desired angle and the reached angle than the control with EO or NA cases. Moreover, a small sum of RMS_EMG from the variable impedance control with motion- dependent models denoted less effort required than the NA (p<0.01, t-test) case or the control with FO (p<0.01, t-test). Results also indicate that the proposed nonlinear and motion-dependent variable impedance control method achieved a smaller angular discrepancy (P<0.05, t-test) than linear variable impedance control.