在轨目标探测任务-资源快速匹配方法

doi:10.12305/j.issn.1001-506X.2026.09.16

摘要/Abstract

摘要：

为提高任务-资源匹配的准确性与时效性，提出自适应注意力双向长短期记忆（adaptive attention bidirectional long short-term memory, DA-BiLSTM）网络算法。算法的输入为待分配在轨目标和资源信息，输出为最佳匹配方案。采用注意力机制确定隐向量，对输入信息进行筛选和集中；通过双向长短期记忆网络实现对多维特征之间关系捕捉，确定网络输出；最后应用贝叶斯算法进行超参数优化。仿真结果表明，与当前流行的卷积神经网络（conventional neural network, CNN）、CNN-LSTM算法进行对比，所提方法在分类准确率、灵敏度、特异性、和F1分数平均提高了22.14%、22.97%、26.85%和16.7%。消融实验验证了网络策略的有效性。

关键词: 太空态势感知, 空间监视网, 任务分配, 资源调度, 深度学习

Abstract:

To enhance the accuracy and timeliness of task-resource matching, an adaptive attention bidirectional long short-term memory (DA-BiLSTM) network algorithm is proposed. The algorithm takes the resident space objective (RSO) and resource information to be allocated as input and outputs the optimal matching scheme. The attention mechanism is used to determine the hidden vectors, filtering and focusing the input information. The BiLSTM network is employed to capture the relationships between multi-dimensional features, to determine the network output. Finally, the Bayesian algorithm is applied for hyperparameter optimization. Simulation result shows that in comparison with the currently popular convolutional convolutional neural network (CNN) and CNN-LSTM algorithms, the proposed method improves accuracy, sensitivity, specificity, and F1 score by 22.14%, 22.97%, 26.85%, and 16.7%, respectively. Ablation experiments validate the effectiveness of the network’s strategies.

Key words: space situational awareness, space surveillance network, task allocation, resource scheduling, deep learning

中图分类号:

V 11

朱彦伟, 陈劲润, 赵晓锋, 龙洗, 乔琛远. 在轨目标探测任务-资源快速匹配方法[J]. 系统工程与电子技术, 2026, 48(9): 3029-3037.

Yanwei Zhu, Jinrun Chen, Xiaofeng Zhao, Xi Long, Chenyuan Qiao. Task-resource fast matching method for resident space objectives detection[J]. Systems Engineering and Electronics, 2026, 48(9): 3029-3037.

