NARMAX_OLS Representation of a Semi-active Dynamic Leg Joint Model for a Paraplegic Subject Using Functional Electrical Stimulation

A nonlinear autoregressive moving average with exogenous input (NARMAX) model structure is used to develop paraplegic dynamic leg joint model and the result is compared to a model built using artificial neural network (ANN). A series of experiments employing Functional Electrical Stimulation (FES) with varying pulse width are conducted to determine the impact of the leg swing angle. The data obtained is used to develop the paraplegic leg joint model. Seven sets of data with 3300 data in each are used to develop the joint model. The joint model thus developed is validated by experimental data from one paraplegic subject. It is revealed that ANN and NARMAX models both can represent the joint dynamics but NARMAX joint model is found to be highly accurate and able to predict the leg movement with better precision as compared to the ANN joint model. In order to improve the performance and the precision of the model, NARMAX is used with orthogonal least square (OLS) algorithm. The established model is then used to predict the behaviour of the underlying system and will be employed in future for the design and evaluation of various control strategies.