Swarm Cooperative Navigation Using Centralized Training and Decentralized Execution

The demand for autonomous UAV swarm operations has been on the rise following success of UAVs in various challenging tasks. Yet conventional control approaches are inadequate coping with scalability, computational requirements, and real-time performance. In this paper, we demonstrate capability emerging multi-agent reinforcement learning (MARL) to successfully efficiently make sequential decisions during collaborative We propose a scalable, real-time, MARL approach navigation where members have arrive at target locations same time. Centralized training decentralized execution (CTDE) used achieve this, combination negative positive is employed reward function. Curriculum facilitate sought performance, especially due high complexity problem which requires extensive exploration. A model that highly resembles respective physical platform proposed framework testing realistic. scalability sizes, speeds, goal positions, environment dimensions, masses showcased (1) load drop-off scenario, (2) formation without requiring any re-training or fine-tuning agents. obtained simulation results proven effectiveness generalizability our cooperative navigation.