一种基于深度学习辅助的SINS/DVL组合导航方法

doi:10.16182/j.issn1004731x.joss.24-0273

摘要/Abstract

摘要：

水下机器人(autonomous underwater vehicle，AUV)导航定位精度一定上程度影响了AUV的工作效率，由于GNSS无法在水下使用，以捷联惯导系统/多普勒计程仪(SINS/DVL)的组合导航系统受到众多青睐。DVL在部分情况下会失效，若直接将DVL隔离，系统将变为纯惯性导航系统，严重影响导航定位的精度。为了应对DVL在部分波束缺失的情况，提出了一种DLinear-informer辅助组合导航算法。通过DLinear对原始输入数据特有的分解方式，增强了算法对AUV非线性信息的提取与学习，提高了速度预测精度。实验结果表明：所提算法能够准确预测DVL失效期间丢失波束的速度，降低组合导航的位置误差，提高了系统的鲁棒性和定位精度。

关键词: 水下机器人, 捷联惯导系统/多普勒计程仪, 组合导航, 波束缺失, DLinear-informer

Abstract:

The navigation and positioning accuracy of an Autonomous Underwater Vehicle (AUV) affects the efficiency of the AUV to a certain extent, and since GNSS cannot be used underwater, the integrated navigation system of Strapdown Inertial Navigation System/ Doppler Velocity Log (SINS/DVL) has been widely favored. However, DVL will fail in some cases, and if DVL is isolated directly, the system will become a pure inertial navigation system, which seriously affects the accuracy of navigation and positioning. In order to cope with the situation that DVL is missing in some beams, a DLinear-Informer assisted integrated navigation algorithm is proposed. Through DLinear's unique decomposition of the original input data, the algorithm enhances the extraction and learning of the nonlinear information of AUV and improves the velocity prediction accuracy. The experimental results show that the proposed algorithm can accurately predict the velocity of the lost beam during the DVL failure, reduce the position error of the integrated navigation, and improve the robustness and positioning accuracy of the system.

Key words: autonomous underwater vehicle(AUV), strapdown inertial navigation system/doppler velocity logger, integrated navigation, missing beams, DLinear-informer

中图分类号:

TP391.9

匡兴红,黄傲威 . 一种基于深度学习辅助的SINS/DVL组合导航方法[J]. 系统仿真学报, 2024, 36(8): 1810-1822.

Kuang Xinghong,Huang Aowei . A Method Based on Deep Learning for Assisting SINS/DVL Integrated Navigation[J]. Journal of System Simulation, 2024, 36(8): 1810-1822.

