Abstract
The path planning of Unmanned Aerial Vehicle (UAV) formations plays a crucial role in mountainous forest monitoring missions. However, path planning is particularly challenging due to steep terrain and dense vegetation, making it difficult to generate optimal flight paths. The goal of UAV formation path planning in forest monitoring is to create safe, feasible flight paths for each UAV, avoiding terrain obstacles and ensuring coordination and safety, ultimately improving the quality of mission accomplishment. This study establishes a mathematical model that incorporates multiple constraints, such as flight distance, collision threats, and path stability, effectively transforming the complex problem of UAV formation path planning into an optimization problem. To address this multi-constraint path planning optimization problem, an Artificial Rabbit Optimization algorithm incorporating Reinforcement Learning and Thermal conduction search strategy (RLTARO) is proposed. The incorporation of multiple strategies aims to improve the balance of exploration and exploitation of the algorithms as well as algorithmic convergence in the face of complex path planning problems. The comprehensive comparison of the RLTARO algorithm with nine advanced algorithms of similar type in the CEC2017 suite demonstrates its outstanding convergence and robustness across various types of optimization problems. The results of path planning experiments conducted on six mountainous forest terrains with varying complexities demonstrate that RLTARO can efficiently and reliably plan flight paths for UAV formations. Furthermore, the Friedman test results from multiple experiments consistently indicate that RLTARO holds significant performance advantages over the comparison algorithms.