An adaptive control system for electrically stimulated muscle: A simulation study

A discrete time adaptive controller for functional neuromuscular stimulation has been developed in this paper. The plant model to be controlled in this study is a musculo-skeletal system including a single- joint limb with two antagonist muscles and a load. The muscles were modeled with a discrete coupled three-factor muscle model recently developed by Shue and coworkers [1]. A model-referenced adaptive control algorithm, sequential nonlinear least-squares (SNLS) algorithm, was used to drive the plant system to trace command signals. Two reference models were used for the estimation of system parameters and control signals for the controllers. One has exactly the structure of the plant model, the other one is a simpler three-factor model. The performances of these algorithms were evaluated on the basis of the responses to a (1) square, (2) trapezoidal, (3) stair up, and (4) stair up and down command signals, respectively. Adaptive controller that has a reference model with exactly the same structure as the plant performed quite well in all cases. It showed even better performance than the controller based on a perfect-matching model. The second controller with an uncoupled model also performed well in the simulations, except for a small deviation from the desired position when an abrupt command change has been applied. The robustness of the algorithms were tested by applying different levels of noise to the plant output. Both algorithms were proved to perform well even in the presence of noise up to 10°of standard deviation.