Automatic and dynamic grounding method based on sensor data for agent-based simulation

Agent-based simulation(ABS) is a promising way to reproduce congestion situations in large-scale facilities and to evaluate the effectiveness of various types of policies for congestion avoidance based on individual behavioral model. Real-world grounding for determining model parameters plays important role to build valid ABS for a specific facility. However, to use ABS continuously for daily decision, parameters should be updated because user characteristics of the facility changes daily or longer term. This study provides a novel grounding method that can automatically and dynamically estimate the parameters of a human behavioral model based on sensor data at a certain interval. To evaluate the method, we conduct simulation experiments using an agent based model to analyze congestion situation in a building. The result with the method can perform congestion prediction with higher accuracy as compared with a conventional method.