MAXIMUM-LIKELIHOOD ADAPTIVE NEURAL CONTROLLER

A concept of model-based neural controller is described, which incorpo rates a model of a controlled system into a neural network architectur e. This concept results in efficient learning requiring small amounts of training data. This is due to the fact that all synapse weights are determined by a relatively small number of model parameters, which ca n be learned quickly. The development in this paper is based on the ML ANS neural network developed previously for classification. The curren t paper addresses the development of a model that is generally applica ble to the problem of multidimensional nonlinear control and that is u seful for model-based neural network development. We describe modifica tions to the MLANS architecture, which are needed to incorporate the c ontrol model into the neural network architecture and for the neural d ynamics to converge to the maximum likelihood values of model paramete rs. Then, we present examples comparing the neural network perfomance to that of the classical least squares controller.