基于深度学习与动力学约束的高速飞行器轨迹预测方法

Trajectory Prediction Method for High-speed Vehicles Based on Deep Learning and Dynamic Constraints

摘要

为满足高速飞行器拦截任务中对高动态、强非线性弹道实时预报的需求,针对传统轨迹预测方法在物理约束缺失、长时序特征捕捉不足及泛化能力弱等方面的局限,提出了一种融合深度学习特征表征与动力学约束的混合驱动预测方法。首先,构建了高速飞行器再入段三自由度动力学模型,并分析气动参数、倾侧角对飞行状态演化的影响;其次,设计了一种基于卷积-双向长短期记忆-自注意力机制的特征表征网络,用于提取飞行状态的时序特征和相关性;最后,将网络预测得到的参数作为动力学模型的驱动项,构建了“参数预测-状态更新-轨迹生成”的闭环递推模型,从而实现数据与物理双驱动的高精度预测,确保预测结果满足物理运动规律。仿真结果表明,该方法在不同场景下均能保持较高的预测精度与稳定性,单步预测耗时保持在毫秒量级,满足实时性要求,且对非合作目标轨迹表现出良好的泛化能力,为高速目标的实时轨迹预报提供了可行的技术路径。

Abstract

To meet the real-time forecasting requirements of highly dynamic and strongly nonlinear trajectories in supersonic vehicle interception missions, and to address the limitations of traditional trajectory prediction methods, such as the lack of physical constraints, insufficient long-term temporal feature extraction, and weak generalization capability, this study proposes a hybrid-driven prediction method that integrates deep feature representation with dynamic constraints. Firstly, a three degree-of-freedom reentry dynamic model of the supersonic vehicle is established, and the effects of aerodynamic parameters and bank-angle commands on state evolution are analyzed. Secondly, a feature representation network based on a convolutional-bidirectional long short-term memory-self-attention architecture is designed to extract temporal dependencies and internal correlations of flight states. Finally, the parameters predicted by the network are incorporated as driving terms into the dynamic equations, forming a closed-loop recursive framework of “parameter prediction-state update-trajectory generation”, thereby achieving high-accuracy prediction through the joint driving of data and physical laws and ensuring consistency with the underlying motion dynamics. Simulation results show that the proposed method maintains high prediction accuracy and stability across different scenarios, with single-step inference remaining at the millisecond level, fully meeting real-time requirements. Moreover, it demonstrates strong generalization capability when applied to non-cooperative target trajectories, providing a feasible technical approach for real-time trajectory forecasting of supersonic vehicles.

关键词

高速飞行器 / 轨迹预测 / 混合驱动模型 / 深度学习 / 动力学约束

Key words

high-speed vehicle / trajectory prediction / hybrid-driven model / deep learning / dynamic constraints

参考文献

[1] 刘双喜, 刘世俊, 李勇, 等. 国外高超声速飞行器及防御体系发展现状[J]. 空天防御, 2023, 6(3): 39-51.

[2] LIU Shuangxi, LIU Shijun, LI Yong, et al. Current developments in foreign hypersonic vehicles and defense systems[J]. Air & Space Defense, 2023, 6(3): 39-51. (in Chinese)

[3] 胡诗国, 方洋旺, 张平, 等. 一种近空间高超声速飞行器的制导律设计与仿真[J]. 弹道学报, 2011, 23(3): 7-12.

[4] HU Shiguo, FANG Yangwang, ZHANG Ping, et al. Design and simulation of guidance law for a kind of near-space hypersonic vehicle[J]. Journal of Ballistics, 2011, 23(3): 7-12. (in Chinese)

[5] 李炯, 张涛, 张金鹏, 等. 基于零效脱靶量的制导估计一体化方法[J]. 弹道学报, 2017, 29(4): 35-39.

[6] LI Jiong, ZHANG Tao, ZHANG Jinpeng, et al. Unitization method of guidance and estimation based on zero-effort-miss[J]. Journal of Ballistics, 2017, 29(4): 35-39. (in Chinese)

[7] 杨春伟, 刘炳琪, 王继平, 等. 基于注意力机制的高超声速飞行器LSTM智能轨迹预测[J]. 兵工学报, 2022, 43(S2): 78-86.

[8] YANG Chunwei, LIU Bingqi, WANG Jiping, et al. LSTM intelligent trajectory prediction for hypersonic vehicles based on attention mechanism[J]. Acta Armamentarii, 2022, 43(S2): 78-86. (in Chinese)

[9] XIE Y, ZHUANG X, XI Z, et al. Dual-channel and bidirectional neural network for hypersonic glide vehicle trajectory prediction[J]. IEEE ACCESS, 2021, 9: 992913-92924.

