Elbow Torque Estimation for Human-Robot Interaction Control

In areas such as rehabilitation and assistance, the use of robotic systems has increased, so several methodologies have been proposed to regulate human-robot interaction. However, despite its adequate performance many times the sensitivity to singularities, the physical limits of the robot actuators and the dependence on user-specific parameters are often overlooked. In this regard, this work proposes an approach to estimate the torque generated during human-robot interaction and test its effectiveness within an impedance controller that regulates this interaction. The torque estimation methodology uses the electromyographic (EMG) signal and the Hill muscle model, and the user-specific parameters are obtained from an optimization process. Experimental tests were performed to validate the estimation algorithm for continuous movements, obtaining correlation coefficients greater than 0.9. In addition, a simulation test was carried out using the estimated torque within an impedance controller to regulate robot-assisted elbow flexion movements, our controller maintaining the impedance error convergence to zero while regulate the human-robot interaction. These results corroborate the effectiveness of our torque estimation and the controller, thus this methodology may be able to improve the design and safety of rehabilitation and assistance robotic systems. As future work, it is planed to perform other tests to optimize the parameters of Hill’s model and prove the performance of the impedance controller under different user conditions.