A Nonlinear Controller Based on the Convolutional Neural Networks

Abstract

This paper focuses on developing a nonlinear controller based on the convolutional neural networks to control different plants. It is assumed that the prior knowledge about the plants is very limited and there are only sensory input-output data history of them. The neural network is trained in supervised learning method without having a target controller. As manipulating data are not picture frames, they are preprocessed and concatenated to form adequate frames required by the convolutional neural networks. A convolutional neural network with a simple structure is proposed for the problem. The trained controller is applied to six different linear and nonlinear plants, one of which is inherently unstable and different from the plants utilized in the training process. Furthermore, an important parameter of this unstable plant is considerably changed and the controller performance is analyzed. The simulation results show that the proposed controller can properly control all plants.