Force Estimation Algorithm using Deep Learning and Electromyogram

Abstract

Human-Robot Interaction (HRI) attracts attention from various fields. For successful HRI, robot has to respond appropriately with recognizing the human intention and force is a good index of it. However, force sensor is expensive and system becomes bulky with it. Therefore, force estimation is necessary to overcome these limitations. For estimation, deep learning approach is reasonable as it can extract highly complex and nonlinear model. However, as input data, position and velocity are not enough because of frictional effect. To compensate it, EMG sensor is used as it reflects the human motion. This paper proposes force estimation algorithm using deep learning and electromyogram.