Synergistic Multi-UAV Path and Surveillance Planning Using APF-Enhanced PPO

Le résumé fourni par la source

In this paper, a novel approach to multi-UAV mission planning is proposed, combining Proximal Policy Optimization (PPO) with Artificial Potential Fields (APF). The methodology efficiently addresses complex objectives such as collision avoidance in three-dimensional environments, evasion of detected threats, and surveillance of target areas. By optimizing UAV behavior within appropriately defined surveillance zones, determined by flight direction and altitude, the proposed approach achieves both efficient path planning and effective surveillance. To facilitate learning, terrain data based on UAV positions is utilized, while static obstacles and detected enemy threats are mapped into EnemyProfileDB, and APF features are generated through the application of an APF filter. An Altitude Filter further refines altitude information, and a Frame Stacking technique integrates temporal observation data to enhance learning stability. These features, combined with UAV positions, target locations, relative positions, and distances between UAVs, are integrated into Fully Connected (FC) layers as comprehensive observation inputs to optimize the reward function and reinforcement learning states. The approach has been validated in various mission environments, demonstrating its ability to synergistically execute path planning and surveillance tasks across multiple UAVs. Ultimately, this method improves the efficiency and effectiveness of multi-UAV mission planning, highlighting its practical applicability to autonomous UAV systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.