In order to improve the autonomous penetration capability of aircraft, a method for generating aircraft maneuver avoidance strategies based on deep reinforcement learning is proposed. In order to solve the sparse reward problem of penetration reinforcement learning, the initialization process and sampling method of deep Q Network are improved, and the convergence efficiency of the learning algorithm is improved. Firstly, a penetration threat scene and an offensive and defensive confrontation motion model are established, and the relative motion state of the interceptor and the typical maneuver mode of the balanced flight section are selected as the state space and the action space. In the learning process, the aircraft is used as an agent to perform a reinforcement learning exploration process through confrontation simulations to train the maneuver value network so that it can finally make autonomous decisions based on the threat situation. Finally, the simulation experiment results show that the algorithm proposed in this paper can effectively generate maneuver penetration strategies and support the intelligent game of aircraft.
Research on Evasion Strategy of Aircraft Based on Deep Reinforcement Learning
Lect. Notes Electrical Eng.
International Conference on Genetic and Evolutionary Computing ; 2021 ; Jilin City, China October 21, 2021 - October 23, 2021
2022-01-04
13 pages
Article/Chapter (Book)
Electronic Resource
English
Orbital Multi-Player Pursuit-Evasion Game with Deep Reinforcement Learning
Springer Verlag | 2024
|Orbital Multi-Player Pursuit-Evasion Game with Deep Reinforcement Learning
Springer Verlag | 2024
|