基于Vondrak-Cepek组合滤波和注意力机制加权的时间比对融合算法

doi:10.12305/j.issn.1001-506X.2025.02.34

摘要/Abstract

摘要：

针对卫星双向时间频率传递(two-way satellite time and frequency transfer, TWSTFT)存在周日效应、短期稳定度不高的问题, 通过引入基于注意力机制的Transformer权值矩阵, 利用Vondrak-Cepek组合滤波的方法将中国科学院国家授时中心(National Time Service Center, NTSC)、德国物理技术研究院(Physikalisch-Technische Bundesanstalt, PTB)之间的TWSTFT和全球定位系统(Global Positioning System, GPS)P3码共视法的时间比对链路进行融合, 分析融合前后链路的性能指标并与没有周日效应、短期稳定度高的GPS精密单点定位(GPS precise point positioning, GPS PPP)时间比对参考链路进行比较。结果表明, 引入注意力机制权值的Vondrak-Cepek组合滤波融合方法与参考链路GPS PPP的标准差为0.310 9 ns, 具有改善TWSTFT周日效应、提升链路整体稳定性的作用。

关键词: Vondrak-Cepek组合滤波, 注意力机制, 时间比对, 数据融合

Abstract:

To solve the problems of the diurnal and low short-term stability in two-way satellite time and frequency transfer (TWSTFT), a Transformer weight matrix based on the attention mechanism is introduced. The Vondrak-Cepek combined filtering method is applied to fuse the TWSTFT and Global Positioning System (GPS) P3 code common view method time alignment links between the National Time Service Center (NTSC) and Physikalisch-Technische Bundesanstalt(PTB). The performance indicators of the links before and after fusion are analyzed and compared with the GPS precise point positioning (GPS PPP) time alignment reference links with no diurnal and high short-term stability. The results show that the standard deviation of Vondrak-Cepek combined filtering fusion method with the attention mechanism weight is 0.310 9 ns comparing with the reference link GPS PPP, which can reduce the diurnal of TWSTFT and enhance the overall stability of the link.

Key words: Vondrak-Cepek combined filtering, attention mechanism, time alignment, data fusion

中图分类号:

P127.1
TN713

刘强, 孙浩冉, 胡邓华, 张爽. 基于Vondrak-Cepek组合滤波和注意力机制加权的时间比对融合算法[J]. 系统工程与电子技术, 2025, 47(2): 673-679.

Qiang LIU, Haoran SUN, Denghua HU, Shuang ZHANG. Time alignment fusion algorithm based on Vondrak-Cepek combined filtering and attention mechanism weighting[J]. Systems Engineering and Electronics, 2025, 47(2): 673-679.

