Multi-Target Optimization Strategy for Unmanned Aerial Vehicle Formation in Forest Fire Monitoring Based on Deep Q-Network Algorithm

Abstract

Forest fires often pose serious hazards, and the timely monitoring and extinguishing of residual forest fires using unmanned aerial vehicles (UAVs) can prevent re-ignition and mitigate the damage caused. Due to the urgency of forest fires, drones need to respond quickly during firefighting operations, while traditional drone formation deployment requires a significant amount of time. This paper proposes a pure azimuth passive positioning strategy for circular UAV formations and utilizes the Deep Q-Network (DQN) algorithm to effectively adjust the formation within a short timeframe. Initially, a passive positioning model for UAVs based on the relationships between the sides and angles of a triangle is established, with the closest point to the ideal position being selected as the position for the UAV to be located. Subsequently, a multi-target optimization model is developed, considering 10 UAVs as an example, with the objective of minimizing the number of adjustments while minimizing the deviation between the ideal and adjusted UAV positions. The DQN algorithm is employed to solve and design experiments for validation, demonstrating that the deviation between the UAV positions and the ideal positions, as well as the number of adjustments, are within acceptable ranges. In comparison to genetic algorithms, it saves approximately 120 s.