Respawning point recommendation by TD-Learning as a content generation of FPS video game-like E-learning

抄録

In this paper, we propose an approach to customize the E-learning of video game-like by trial and error. A virtual training environment is getting to be common in military training, however, it is still underway to use it as a self-learning tool because of a lack of suitable training curricula for each trainee. First Person Shooting game (FPS) environment which is adequate for such the training, but a lot of characters and objects there which can be considered as the customizing point may cause combinatorial problems in traditional approaches. We show our method based on respawning point can present tasks to trainees by reinforcement learning and they can reach the goal faster than other content generation methods.