[논문]Long Short-Term Memory Network for INS Positioning During GNSS Outages: A Preliminary Study on Simple Trajectories

Yujin Shin; Cheolmin Lee; Doyeon Jung; Euiho Kim

doi:10.11003/jpnt.2024.13.2.137

Long Short-Term Memory Network for INS Positioning During GNSS Outages: A Preliminary Study on Simple Trajectories 원문보기

Journal of Positioning, Navigation, and Timing, v.13 no.2, 2024년, pp.137 - 147

Yujin Shin (Autonomous Vehicle Intelligent Robotics Program, Hongik University) , Cheolmin Lee (Department of Mechanical Engineering, Hongik University) , Doyeon Jung (Department of Mechanical Engineering, Hongik University) , Euiho Kim (Department of Mechanical & System Design Engineering, Hongik University)

Abstract ▼ AI-Helper

This paper presents a novel Long Short-Term Memory (LSTM) network architecture for the integration of an Inertial Measurement Unit (IMU) and Global Navigation Satellite Systems (GNSS). The proposed algorithm consists of two independent LSTM networks and the LSTM networks are trained to predict attitudes and velocities from the sequence of IMU measurements and mechanization solutions. In this paper, three GNSS receivers are used to provide Real Time Kinematic (RTK) GNSS attitude and position information of a vehicle, and the information is used as a target output while training the network. The performance of the proposed method was evaluated with both experimental and simulation data using a lowcost IMU and three RTK-GNSS receivers. The test results showed that the proposed LSTM network could improve positioning accuracy by more than 90% compared to the position solutions obtained using a conventional Kalman filter based IMU/GNSS integration for more than 30 seconds of GNSS outages.

주제어

표/그림 (17)

그림 Fig. 1. Schematic process of an IMU mechanization.
$Fig. 2. Architecture of IMU-GNSS loosely coupled KF integrated navigation system. The position and attitude of a vehicle calculated from three GNSS antennas measurements are denoted as <TEX>$\widetilde{r}^e$</TEX> and <TEX>$\widetilde{\varphi}^e$</TEX>, respectively$ 그림 Fig. 2. Architecture of IMU-GNSS loosely coupled KF integrated navigation system. The position and attitude of a vehicle calculated from three GNSS antennas measurements are denoted as $\widetilde{r}^e$ and $\widetilde{\varphi}^e$, respectively
그림 Fig. 3. Arrangement of three orthogonally placed GNSS antennas for attitude and position measurements.
$Fig. 4. Schematic architecture of LSTM block with gating units. ft', it ot and <TEX>$\widetilde{C}_t$</TEX> are forget gate, input gate, output gate and cell input state, respectively.$ 그림 Fig. 4. Schematic architecture of LSTM block with gating units. f_t', i_t o_t and $\widetilde{C}_t$ are forget gate, input gate, output gate and cell input state, respectively.
그림 Fig. 5. Process of training LSTM network when GNSS measurements are available. The input data for the network has a following sequential form, x_seq =[x_t-SL+1, ⋯,x_t-1, x_t] where SL stands for the sequence length of the network.
그림 Fig. 6. Using trained LSTM network on GNSS outages. The input data for the network has a following sequential form, x_seq =[ x_t-SL+1,⋯, x_t-1, x_t] where SL stands for the sequence length of the network.
그림 Fig. 7. Configuration of mobile robot and onboard sensor system.
그림 Fig. 8. Two trajectories of mobile robot from the experiments, oval shape driving (top) and figure eight driving path (bottom). The average speed in oval shape driving was 2.79 m/s and 1.48 m/s in figure eight driving.
표 Table 1. Sensor specification of MPU9250 & Ublox F9P.
그림 Fig. 9. IMU measurements during the oval shape driving.
그림 Fig. 10. Results on each parameter combination on oval shape driving case (top), figure eight driving (bottom).
표 Table 2. Hyperparameters used to train the network.
그림 Fig. 11. Positioning results on oval shape driving case (top), figure eight driving (bottom) for 60 seconds GNSS outages.
그림 Fig. 12. Reference vehicle driving route data at the speed of 13.93 m/s divided into train and test data.
그림 Fig. 13. Results on each parameter combination on highway driving data.
그림 Fig. 14. Test results on 40 seconds GNSS outages. RMSE from INS are 42.18 m (East), 118.72 m (North), while RMSE from the trained network are 1.48 m (East), 3.48 m (North).
표 Table 3. Position RMSE on 60s GNSS outages.