图/表 23

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

图10

图11

表2

图12

图13

图14

表3

图15

图16

图17

表4

图18

表5

参考文献 19

1	郑为, 邹启明, 倪文玺. 水下无人潜航器避障与导航仿真系统设计与应用[J]. 系统仿真学报, 2016, 28(1): 91-98.
	Zheng Wei, Zou Qiming, Ni Wenxi. Design and Application of Underwater Unmanned Vehicle Simulation System for Navigation and Obstacle Avoidance[J]. Journal of System Simulation, 2016, 28(1): 91-98.
2	刘沛佳. INS/DVL组合导航关键技术研究[D]. 北京: 北京理工大学, 2018.
	Liu Peijia. Research on Key Technologies of INS/DVL Integrated Navigation[D]. Beijing: Beijing Institute of Technology, 2018.
3	宋世磊. SINS/DVL水下组合导航技术研究[D]. 镇江: 江苏科技大学, 2022.
	Song Shilei. Research on SINS/DVL Integrated Navigation Technology[D]. Zhenjiang: Jiangsu University of Science and Technology, 2022.
4	龙瑛芝. 水下声学定位声速改正及定位算法研究[D]. 西安: 长安大学, 2023.
	Long Yingzhi. Research on Underwater Acoustic Positioning Sound Velocity Correction and Positioning Algorithm[D]. Xi'an: Chang'an University, 2023.
5	杨玉孔. 惯性与DVL计程仪组合导航技术研究[J]. 光学与光电技术, 2023, 21(5): 125-130.
	Yang Yukong. Study on Inertial and DVL High Precision Integrated Navigation Technology[J]. Optics & Optoelectronic Technology, 2023, 21(5): 125-130.
6	Miller P A, Farrell J A, Zhao Yuanyuan, et al. Autonomous Underwater Vehicle Navigation[J]. IEEE Journal of Oceanic Engineering, 2010, 35(3): 663-678.
7	Klein Itzik, Lipman Yeshaya. Continuous INS/DVL Fusion in Situations of DVL Outages[C]//2020 IEEE/OES Autonomous Underwater Vehicles Symposium (AUV). Piscataway: IEEE, 2020: 1-6.
8	Mirabadi A, Mort N, Schmid F. Fault Detection and Isolation in Multisensor Train Navigation Systems[C]//UKACC International Conference on Control'98. Piscataway: IEEE, 1998: 969-974.
9	曾观林, 冯国虎. 水下SINS/DVL组合导航误差抑制综述[J]. 现代防御技术, 2023, 51(4): 25-35.
	Zeng Guanlin, Feng Guohu. Review on the Error Suppression of Underwater SINS/DVL Integrated Navigation[J]. Modern Defence Technology, 2023, 51(4): 25-35.
10	Karmozdi Ali, Hashemi Mojtaba, Salarieh Hassan, et al. Implementation of Translational Motion Dynamics for INS Data Fusion in DVL Outage in Underwater Navigation[J]. IEEE Sensors Journal, 2021, 21(5): 6652-6659.
11	潘绍华, 徐晓苏, 张亮. 基于卡方检测和相关向量机的DVL异常信息处理机制[J]. 中国惯性技术学报, 2022, 30(4): 461-468.
	Pan Shaohua, Xu Xiaosu, Zhang Liang. DVL Exception Information Processing Mechanism Based on Chi-square Test and RVM[J]. Journal of Chinese Inertial Technology, 2022, 30(4): 461-468.
12	Jin Kaidi, Chai Hongzhou, Su Chuhan, et al. A Performance-enhanced DVL/SINS Integrated Navigation System Based on Data-driven Approach[J]. Measurement Science and Technology, 2023, 34(9): 095120.
13	Zhu Jiupeng, Li An, Qin Fangjun, et al. A Novel Hybrid Method Based on Deep Learning for an Integrated Navigation System during DVL Signal Failure[J]. Electronics, 2022, 11(19): 2980.
14	Cohen Nadav, Yampolsky Zeev, Klein Itzik. Set-transformer BeamsNet for AUV Velocity Forecasting in Complete DVL Outage Scenarios[C]//2023 IEEE Underwater Technology (UT). Piscataway: IEEE, 2023: 1-6.
15	Cohen Nadav, Klein Itzik. BeamsNet: A Data-driven Approach Enhancing Doppler Velocity Log Measurements for Autonomous Underwater Vehicle Navigation[J]. Engineering Applications of Artificial Intelligence, 2022, 114: 105216.
16	Zhou Haoyi, Zhang Shanghang, Peng Jieqi, et al. Informer: Beyond Efficient Transformer for Long Sequence Time-series Forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(12): 11106-11115.
17	Tian Xinyu, Zheng Qinghe, Yu Zhiguo, et al. A Real-time Vehicle Speed Prediction Method Based on a Lightweight Informer Driven by Big Temporal Data[J]. Big Data and Cognitive Computing, 2023, 7(3): 131.
18	Wang Di, Xu Xiaosu, Yao Yiqing, et al. A Novel SINS/DVL Tightly Integrated Navigation Method for Complex Environment[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(7): 5183-5196.
19	Cohen Nadav, Klein Itzik. LiBeamsNet: AUV Velocity Vector Estimation in Situations of Limited DVL Beam Measurements[C]//OCEANS 2022, Hampton Roads. Piscataway: IEEE, 2022: 1-5.

参数	值
Learning rate	0.000 1
Epoces	50
Batch size	32
Num_heads	8
Encoder_layers	2
Decoder_layers	1
Dropout	0.001
Hidden_layers	512

模型	MAE	STD
MLP	0.049 4	0.079 2
CNN	0.033 7	0.018 8
LSTM	0.029 4	0.026 9
DLinear-Informer	0.012 6	0.016 3

模型	MAE		STD
模型	Beam3	Beam4	Beam3	Beam4
MLP	0.139 5	0.053 1	0.145 8	0.062 1
CNN	0.062 5	0.050 8	0.064 3	0.043 6
LSTM	0.036 2	0.021 2	0.056 5	0.032 7
DLinear-Informer	0.026 9	0.014 8	0.038 1	0.019 1

模型	MAE			STD
模型	Beam2	Beam3	Beam4	Beam2	Beam3	Beam4
MLP	0.149 0	0.141 9	0.064 9	0.133 4	0.117 7	0.050 8
CNN	0.091 7	0.080 0	0.054 5	0.093 8	0.092 6	0.060 3
LSTM	0.051 7	0.050 2	0.024 0	0.088 4	0.079 7	0.038 9
DLinear-Informer	0.027 9	0.028 9	0.014 5	0.038 6	0.041 1	0.018 8

类型	1个束波丢失	2个束波丢失		3个束波丢失
类型	Beam4	Beam3	Beam4	Beam2	Beam3	Beam4
p	0.54	0.81	0.40	0.49	0.66	0.37
t	0.61	-0.24	0.84	-0.69	-0.45	0.90