Research on Autonomous Behavior Decision of UAV Cluster with Limited Communication Bandwidth

Abstract Communication is the basis of UAV(Unmanned Aerial Vehicle,UAV) cluster cooperative decision control. In the battlefield environment with scarce communication resources, it is of great significance to realize the efficient communication of UAV cluster. At present, Multi-Agent Reinforcement learning method is widely used in the research of Autonomous Cooperation Strategy of UAV cluster, but most methods do not consider the impact of limited communication resources. Based on the partial observable Markov decision process, a communication based autonomous behavior decision model of UAV cluster is proposed; Based on the research hypothesis and information theory, a reward strategy based on bandwidth utilization is designed. The reward value under this strategy is adjusted with the change of available bandwidth, which can guide the UAV whether to send messages at each step in real time; In the process of model training and execution, the long-term and short-term memory network is used as the memory unit to ensure the effectiveness of UAV cooperative decision-making. The effectiveness of the proposed method is verified by air confrontation simulation experiment, which provides a method reference for autonomous cooperative decision-making of UAV cluster in bandwidth limited combat scenario.