Teleoperation of a Robot Arm in 2D Catching Movements using EMG Signals and a Bio-inspired Motion Law

This paper presents a methodology of robot arm teleoperation, using electromyographic (EMG) signals and a bio-inspired motion law. The methodology is implemented in planar catching movements, in situations that the user reaches and grasps objects lying on a table in front of him. EMG signals from the flexor and extensor muscles of both the elbow and the wrist joint are used to predict the elbow and wrist joint angle. This is done by using two auto-regressive moving average with exogenous output (ARMAX) models, one for each joint. A position tracker is attached in the user's upper arm, before the elbow joint, and is used for the application of the bio-inspired motion law. This law states that the trajectory of the human hand during planar reaching tasks lays on a straight line. Thus, by applying this motion law at the predicted hand trajectory, the errors of the joint angle estimation through the ARMAX model are reduced. The grasping intention of the user, after reaching the target is decoded through a discrimination algorithm based on feature extraction of the EMG signals from the forearm. The experimental results show that the two ARMAX model estimations for the joint angles, in conjunction with the motion law, are able to predict the user's motion with high accuracy, within different target points and various movement velocities