Abstract:
Wildfires are among the most dangerous and least predictable natural hazards, necessitating continuous real-time monitoring of the advancing fire front. Traditional assets – such as satellite imagery or ground observation posts – often lack the responsiveness and completeness required for timely decision-making. A promising alternative is the use of a swarm of unmanned aerial vehicles (UAVs); however, effective coordination under dynamic conditions and constrained communication and computational resources calls for dedicated control algorithms. This work presents a multi-layer control strategy for a UAV swarm that integrates three components. Horizontal navigation is governed by the Artificial Potential Field (APF) method, which attracts agents to the fire-front line while repelling them from obstacles and neighboring vehicles. A distributed consensus protocol stabilizes flight at a common reference altitude, ensuring comparable viewing geometry. To achieve an even distribution along the front, a modified Particle Swarm Optimization (PSO) layer is employed, reducing competition between drones and unnecessary maneuvering. We report a series of computational experiments comparing classical APF with the hybrid APF+PSO scheme for minimizing a swarm motion performance functional. The hybrid approach lowers the objective by nearly an order of magnitude relative to APF alone, reduces behavioral variability, maintains a coordinated altitude without pronounced overshoot, and ensures reliable obstacle avoidance in the presence of a moving front. Minimum inter-drone separations did not fall below the 5-m safety threshold, confirming compliance with safety requirements. The algorithm sustains stable tracking of the moving front over the entire simulation horizon while preserving correct obstacle clearance. Overall, the proposed strategy combines computational simplicity with high reliability. Future work will incorporate onboard sensing (video and thermal cameras), modeling of wind and terrain effects, analysis of communication delays and losses, and extension of the approach to larger UAV teams.
Fulltext:
PDF file (1026 kB)