Loitering Munition Interception Decision-Making Technology Based on Deep Reinforcement Learning

The loitering munition is a weapon system that integrates drone technology with ammunition technology, capable of conducting extended autonomous cruising, reconnaissance, identification, and fire strike missions. It finds extensive applications in modern unmanned warfare. The main interception method for such unmanned aerial attack weapons is missile interception. However, there is a large cost disparity between missiles and loitering munitions. To achieve low-cost interception, utilizing unmanned aerial vehicles for interception has become a major research direction. In order to enhance interception effectiveness and achieve intelligent interception, this paper constructs a three-dimensional interception scenario using loitering munition as the interception method. Employing the Deep Q-Network (DQN) algorithm, it trains the intelligent decision-making capabilities of loitering munition. The paper designs neural networks and reward functions to train the maneuver strategic decision model and tests it against scenarios involving various enemy evasion strategies. The results indicate that the trained model possesses interception capability, enabling it to adjust its maneuvering during flight to track and intercept targets.