Unmanned aerial vehicle air combat confrontation target tracking method based on deep reinforcement learning

Abstract

The invention provides an unmanned aerial vehicle air combat confrontation target tracking method based on deep reinforcement learning, belongs to the technical field of unmanned aerial vehicle air combat confrontation, and solves the technical problem that an own unmanned aerial vehicle cannot rapidly track a target unmanned aerial vehicle in real time in an unknown environment. According to the technical scheme, the method comprises the following steps: S1, carrying out stack type data processing on a UWB positioning system, and providing position data for subsequent tracking control; s2, establishing a target detection algorithm YOLO-V5 to perform target detection on the single target tracking task; s3, designing a reward function, an action space and a state space based on a TD3 algorithm, and establishing an unmanned aerial vehicle target tracking control model; and S4, analyzing and processing data in the unmanned aerial vehicle target tracking process according to the trained unmanned aerial vehicle target tracking control model, and obtaining the effectiveness of the method in air combat confrontation. According to the method, the target unmanned aerial vehicle can be rapidly and accurately tracked, the tracking uncertainty is given, and the robustness of the model is enhanced.

Images