Modeling and Online-Identification of Electrically Stimulated Antagonistic Muscles for Horizontal Shoulder Abduction and Adduction

Spagnol P., C. Klauer, F. Previdi, J. Raisch, T. Schauer

A comparison of two different models of a pair of antagonistic muscles for horizontal shoulder abduction and adduction is presented. The proposed models are based on the so called Hill-model: a mechanical framework (inertia, dampers and springs) is used for their development. The models consider as inputs the estimates of the activation states of the muscles based on digital filtering of the evoked electromyogram (eEMG). Model outputs are angular velocity and position. Both models show good results in terms of performance, but they are different in terms of number of parameters that need to be identified and in terms of physical interpretation. One of the models, in fact, describes the muscle as a spring that generates torque by changing its stiffness parameter depending on its activation level. In order to enable adaptive model-based feed-forward and feedback control strategies for angular position/velocity control, an online-identification method based on an Extended Kalman Filter (EKF) is introduced for one of the two models. Simulation and experimental results show the good performance in terms of convergence time and accuracy of the estimation.