Co-Contraction of Antagonist Muscles of Human Limb using Neural Network-based Control

This paper considers co-contraction of antagonist muscles of human limb using neural network (NN)-based control. Some experimental results indicate that the antagonist muscle and agonist muscle work at the same time during the knee extension and flexion, then it is called co-contraction. However, control laws to adjust electrical impulses for the antagonist muscle by neuromuscular electrical stimulation (NMES) are almost not shown in previous researches. This paper proposes that the control of agonist muscle and antagonist muscle is controlled independently. Moreover, NN-based control is applied to control of antagonist muscle, because the learning components is needed for the unknown disturbances. The set-point regulation control and trajectory tracking control are implemented with the no co-contraction and NN-based co-contraction. It is shown that the co-contraction is useful to the convergence of the joint angle by the attenuation of the overshoot.