Maneuvering decision in short range air combat for Unmanned Combat Aerial Vehicles

In order to the requirement of flexibility and real-time for maneuvering decision in short range air combat, a maneuvering decision model in short range air combat based on reinfbrcement learning is presented. The model has two characteristics, one is select the situational evaluation function to build the state space, what make the discrete state space get more representative. The other is use the thinking of Monte-Carlo reinforcement learning, the result of air combat was used as the basis for returning the reward, what ensure the continuity of maneuver movement. The model is use the control valve as the decision result and the decision time is less than 0.001 seconds. At last, the feasibility of the decision model was verified by two different experiments.