Real-Time Fire Escape Path Guidance System through YOLO11 and Dynamic A* Algorithm
Year: 2026
Type: article
Abstract: We present a closed-loop fire escape guidance system that integrates YOLO11 flame detection, map-aligned hazard mapping, and event-triggered A* replanning. Corridor-camera detections are converted into grid cells, expanded by a 3 × 3 safety buffer, and updated by a time-dependent four-neighbor fire-spread model. A* recomputes routes to the nearest stairwell when a new flame appears, fire spreads, a restored cell reopens, or the start position changes. The system displays the live hazard map and route and transmits alarms through the ServerChan Turbo API to WeChat (version 3.9.10.19, Tencent, Shenzhen, China). Each alarm contains the detection time, start position, path length, and flame count. On the fixed validation split, the deployment YOLO11m checkpoint achieves 77.34% [email protected]; on a duplicate-aware group-disjoint test set, the final model achieves 72.36%. A single A* call averages 0.407 ± 0.011 ms on an Intel Core i5-14600KF CPU (Intel Corporation, Santa Clara, CA, USA). In a controlled 60 s scenario with 3 s updates, the system finds a fire-free, buffer-clear route at 30.0% of replanning events, and the common-timeline evacuation run reaches the stair region in 15.83 s. Grid maps from 21 × 21 to 81 × 81 are processed successfully.
Source: AI Engineering
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7225822476
FWCI:
Cites:
Cited by:
Related to: