SwarmSAR-FW: Drone-Based Swarm Search and Rescue in Fire and Water Disasters

Abstract

Fire and water disasters impose critical constraints on search-and-rescue (SAR) operations through rapidly evolving hazards, degraded visibility, and narrow intervention windows. Conventional SAR operations that rely on human operators and single-UAV approaches have limited scalability under such conditions. To address this, our research introduces SwarmSAR-FW, a coordinated multi-drone framework for integrated fire and flood disaster response that employs cooperative surveying, dynamic task allocation, and real-time communication to enable parallel coverage and adaptive deployment. In aquatic scenarios, the system integrates lightweight wearable devices that provide victim localization, physiological monitoring, drift prediction, and priority-based triage. Autonomous life-jacket deployment achieved visual detection within 2.1 s and a mean delivery accuracy of 1.8 m. The prioritization algorithm correctly ranked critical victims in 98.7% of trials, while the swarm-based task reassignment introduced 4.1 s of coordination latency and 25.4 s of end-to-end assistance time when tested with two physical and one simulated drone. For fire environments, the framework supports multi-altitude detection, panoramic reconstruction, safe access-point identification, and temporally weighted fire localization through aggregated visual observations. Experiments with physical drones using controlled visual injection demonstrate reliable fire-region identification and robust entry/exit estimation in dense smoke. SwarmSAR-FW offers a scalable SAR solution that improves response time, prioritization accuracy, and responder safety relative to single-UAV deployments, thereby addressing fundamental limitations in conventional disaster-response paradigms.