An Air Combat UCAV Autonomous Maneuver Decision Method Based on LSTM Network and MCDTS

This paper studies the integrated process of air combat UCAV action recognition and autonomous maneuver decision-making in combination with the aerial battlefield situation. The key points of this paper are recognizing the maneuver action of UCAV under limited information sets and generating the optimal maneuver strategies for autonomous decision-making when confronting an intelligent enemy. By introducing a large number of air combat data sets to train the LSTM network, and inputting the real-time information of the enemy into the network, the maneuver type can be recognized as auxiliary information for the Monte Carlo double tree search (MCDTS) autonomous decision process. The UCT function of the search tree is designed as a combination of the game process reward and the air combat result reward, which can guide the decision tree to approach the optimal maneuver strategies more reasonably and faster. The simulation results of air combat show that the strategy is feasible when confronting an intelligent enemy.