Practical method for designing robust RBF-MRAC neurocontrollers: The first-order case

The model reference adaptive control (MRAC) system built with radial basis function (RBF) neural network proved to be useful in controlling nonlinear plants. In this paper a novel method is proposed for designing RBF-MRAC neurocontrollers. The resulted RBF network is optimal in minimal norm-product sense, and the method can be used in case of plants' known nonlinear function. The robustness of the resulted controller was investigated with simulations concerning both parameter tolerances of the nonlinearity and abrupt disturbances. In case of unknown nonlinearity a heuristic method is given, which also resulted in a robust solution. The simulation results of the neurocontroller are also presented in the paper in case of several nonlinearities.