ComSAC: Multi-agent unmanned aerial vehicle swarm communication collaboration algorithm based on Soft Actor-Critic
Jiawei Hu, Binxue Sun · 2024
This paper proposes a multi-agent communication and coordination framework for unmanned aerial vehicle (UAV) swarms based on the Soft Actor-Critic (SAC) algorithm. The proposed framework incorporates a master-slave communication model that effectively addresses challenges such as non-stationarity, partial observability, and high learning complexity in multi-agent systems. Experimental results in the simulation environment show that, compared to existing multi-agent methods, the proposed framework significantly enhances UAV task efficiency and communication coverage. Specifically, the communication coverage increased by 74.35%, the target search time was reduced by 57.04%, and the rate of redundant searches decreased by 23.21%. Particularly in complex environments and multi-target tasks, the proposed framework demonstrates exceptional stability and adaptability, providing a novel solution for efficient collaboration of UAV swarms in real-world applications.