基于蒙特卡洛树搜索和强化学习的近距空战机动决策方法
By constructing a virtual air combat environment and a neural network model, and combining Monte Carlo tree search and reinforcement learning, the problems of high-dimensional state and delayed feedback in autonomous air combat maneuver decision-making of aircraft were solved, and fast, globally optimal maneuver decision-making was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN UNIV
- Filing Date
- 2023-08-28
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for autonomous air combat maneuver decision-making methods for aircraft in high-dimensional state spaces and with long-term delayed feedback suffer from the curse of dimensionality, slow learning, and instability, and it is difficult to find a balance between exploring unknown states and utilizing known states.
A virtual air combat environment is constructed using Monte Carlo tree search and reinforcement learning methods. By combining a policy deep neural network and a value neural network, and collecting experience sample pool data through Selfplay, the policy deep neural network and the value neural network are trained offline to optimize maneuver decisions.
In high-dimensional state spaces, it can quickly find optimal action strategies, improve decision-making speed and global optimality, adapt to environmental uncertainty, and improve the accuracy and efficiency of autonomous maneuver decision-making.
Smart Images
Figure CN120277980B_ABST