An Advanced Life Support System (ALSS) achieves life support in the ultimate environment by regenerating and circulating materials in the system. Because of their complexity of material circulation in ALSS, it is difficult to control whole-system material circulations and to handle on abnormal states in the system. Because of this, it is recommended that the material circulation in the system is controlled automatically, and if unpredictable events occur, it is preferable that the system adapts autonomously. We had proposed new autonomous control method based on a hierarchical control, automatic scheduling, and Multi-Agent Reinforcement Learning method, and had applied to a simple ALSS model. We also considered if a number of equipment have a trouble at the same time, how to be reacted autonomously by control system. Especially, using mental model, we constructed the mechanism to determine what order should the problem be handled by control system automatically. Last year, we started to consider how to build mental model automatically, because the model is simple however, there are a lot of challenges. In this year, we blush up the way to construct the mental model automatically. Especially, we focus on how to reduce the calculation amount of mental model creation in more complex circulation model.