On-line neural network control applications

The main objective of this work is to present an exploration of potential applications of online trained neural networks. In particular the work contains computer simulation results obtained from: implementing an internal model neural network-based control scheme, and a model reference neural network-based adaptive control scheme. The design of the online learning algorithm used to adapt the neural networks weights is based on the theory of continuous-time variable structure control systems. Both the selection of the neural network adaptation gain and the parameters of the low-pass filter involved in the scheme have been selected by trial and error. Only stable open-loop systems have been studied.<
>