Model-based state space segmentation for finding suitable procedure in game task

When human beings perform planning, it seems that they segment the process to achieve a goal. In this paper, we propose the technique for state space segmentation based on primitive local predictors to predict phenomenon. This techinique achieves temporal and spatial segmentation using the event cognition with detecting a switch of multiple local predictors and the cognition of situation used each local predictor. To evaluate the proposed techinique, we perform a computer simulation using a simple ping pong game. Then, we present that the representation acquired with state space segmentation enables a long time prediction.