Nonlinear modelling and control of electrically stimulated muscle: A local model network approach (Q4374752)

The authors consider mathematical models which can describe the nonlinear behaviour of human muscle and which are controller oriented. The presented dynamic muscle models are based on force measurements taken from isometric contracting muscles stimulated by irregular pulse trains. Linear models use linear autoregressive moving average methods with exogenous inputs (ARX). The user has to find a suitable sampling period, the optimal dynamical order, the optimal delay parameters and the coefficients (parameter identification) with the help of the least square methods and cross-validation tests. The linear models are often not powerful enough, therefore local nonlinear models are considered. They employ information from the linear models. The pole placement technique is applied locally for each linear model, the resulting set of controllers is integrated into a network using validity functions. The performance of the synthesized feedback controllers and of the robust linear controllers is compared in simulation experiments. Many figures help to understand the technical and biological background.