An Example of Synthesis on the Neural Networks for the Linear Objects with Multiple Inputs

Abstract

The method of synthesis of the neural controllers using, that is based on the reinforcement learning approaches, is proposed in the paper. The proposed method should be applied to multichannel objects with several control inputs and a single output. Such objects are complex objects from the controller synthesis point of view and the application of the known classical methods particularly of the polynomial synthesis method might not give results. On the other hand, the synthesis of controllers with neural networks allows to regulate such types of objects. An example of the application of the proposed method to the object with two inputs and one output is shown in the paper.