图/表 11

图1

图2

图3

表1

图4

图5

图6

图7

图8

表2

表3

参考文献 16

1	JIANGZ H,ZHANGV,PARKERT E,et al.Improving two-way satellite time and frequency transfer with redundant links for UTC generation[J].Metrologia,2019,56(2):025005. doi: 10.1088/1681-7575/aafced
2	付栋,彭竞,马明,等.基于钟差检验的GNSS授时欺骗检测与识别[J].系统工程与电子技术,2022,44(3):948-955. doi: 10.12305/j.issn.1001-506X.2022.03.27
	FUD,PENGJ,MAM,et al.GNSS time spoofing detection and discrimination based on clock bias hypothesis test[J].Systems Engineering and Electronics,2022,44(3):948-955. doi: 10.12305/j.issn.1001-506X.2022.03.27
3	武文俊. 卫星双向时间频率传递的误差研究[D]. 西安: 中国科学院研究生院(国家授时中心), 2012.
	WU W J. Research on two-way staellite time and frequency transfer errors[D]. Xi'an: Graduate University of Chinese Academy of Sciences (National Time Service Center), 2012.
4	武文俊.卫星双向时间频率传递的误差研究[J].天文学报,2013,54(4):403-404. doi: 10.3969/j.issn.0001-5245.2013.04.011
	WUW J.A research on errors in two-way satellite time and frequency transfer[J].Acta Astronomica Sinica,2013,54(4):403-404. doi: 10.3969/j.issn.0001-5245.2013.04.011
5	王威雄,董绍武,武文俊,等.卫星双向时间频率传递链路校准及其不确定度分析[J].仪器仪表学报,2018,39(12):64-72.
	WANGW X,DONGS W,WUW J,et al.Link calibration of two-way satellite time and frequency transfer and its uncertainty analysis[J].Chinese Journal of Scientific Instrument,2018,39(12):64-72.
6	王翔,宋会杰,郭栋,等.卫星双向时间频率传递链路性能优化方法研究[J].时间频率学报,2022,45(4):254-261.
	WANGX,SONGH J,GUOD,et al.Research on performance optimization method of two-way satellite time and frequency transfer link[J].Journal of Time and Frequency,2022,45(4):254-261.
7	JIANGZ,PETITG.Combination of TWSTFT and GNSS for accurate UTC time transfer[J].Metrologia,2009,46(3):305-314. doi: 10.1088/0026-1394/46/3/019
8	王威雄,董绍武,武文俊,等.Vondrak-Cepek组合滤波在北斗共视和卫星双向时间比对融合中的应用[J].国防科技大学学报,2021,43(6):17-25.
	WANGW X,DONGS W,WUW J,et al.Application of Vondrak-Cepek combined filtering in the fusion of BeiDou CV and TWSTFT[J].Journal of National University of Defense Technology,2021,43(6):17-25.
9	毛远宏,孙琛琛,徐鲁豫,等.基于深度学习的时间序列预测方法综述[J].微电子学与计算机,2023,40(4):8-17.
	MAOY H,SUNC C,XUL Y,et al.A survey of time series forecasting methods based on deep learning[J].Microelectronics & Computer,2023,40(4):8-17.
10	HOCHREITERS,SCHMIDHUBERJ.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735
11	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proc. of the 31st Conference on Neural Information Processing Systems, 2017: 6000-6010.
12	黄艳. GPS P3码共视算法的研究[D]. 北京: 北京工业大学, 2009.
	HUANG Y. The research of the common-view comparision system based on GPS P3 code receiver[D]. Beijing: Beijing University of Technology, 2009.
13	BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate[C]//Proc. of the 3rd International Conference on Learning Representations, 2015.
14	LUONG M T, PHAM H, MANNING C D. Effective approaches to attention-based neural machine translation[C]//Proc. of the Conference on Emprical Methods in Natrual Language Processing, 2015: 1412-1421.
15	VONDRAKJ,CEPEKA.Combined smoothing method and its use in combining Earth orientation parameters measured by space techniques[J].Astronomy and Astrophysics Supplement Series,2000,147(2):347-359. doi: 10.1051/aas:2000306
16	王威雄. 守时系统国际时间比对数据融合方法研究[D]. 西安: 中国科学院大学(国家授时中心), 2021.
	WANG W X. Research on combination method of international time comparision data of timekeeping system[D]. Xi'an: University of Chinese Academy of Sciences(National Time Service Center), 2021.

周期	平滑系数	T/T′
周期	平滑系数	0.1	0.3	0.5	0.7	0.99
P=0.5	ε	437 540	1 687 600	3 937 800	9 188 300	389 850 000
P=0.5	ε	2 771	10 687	24 937	58 186	2 468 700
P=1	ε	6 837	26 370	61 529	143 570	6 091 400
P=1	ε	173	668	1 559	3 637	154 300
P=1.5	ε	600	2 315	5 402	12 604	534 770
P=1.5	ε	34	132	308	718	30 478