[10] ZHANG J, XIONG J, LI L, et al. Motion state recognition and trajectory prediction of hypersonic glide vehicle based on deep learning[J]. IEEE ACCESS, 2022, 10: 21095-21108.

[11] BARTUSIAK E R, NGUYEN N X, CHAN M W, et al. A machine learning approach to classify hypersonic vehicle trajectories[C]// Proceedings of 2021 IEEE Aerospace Conference. Big Sky, MT: IEEE, 2021.

[12] BARTUSIAK E R, JACOBS M A, CHAN M W, et al. Predicting hypersonic glide vehicle behavior with stochastic grammars[J]. IEEE Transactions on Aerospace and Electronic Systems, 2024, 60(1): 1208-1223.

[13] 李佳丽, 郭杰, 唐胜景. 面向高超声速滑翔目标的多模型多意图融合轨迹预测[J]. 宇航学报, 2024, 45(2): 167-180.

[14] LI Jiali, GUO Jie, TANG Shengjing. Trajectory prediction based on multi-model and multi-intent fusion for hypersonic gliding targets[J]. Journal of Astronautics, 2024, 45(2): 167-180. (in Chinese)

[15] 张凯, 熊家军, 李凡, 等. 基于意图推断的高超声速滑翔目标贝叶斯轨迹预测[J]. 宇航学报, 2018, 39(11): 1258-1265.

[16] ZHANG Kai, XIONG Jiajun, LI Fan, et al. Bayesian trajectory prediction for a hypersonic gliding reentry vehicle based on intent inference[J]. Journal of Astronautics, 2018, 39(11): 1258-1265. (in Chinese)

[17] HU Y, GAO C, LI J, et al. Novel trajectory prediction algorithms for hypersonic gliding vehicles based on maneuver mode on-line identification and intent inference[J]. Measurement Science and Technology, 2021, 32(11): 115012.

[18] ZHOU D, HUANG D, SHEN Y, et al. An interpretable machine learning model for trajectory prediction based on nonlinear dynamics mechanism constraints: applications for HVs[J]. Neural Computing and Applications, 2023, 36(8): 4083-4100.

[19] LU S, QIAN Y. Enhanced trajectory forecasting for hypersonic glide vehicle via physics-embedded neural ODE[J]. Drones, 2024, 8: 377.

[20] SUN L, YANG B, MA J. Trajectory prediction in pipeline form for intercepting hypersonic gliding vehicles based on LSTM[J]. Chinese Journal of Aeronautics, 2023, 36(5): 421-433.

[21] ZHANG J, XIONG J, LAN X, et al. Trajectory prediction of hypersonic glide vehicle based on empirical wavelet transform and attention convolutional long short-term memory network[J]. IEEE Sensors Journal, 2022, 22(5): 4601-4615.

[22] 张君彪, 熊家军, 兰旭辉, 等. 一种高超声速滑翔飞行器轨迹智能预测方法[J]. 宇航学报, 2022, 43(4): 413-422.

[23] ZHANG Junbiao, XIONG Jiajun, LAN Xuhui, et al. An intelligent prediction method of hypersonic glide vehicle trajectory[J]. Journal of Astronautics, 2022, 43(4): 413-422. (in Chinese)

[24] 卢青. 高超声速飞行器制导控制与评估方法研究[D]. 西安:西北工业大学, 2018.

[25] LU Qing. Research of guidance control and evaluation for hypersonic vehicle[D]. Xi'an: Northwestern Polytechnical University, 2018. (in Chinese)

[26] 王智, 唐硕, 闫晓东. 高超声速滑翔飞行器约束预测校正再入制导[J]. 飞行力学, 2012, 30(2): 175-180.

[27] WANG Zhi, TANG Shuo, YAN Xiaodong. Constrained predictor-corrector reentry guidance for hypersonic glide vehicle[J]. Flight Dynamics, 2012, 30(2): 175-180. (in Chinese)

[28] 章吉力, 刘凯, 樊雅卓, 等. 考虑禁飞区规避的空天飞行器分段预测校正再入制导方法[J]. 宇航学报, 2021, 42(1): 122-131.

[29] ZHANG Jili, LIU Kai, FANG Yazhuo, et al. A piecewise predictor-corrector re-entry guidance algorithm with no-fly zone avoidance[J]. Journal of Astronautics, 2021, 42(1): 122-131. (in Chinese)

基金资助

国家自然科学基金项目(62373307)

国家级大学生创新训练项目(202410699200)

陕西省自然科学基础研究计划资助项目(2025JC-YBQN-585)

中央高校基本科研业务费专项资金资助()