Abstract

This thesis proposes an autonomous leader–follower Unmanned Aerial Vehicle (UAV) framework for real-time wildfire detection and suppression. In recent years, wildfires have become increasingly frequent and destructive, posing severe threats to ecosystems, infrastructure, and human life. Rapid detection and precise suppression during the early stages of fire propagation are therefore critical for effective wildfire management. UAVs have emerged as a promising solution due to their high mobility, flexible deployment, and ability to operate in hazardous and inaccessible environments. However, autonomous wildfire suppression using UAVs remains a challenging problem, primarily due to limited payload capacity, the need for accurate fire localization, and the absence of robust strategies for suppressing line-shaped fire fronts, which represent the most common wildfire scenarios.
To address these challenges, this thesis proposes role-based architecture in which a leader UAV is dedicated to fire detection and localization, while one or more follower UAVs perform targeted suppression tasks. The leader UAV employs deep learning–based fire detection combined with synchronized RGB and thermal imaging to achieve reliable fire identification under varying environmental and low-visibility conditions. Detected fire regions are refined and georeferenced, and the resulting fire coordinates are transmitted to a ground station, where optimized suppression paths are generated and assigned to the follower UAVs. Each follower UAV is equipped with a custom-designed dual-tank water-dropping mechanism that enables precise and controlled release of fire retardant along computed firefighting lines. By decoupling detection and suppression tasks, the proposed architecture improves mission endurance, enhances payload utilization, and increases system robustness and scalability overall. The effectiveness of the proposed framework is validated through real-world outdoor flight experiments using a DJI Matrice 300 RTK UAV equipped with an H20T RGB–thermal camera and a custom-designed water tanker system.