Neural-network-based adaptive control using sliding modes for nonlinear unknown discrete-time systems

Neural-network-based adaptive sliding-mode control methodologies are proposed for the tracking problem of nonlinear discrete-time input-output systems. The unknown dynamics of the system are approximated via radial basis function neural networks. A fixed structure neural network control scheme and a dynamic structure neural network control scheme are developed. The control laws are based on the sliding mode control and simple to implement. The discrete-time adaptive laws for tuning the neural network are presented using the adaptive filtering algorithm with residue upper-bound compensation. Simulation studies of these approaches demonstrate their validity and effectiveness.