EconPapers    
Economics at your fingertips  
 

Coordinating dynamic signage for evacuation guidance: A multi-agent reinforcement learning approach integrating mesoscopic crowd modeling and fire propagation

Chuan-Zhi Thomas Xie, Qihua Chen, Bin Zhu, Eric Wai Ming Lee, Tie-Qiao Tang, Xianfei Yin, Zhilu Yuan and Botao Zhang

Chaos, Solitons & Fractals, 2025, vol. 194, issue C

Abstract: Recurrent fire outbreaks in indoor crowd-gathering facilities, particularly those where pedestrians are unfamiliar with the spatial layout and visibility is limited, present significant challenges to evacuation safety and efficiency. Under such conditions, traditional static signage, which directs pedestrians to the nearest exit without accounting for real-time crowd and fire dynamics, often fails to provide effective guidance. To address these limitations, under the context of dynamic signage, we propose an integrated method capable of offering coordinated, real-time directional guidance of multiple signs, i.e.,: i) the extension of a sub-region-based mesoscopic model (Cellular Transmission Model, CTM) for crowd movement simulation; ii) the adoption of PyroSim to simulate the dynamic propagation of fire and its byproducts; iii) integrating real-time simulation results from i) and ii) into a dynamic environment to optimize signage directions using the Multi-Agent Reinforcement Learning (MARL)-based QMIX algorithm, with multi-objective goals addressing evacuation efficiency, congestion levels, and fire-induced risks simultaneously. Advancements of this paper can be summarized as: i) in terms of environment construction for dynamic crowd evacuation guidance, our approach represents one of the first to integrate sub-regional mesoscopic crowd modeling with dynamic fire propagation simulation. This integration naturally aligns with the granularity of directional guidance, where pedestrians within the same sub-region receive uniform instructions; ii) regarding real-time directional guidance generation for multi-sign, our method extends the discrete MARL algorithm QMIX, which is well-suited for the discrete action space of each sign (i.e., forward, backward, left, right). This extension effectively manages the high-dimensional challenge of coordinating multiple signs simultaneously while optimizing both evacuation efficiency and safety; iii) from the perspective of model application, we demonstrate the effectiveness of our CTM-PyroSim-QMIX framework in a fire evacuation scenario in a real-world karaoke venue, characterized by low visibility and pedestrians' unfamiliarity with the layout. Benchmarking against the traditional static signage approach, we show that the directional guidance generated by our method enhances evacuation efficiency and reduces fire-related and congestion-induced hazards across 10 single and dual fire source cases. Specifically, the maximum improvements observed in evacuation efficiency, fire-related hazards, and congestion-related risks are approximately over 30 %, 50 % and 70 %, respectively.

Keywords: Evacuation; Dynamic signage; Crowd modeling; Fire simulation; Reinforcement learning (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0960077925002590
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:chsofr:v:194:y:2025:i:c:s0960077925002590

DOI: 10.1016/j.chaos.2025.116246

Access Statistics for this article

Chaos, Solitons & Fractals is currently edited by Stefano Boccaletti and Stelios Bekiros

More articles in Chaos, Solitons & Fractals from Elsevier
Bibliographic data for series maintained by Thayer, Thomas R. ().

 
Page updated 2025-04-08
Handle: RePEc:eee:chsofr:v:194:y:2025:i:c:s0960077925002590