Skip to Main Content
In this paper, the previously proposed neuro-sliding mode controller for SISO systems by the authors is modified for MIMO case. The structure benefits from the power of sliding mode control and nonlinear function approximation ability of the neural networks. The controller is a two layer feed-forward neural network and weight updates are done using backpropagation algorithm. The error function that is introduced to the neural network is such that the states of the system are restricted to belong to a certain manifold in state space. Different from the works done until now, in this work the aim is not calculating the equivalent control but instead finding the control input by just minimizing a certain error function. Simulation results demonstrate the performance of the controller.