Master-sleve system based on virtual internal model using Deep Q-Network

Abstract

In the existing master-slave system, the design of the control parameters is performed in the heuristics, and veteran designers and workers are needed. In this paper, we aimed to realize the ideal behavior for each person more easily, devised a method incorporating DeepQ-Network (DQN) in the virtual internal model. As a result of verifying the method using simulation, we could not realize master - slave control. but could show the learning possibility of the model.