This paper deals with the position control of an induction motor using the secondary resistance control method. In the secondary resistance control method, the generated torque is changed by discretely switching the resistance connected to the secondary circuit, thus the positioning control becomes a discrete-valued input control problem of a nonlinear system. For this problem, we try to apply Deep Q-Network and propose to learn a Q-function with an additional integrator in the state, which is

more »

... xtended in terms of the internal model principle to provide robustness against disturbances. We also propose a data expansion method that exploits the symmetry of the system in order to obtain sufficient training data for offline learning. The effectiveness of the proposed method is demonstrated by numerical experiments on a simulator.