方法	最大值	最小值	标准差	均值
TW+P3	0.421 5	-4.275 9	1.234 0	-1.538 5
TW′+P3′	-0.039 4	-4.301 9	1.158 3	-1.506 5
TWSTFT	3.626 0	-7.238 0	1.756 1	-1.542 9
GPS P3	1.472 0	-5.906 0	1.403 5	-1.538 5
GPS PPP	0.613 0	-4.333 0	1.251 3	-1.542 4

方法	最大值	最小值	标准差
TW+P3-GPS PPP	0.444 9	-0.392 5	0.183 0
TW′+P3′-GPS PPP	0.751 2	-0.699 6	0.310 9
TWSTFT-GPS PPP	5.207 0	-4.200 0	1.284 9
GPS P3-GPS PPP	1.638 0	-1.991 0	0.685 2

[1]	陈晓萱, 徐书文, 胡绍海, 马晓乐. 基于卷积与自注意力的红外与可见光图像融合[J]. 系统工程与电子技术, 2024, 46(8): 2641-2649.
[2]	应文健, 程雨森, 王旋, 孙世岩. 基于研制阶段数据融合的舰炮制导弹药测试性评估方法[J]. 系统工程与电子技术, 2024, 46(8): 2730-2737.
[3]	吴嘉俊, 苏春, 张玉茹. 基于双重自注意力机制和长短时记忆网络的剩余寿命预测[J]. 系统工程与电子技术, 2024, 46(6): 1986-1994.
[4]	薛丽莎, 葛瑞星, 朱宇轩, 鲍雁飞. 基于SABEGAN的通信干扰信号生成与效能分析[J]. 系统工程与电子技术, 2024, 46(6): 2155-2163.
[5]	徐奕禹, 陈长风, 袁雪林, 陈正坤, 周志健. 基于高度计辅助的GNSS欺骗干扰检测[J]. 系统工程与电子技术, 2024, 46(5): 1484-1492.
[6]	刘绍华, 杜康, 佘春东, 杨傲. 基于CenterNet的多教师联合知识蒸馏[J]. 系统工程与电子技术, 2024, 46(4): 1174-1184.
[7]	邱建杰, 蔡益朝, 李浩, 黄权印. 基于动态估计反馈的灰色理论航迹关联算法[J]. 系统工程与电子技术, 2024, 46(4): 1401-1411.
[8]	方小宇, 黄丽佳. 基于全局位置信息和残差特征融合的SAR船舶检测算法[J]. 系统工程与电子技术, 2024, 46(3): 839-848.
[9]	张春杰, 王冠博, 陈奇, 邓志安. 基于纯自注意力机制的毫米波雷达手势识别[J]. 系统工程与电子技术, 2024, 46(3): 859-867.
[10]	赵晓枫, 牛家辉, 刘春桐, 夏玉婷. 基于三维注意力与混合卷积的高光谱图像分类[J]. 系统工程与电子技术, 2023, 45(9): 2673-2680.
[11]	李海军, 孔繁程, 林云. 基于改进YOLOv5s的红外舰船检测算法[J]. 系统工程与电子技术, 2023, 45(8): 2415-2422.
[12]	邓喆, 雷菁, 孙承哲. 跳频信号盲检测的半监督干扰对消方法[J]. 系统工程与电子技术, 2023, 45(7): 2236-2248.
[13]	赵庆媛, 赵志强, 叶春茂, 鲁耀兵. 基于自注意力的双波段预警雷达微动融合识别[J]. 系统工程与电子技术, 2023, 45(3): 708-716.
[14]	曹鹏宇, 杨承志, 陈泽盛, 王露, 石礼盟. 基于深度残差收缩注意力网络的雷达信号识别方法[J]. 系统工程与电子技术, 2023, 45(3): 717-725.
[15]	闫啸家, 梁伟阁, 张钢, 佘博, 田福庆. 基于RCNN-ABiLSTM的机械设备剩余寿命预测方法[J]. 系统工程与电子技术, 2023, 45(3): 931-940.