A neural-network-based adaptive tracking controller design for a class of nonlinear systems

A direct adaptive neural network controller is proposed for a class of nonlinear systems. The only restriction for the plant is that the input-output gain has to be strictly monotonic. An artificial neural network is adapted to identify the inverse mapping between the input and the desired state differential. By incorporating sliding mode control dynamics into this mapping, the output of this neural network is used as a direct tracking command. A compensator is added to prevent system dynamics from staying outside the desired range. Under the assumption that a neural network can represent the nonlinear mapping to a chosen degree of accuracy, the algorithm is proven to be globally stable, with tracking errors converging to a neighborhood of zero. A stable weight adjustment mechanism is determined in terms of Lyapunov theory. A simulation is performed to validate the proposed adaptive control scheme.