Neural Network Learning Applied To The Control Of Unknown Systems

A neural network (NN) learning controller which is capable of improving its performance in the control of a nonlinear plant of unknown dynamics is described in this paper. This learning controller is based on a gradient-free neural network learning algorithm. Compared to previous neural network learning controllers, the proposed controller does not require information about the plant such as sensitivity nor a plant emulator. Therefore, the controller is more robust and requires a much smaller number of neurons to implement the controller. Simulation has been carried out to study the performance of this new controller in comparison with traditional linear controllers. The new controller has shown fast learning and small tracking error in the control of a pendulum.