图/表 19

图1

表1

图2

图3

图4

表2

图5

表3

图6

表4

图7

图8

图9

表5

表6

表7

表8

表9

图10

参考文献 31

1	Liu A R, Xu X L, Xiong Y Q, et al. Maneuver strategies of Starlink satellite based on SpaceX-released ephemeris[J]. Advances in Space Research, 2024, 74 (7): 3157. doi: 10.1016/j.asr.2024.06.038
2	Kozhaya S, Kassas Z M. A first look at the OneWeb LEO constellation: beacons, beams, and positioning[J]. IEEE Trans. on Aerospace and Electronic Systems, 2024, 60 (5): 7528. doi: 10.1109/taes.2024.3410252
3	翟光, 王妍欣, 孙一勇. 基于低轨星网的多目标协同跟踪滤波技术[J]. 系统工程与电子技术, 2022, 44 (6): 1957.
4	Xue C B, Cai H, Gehly S, et al. Review of sensor tasking methods in space situational awareness[J]. Progress in Aerospace Sciences, 2024, 147, 101017. doi: 10.1016/j.paerosci.2024.101017
5	Long X, Yang L P, Qiao C Y. Emergency scheduling based on event triggering and multi-hierarchical planning for space surveillance network[J]. Information Sciences, 2024, 337, 120486. doi: 10.2139/ssrn.4577266
6	Nicholas R, Brandon A J. Risk-aware sensor scheduling and tracking of large constellations[J]. Advances in Space Research, 2021, 68 (6): 2530. doi: 10.1016/j.asr.2021.04.042
7	Cai H, Yang Y, Gehly S, et al. Sensor tasking for search and catalog maintenance of geosynchronous space objects[J]. Acta Astronautica, 2020, 175, 234. doi: 10.1016/j.actaastro.2020.05.063
8	Wang J L, Zheng X Y, Shen J K, et al. A modified weighting scheme for the automatic tasker of space surveillance network[C]//IEEE International Conference on Signal Processing, 2022: 524.
9	Long X, Cai W W, Yang L P, et al. Mission scheduling of multi-sensor collaborative observation for space surveillance network[J]. Journal of Systems Engineering and Electronics, 2023, 34 (4): 906. doi: 10.23919/JSEE.2023.000104
10	Qin J H, Bai X, Du G M, et al. Multisatellite scheduling for moving targets using the enhanced hybrid genetic simulated annealing algorithm and observation strip selection[J]. IEEE Trans. on Aerospace and Electronic Systems, 2024, 60, 5773. doi: 10.1109/TAES.2024.3397958
11	Yao F, Chen Y G, Wang L, et al. A bilevel evolutionary algorithm for largescale multiobjective task scheduling in multiagile earth observation satellite systems[J]. IEEE Trans. on Systems, Man, and Cybernetics: Systems, 2024, 54, 3512. doi: 10.1109/tsmc.2024.3359265/mm1
12	Bryan D L, Carolin E F. Space situational awareness sensor tasking: comparison of machine learning with classical optimization methods[J]. Journal of Guidance, Control, and Dynamics, 2020, 43 (2): 262. doi: 10.2514/1.g004279
13	Du Y H, Wang T, Xin B, et al. A data-driven parallel scheduling approach for multiple agile earth observation satellites[J]. IEEE Trans. on Evolutionary Computation, 2024, 24, 679.
14	Wu G H, Mao X, Chen Y G, et al. Coordinated scheduling of air and space observation resources via divide and conquer framework and iterative optimization[J]. IEEE Trans. on Aerospace and Electronic Systems, 2023, 59, 3631. doi: 10.1109/TAES.2022.3228832
15	Dhingra N, Dejac C, McGuire C. Machine learning for space domain awareness sensor scheduling[C]//Advanced Maui Optical and Space Surveillance Technologies Conference, 2024: 86.
16	Peng M S, Daniel J, Thomas G R, et al. Space-based sensor tasking using deep reinforcement learning[J]. The Journal of the Astronautical Sciences, 2022, 69 (6): 1855. doi: 10.1007/s40295-022-00354-8
17	Chen X Y, Gu W C, Dai G M, et al. Data-driven collaborative scheduling method for multi-satellite data-transmission[J]. Tsinghua Science and Technology, 2024, 29, 1463. doi: 10.26599/TST.2023.9010131
18	Li J, Li J, Jing N, et al. A satellite schedulability prediction algorithm for EO SPS[J]. Chinese Journal of Aeronautics, 2013, 3, 705. doi: 10.1016/j.cja.2013.04.058
19	邢立宁, 王原, 何永明, 等. 基于BP神经网络的星上任务可调度性预测方法[J]. 中国管理科学, 2015, 23 (S1): 117.
20	Wang C K, Zheng W Z, Zhu Z, et al. Introspective deep metric learning[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2024, 46 (4): 1964. doi: 10.1109/TPAMI.2023.3312311
21	Wen X H, Zhou M C. Evolution and role of optimizers in training deep learning models[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11 (10): 2039. doi: 10.1109/JAS.2024.124806
22	Siew P M, Smith T, Ponmalai R, et al. Scalable multi-agent sensor tasking using deep reinforcement learning[C]//Advanced Maui Optical and Space Surveillance Technologies Conference, 2023.
23	Wang M, Zhou Z B, Chang Z X, et al. Deep reinforcement learning for agile earth observation satellites scheduling problem with variable image duration[J]. Applied Soft Computing, 2025, 169, 112575. doi: 10.1016/j.asoc.2024.112575
24	Pu X Y, Xu F. Low-rank adaption on transformer-based oriented object detector for satellite onboard processing of remote sensing images[J]. IEEE Trans. on Geoscience and Remote Sensing, 2024,
25	Li C L, Zhang J W, Huo B H, et al. DHQ-DETR: distributed and high-quality object query for enhanced dense detection in remote sensing[J]. Remote Sensing, 2025, 17 (3): 514. doi: 10.3390/rs17030514
26	Rojhani N, Sadeghibakhi M, Passafiume M, et al. Application of PCA and unsupervised deep learning in bird and drone discrimination based on FMCW radar measurements[J]. IEEE Geoscience and Remote Sensing Letters, 2024, 21, 3510805. doi: 10.1109/lgrs.2024.3487008
27	Qin Y H, Li Z N, Xie S L, et al. Non-line-of-sight multipath classification method for BDS using convolutional sparse autoencoder with LSTM[J]. Tsinghua Science and Technology, 2025, 30 (1): 68. doi: 10.26599/TST.2024.9010004
28	United States Space Command. Space-track[EB/OL]. [2024-11-30]. https://www.space-track.org.
29	Chen J R, Yang L P, Huang H, et al. A task-priority’s time-density based resource allocation method for resident space objectives catalog maintaining[C]//4th International Conference on Autonomous Unmanned Systems, 2024.
30	Xu F L, Mei S H, Zhang G, et al. Bridging CNN and transformer with cross attention fusion network for hyperspectral image classification[J]. IEEE Trans. on Geoscience and Remote Sensing, 2024, 62, 5522214.
31	Nathaniel S Y T, Mau-Luen T, Sing Y C, et al. Accurate and fast classification of natural disasters using CNN-LSTM and inference acceleration[J]. Journal of Communications Software and Systems, 2024, 20 (1): 58.

类别	序号	内容
探测窗口特征	1	探测资源的窗口数量
	2	探测资源窗口平均开始时间
	3	探测资源窗口平均结束时间
	4	探测资源窗口最早开始时间
	5	探测资源窗口最早结束时间
	6	探测资源窗口最晚开始时间
	7	探测资源窗口最晚结束时间
	8	探测资源窗口平均时长
	9	探测资源窗口平均间隔
	10	探测资源窗口总计时长
RSO 轨道特征	11	RSO的轨道六根数（半长轴、偏心率、轨道倾角、升交点赤经、近地点幅角、真近点角）
探测任务特征	12	探测任务权重
	13	探测任务需求
	14	探测任务频率
	15	探测任务执行时间

类型预测	实际类型
类型预测	正	负
正	TP	FP
负	FN	TN

名称	准确率/%	灵敏度	特异性	F1分数
DA-BiLSTM	72.2	0.63	0.68	0.74
CNN	62.1	0.47	0.52	0.59
CNN-LSTM	56.1	0.49	0.49	0.61

组别	特征参数
第1组	优先级、探测需求、探测次数、探测时长、各测站窗口数目、轨道六根数
第2组	各测站窗口数目、平均窗口开始时间、平均窗口结束时间、窗口持续总时间、平均窗口持续时间、窗口平均间隔、窗口最早开始时间、窗口最早结束时间、窗口最晚开始时间和窗口最晚结束时间、轨道倾角、升交点赤经、近地点幅角、平近点角
第3组	优先级、探测时长、各测站窗口数目、轨道倾角、升交点赤经、近地点幅角、平近点角

组别	准确率
本文	72.2
第1组	48.5
第2组	60.6
第3组	48.2