Efficient Crowd Evacuation Guidance with Multiple Visual Signage Using a Middle-Range Agent Model and Black-Box Optimization

To facilitate efficient crowd evacuations, we propose evacuation guidance systems with multiple visual signs to effectively convey helpful evacuation information to the evacuees. However, conventional signage systems use signs with fixed evacuation information that cannot respond to dynamically changing evacuation environments. Therefore, to address this issue, adaptive evacuation signage systems that alter evacuation information depending on situational changes have been proposed. This study addresses this problem by employing the black-box optimization technique using multi-agent simulation as the objective function. We assume a realistic evacuation environment consisting of central cores with complex aisles and broad surrounding space. Then, we simulate guiding a crowd of evacuees. The two problem settings for the visual evacuation signage assignment are guiding a crowd to a specific target and minimizing the total evacuation time. Our approach generated near-optimal solutions for the first set and a reasonably good solution for the second set.