Networked control of unmanned vehicle using wavelet-based generalized predictive controller

We present a networked control system for nonlinear plants with unknown dynamics using a wavelet neural network (WNN) with feedforward component and generalized predictive controller (GPC). The WNN with feedforward term is used to adaptively model the nonlinear plant using its input and output history. The WNN uses the traditional back propagation algorithm and the feedforward term uses the recursive least-squares error algorithm for training. The model predictive controller predicts the optimal plant input in the presence of sensor and actuator delay. We apply our approach to the control of an unmanned autonomous vehicle and demonstrate its efficacy in the presence of fixed and random network delays. Simulation results show that adding the feedforward component to WNN drastically reduces the number of hidden layer nodes of the WNN and makes it more suited to online identification and control.