[4J2-GS-5] Agents:

[4J2-GS-5-04] Mitigating Adverse Effects of System Failures on Dynamic Distributed Evacuation Guidance System

〇Akira Tsurushima¹ (1. SECOM CO., LTD.)

Keywords:Multi-agent simulation, Distributed problem solving, Evacuation guidance, Resilience

Evacuation environments, such as those during fires, are harsh for most systems, and evacuation support systems themselves may not always function properly. To address this issue, we are researching a dynamic evacuation guidance system using distributed processing. This system enables evacuation signs installed throughout a building to independently communicate via peer-to-peer connections, providing efficient evacuation guidance as a whole system. Since it does not have a single point of failure, the system can maintain functionality even if some components fail. Evaluation using multi-agent simulation showed that in scenarios where components fail in proportion to the intensity of the fire, this system provides evacuation guidance almost equivalent to scenarios with no failures. On the other hand, in scenarios where failures occur randomly, the performance of evacuation guidance decreases. This study addresses this issue by introducing two strategies: spatial interpolation and persistent guidance, demonstrating that the system can perform almost equivalently even in random failure scenarios.