A Combat Decision Support Method Based on OODA and Dynamic Graph Reinforcement Learning

Modern network centric joint combat greatly increases the complexity of war, and the demand for intelligent decision-making is becoming stronger. In this paper, a combat decision support method based on graph neural network and reinforcement learning is proposed. Based on OODA cycle theory, TSDI combat network is constructed, and the characteristics of nodes, OODA loops and network topology are extracted. A graph neural network GMQN which can adapt to dynamic network and dynamic action space is designed as the built-in neural network of reinforcement learning. The global joint action is obtained by combining GMQN and Kuhn-Munkras algorithm. Finally, the effectiveness and advantage of the proposed method are verified on the joint combat simulation platform.