Mohammad Hasan Olyaei Torqabeh - Faculty of Electrical Engineering, Sadjad University of Technology, Mashhad, Iran
Hasan Jalali - Faculty of Electrical Engineering, Sadjad University of Technology, Mashhad, Iran
Ali Olyaei Torqabeh - Department of Computer Engineering, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, Iran
Amin Noori - Faculty of Electrical Engineering, Sadjad University of Technology, Mashhad, Iran

This paper discusses solving the Grid World with Changing Obstacles (GWCO) problem with the Deep Reinforcement Learning (Deep RL) method. In the GWCO problem, obstacles move on specific paths. Moving these obstacles turns this problem into a dynamic problem. Due to the changing environment of the problem and the high number of state-action, the Deep RL method is used to solve the GWCO problem. In this paper, we refer to the methods of Reinforcement Learning, Deep Learning, and Deep RL, and some of their applications. In the final section, by comparing the three types of learning rate (α), the simulation results are compared and it can be concluded that for the GWCO problem, the learning rate is better set to 0.001. Simulation of this paper is done with the powerful Python software and Tensorflow

Grid World, Deep Reinforcement Learning, Deep Learning, RL, Python, Tensorflow