Abstract: Unmanned Combat Aerial Vehicle (UCAV) is usually a continuous and multi-round missile attack and defense confrontation process. In the process of evading incoming air-to-air missiles, UCAV should comprehensively consider the impact of maneuvers on the entire air combat confrontation mission, instead of just focusing on security factors. In this paper, the UCAV autonomous evasive maneuver strategy generation method is proposed under the coupling of tactical requirements such as miss distance, energy consumption and terminal situation advantage. The three-dimensional pursuit and escape model of UCAV-missile and the state space, action space and reward function model of UCAV autonomous avoidance are established. LSTM- Dueling DDQN (Long Short-Term Memory-Dueling Double Deep-Q Network) algorithm is proposed for this model. The algorithm fuses Double DQN (Double Deep-Q Network) and Dueling DQN (Dueling Deep-Q Network) network models, and uses LSTM network to extract timing features. Aiming at the demand conflict problem in the local coupling process, the Chebyshev method is constructed to intuitively target the Pareto strategy solution set for different local demand emphasis levels, reflecting the contradiction and coupling of local demands in air combat maneuver avoidance. Simulation experiments and result analysis show that the proposed method has good convergence speed and learning effect, and it is feasible and effective to solve the problem of autonomous evasive maneuvers in air combat for the multiple tactical requirements. The obtained evasive maneuvers can reflect different evasive tactical requirements while ensuring UCAV's own safety.

Key words: Air Combat maneuver, Autonomous Evasive, Tactical Requirements, UCAV, LSTM-Dueling DDQN