Contents I. Introduction
There has been considerable attention over the years on researches using the neural-network-based control technique when solving the control problems. Many authors have suggested the neural networks (NNs) as powerful building blocks for a wide class of complex nonlinear system control strategies when there exists no complete model information or, even, a controlled plant is considered as a “black box” [1]–[3]. A comprehensive survey on neural control can be founded in [4]. The most useful property of NNs in control is their ability to uniformly approximate arbitrary input-output linear or nonlinear mappings. Based on this property, the NN-based controllers have been developed to compensate the effects of nonlinearities and system uncertainties in control systems. For feedforward NN, because all the weights are updated during each learning cycle, the learning is essentially global in nature and slow. As well, the ability of function approximation is sensitive to training data. Thus, the effectiveness of a general multiplayer NN is limited in problems requiring online learning. In the recent developments in [2] and [5]–[7], using adaptive neural-network control, the asymptotic error convergence can be guaranteed. The adaptive neural controllers are based on conventional adaptive control techniques where the NNs are employed to approximate the unknown nonlinear characteristics of the system dynamics. The objective of these controllers is to minimize the trajectory tracking error for the system while the NN parameters are tuned online.
