Abstract: Drawing inspiration from the triumph of artificial intelligence in complex board games, we propose a novel game-theoretic framework for optimizing decision-making in high-speed vehicle (HSV) pursuit‒evasion game scenarios. The interaction between HSVs and defensive systems is reframed as a high-stakes game characterized by extreme dynamics, compressed decision windows, and partial observability; this presents computational challenges that mirror the strategic depth of board games like Go. To overcome the limitations of existing methods, we adapt the Monte Carlo tree search (MCTS) algorithm to a continuous domain. The adapted MCTS method can be applied to handle HSV-specific kinematics and interceptor constraints, resulting in a framework that implements a cycle of autonomous detection, online MCTS search, and optimal execution. Furthermore, we posit that MCTS offers distinct advantages over alternative algorithms, particularly in terms of adaptability to dynamic scenarios, real-time performance, and interpretability. Overall, this study establishes MCTS as a rational and promising methodology for advancing autonomous decision-making in high-speed adversarial engagements.

棋盘谋略能否重塑高速对抗的格局？

[1]CaoXY, NingX, WangZ, et al., 2025. Intelligent sequential multi-impulse collision avoidance method for non-cooperative spacecraft based on an improved search tree algorithm. Chinese Journal of Aeronautics, 38(4):103205.

[2]ChenWX, HuYD, GaoCS, et al., 2024. Intercepting hypersonic glide vehicle: progress and prospect of guidance technology. Journal of Astronautics, 45(6):799-814 (in Chinese).

[3]DingYB, YueXK, ChenGS, et al., 2022. Review of control and guidance technology on hypersonic vehicle. Chinese Journal of Aeronautics, 35(7):1-18.

[4]GongHR, GongSP, LiJF, 2020. Pursuit–evasion game for satellites based on continuous thrust reachable domain. IEEE Transactions on Aerospace and Electronic Systems, 56(6):4626-4637.

[5]HuiYL, LiXM, LiangC, et al., 2025. Endo-atmospheric maneuver penetration strategy based on generative adversarial reinforcement learning. Chinese Journal of Aeronautics, 38(4):103200.

[6]JiangW, LyuYX, LiYF, et al., 2022. UAV path planning and collision avoidance in 3D environments based on POMPD and improved grey wolf optimizer. Aerospace Science and Technology, 121:107314.

[7]RivièreB, LathropJ, ChungSJ, 2024. Monte Carlo tree search with spectral expansion for planning with dynamical systems. Science Robotics, 9(97):eado1010.

[8]ShiL, WangXS, ChengYH, 2025. Safe reinforcement learning for zero-sum games of hypersonic flight vehicles. IEEE Transactions on Vehicular Technology, 74(1):191-200.

[9]SilverD, HuangA, MaddisonCJ, et al., 2016. Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587):484-489.

[10]SilverD, HubertT, SchrittwieserJ, et al., 2018. A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play. Science, 362(6419):1140-1144.

[11]SongZL, ChinKW, YangCL, et al., 2024. Methods to assign UAVs for K-coverage and recharging in IoT networks. IEEE Transactions on Mobile Computing, 23(4):2504-2519.

[12]ŚwiechowskiM, GodlewskiK, SawickiB, et al., 2023. Monte Carlo tree search: a review of recent modifications and applications. Artificial Intelligence Review, 56(3):2497-2562.

[13]XuJW, OtaK, DongMX, 2023. Aerial edge computing: flying attitude-aware collaboration for multi-UAV. IEEE Transactions on Mobile Computing, 22(10):5706-5718.

[14]YanT, CaiYL, XuB, 2020. Evasion guidance algorithms for air-breathing hypersonic vehicles in three-player pursuit–evasion games. Chinese Journal of Aeronautics, 33(12):3423-3436.

[15]YuZ, HuoMY, WangYG, et al., 2025. A UAV mission planning method based on improved Monte Carlo tree search. Journal of Astronautics, 46(5):874-883 (in Chinese).

Similar articles