Abstract: To address the communication planning problem in complex battlefield environments, a communication planning decision-making model for distributed combat systems is designed. Based on multi-agent deep reinforcement learning algorithm, this model can provide decision support for combat units and communication nodes within the combat system, enabling each combat unit to obtain long-term stable and efficient communication services. Different from the previous methods that adopt a centralized decision-making paradigm, the proposed method models each combat unit and communication nodes in the environment as independent decision-making agents, and adopts a distributed decision-making mode. Each agent can make rapid decisions solely based on its own observation data without communication with other agents. Experimental results show that the proposed method can effectively assist combat units and communication facilities in decision-making, achieving better decision quality compared with existing benchmark algorithms.