Dual-Mode UAV Collaboration in Unknown Environments: A Frontier-Integrated MASAC Approach with Dynamic Role Specialization

Multiple Unmanned Aerial Vehicles (UAVs) systems have shown significant potential in disaster management, particularly in areas such as rapid damage assessment, search and rescue (SAR), and supply delivery. However, current search and rescue operations often suffer from suboptimal area coverage, limited search efficiency, and weak robustness in multi-agent coordination. This research aims to achieve rapid and comprehensive coverage search in unknown environments to ensure timely detection of rescue targets. To this end, we integrate a frontier exploration algorithm into the Multi-Agent Soft Actor-Critic algorithm (MASAC) to guide UAVs towards unexplored areas, thereby improving coverage efficiency. Effective collaboration among UAVs is achieved through distributed information sharing and fixed role assignment. Simulation results demonstrate that the proposed approach offers faster search performance and improved adaptability in both obstacle-free and obstacle scenarios, offering faster search performance and improved adaptability.