AI 본문요약
AI-Helper

문제 정의

The previously mentioned prior arts appear to neglect the significant correlation between a vehicle's motion over time, and many have overlooked the importance of integrating sequential IMU and GNSS measurements into AI models to connect past and present navigation solutions. Therefore, this paper aims to derive navigation solution based on time-series sensor data from multiple previous epochs as input, rather than from a single-epoch data, as is done with neural networks. With this point, the paper further explores the IMU-GNSS integration problem using LSTM network that trained from the sequences of IMU measurements, mechanization solutions, and GNSS measurements to provide more reliable navigation solutions during GNSS outages.

제안 방법

For the two-dimensional path of a wheeled robot mainly covered in this paper, position information can be estimated using only heading values computed from the positions of two GNSS receivers. However, this paper used three GNSS receivers to compute three-axis attitude information with the intention of extending this method to platforms having more complex trajectories, such as drones.
With this point, the paper further explores the IMU-GNSS integration problem using LSTM network that trained from the sequences of IMU measurements, mechanization solutions, and GNSS measurements to provide more reliable navigation solutions during GNSS outages. In this paper, as a preliminary study for designing LSTM networks to derive navigation solution during GNSS outages, a specific input-output configuration of the LSTM network is proposed, and the results are provided as a function of the network sequence length and the number of hidden units for each GNSS outage time on simple vehicle trajectories. Two distinct LSTM networks were used to estimate attitude and velocity, respectively, and the attitude information computed from positions of three GNSS receivers were used to enhance the network training precision.
In this study, LSTM network was implemented to build a relationship between a navigation solution from IMU measurements through mechanizations and a precise RTK level navigation solution to provide accurate navigation solution during GNSS outages. The main objective of the trained LSTM network is to predict navigation solution on GNSS outages based on the sequence of IMU measurements and previous navigation solution.
This paper proposed using LSTM to provide more reliable navigation solutions than using INS during GNSS outages. The proposed algorithm consists of two independent LSTM networks to predict current attitude and velocity of the vehicle individually from the sequence of IMU measurements and mechanization solutions. During GNSS outages, the outputs of the trained network are used to compute position solutions.
In this paper, as a preliminary study for designing LSTM networks to derive navigation solution during GNSS outages, a specific input-output configuration of the LSTM network is proposed, and the results are provided as a function of the network sequence length and the number of hidden units for each GNSS outage time on simple vehicle trajectories. Two distinct LSTM networks were used to estimate attitude and velocity, respectively, and the attitude information computed from positions of three GNSS receivers were used to enhance the network training precision. For the two-dimensional path of a wheeled robot mainly covered in this paper, position information can be estimated using only heading values computed from the positions of two GNSS receivers.
Therefore, this paper aims to derive navigation solution based on time-series sensor data from multiple previous epochs as input, rather than from a single-epoch data, as is done with neural networks. With this point, the paper further explores the IMU-GNSS integration problem using LSTM network that trained from the sequences of IMU measurements, mechanization solutions, and GNSS measurements to provide more reliable navigation solutions during GNSS outages. In this paper, as a preliminary study for designing LSTM networks to derive navigation solution during GNSS outages, a specific input-output configuration of the LSTM network is proposed, and the results are provided as a function of the network sequence length and the number of hidden units for each GNSS outage time on simple vehicle trajectories.

성능/효과

The proposed method has been verified with both experimental data and simulated data and showed more than 90% improvements on positioning accuracy compared with INS on GNSS outages.
14 shows test results on 40 seconds GNSS outages. The predicted LSTM closely matches with the reference trajectory and the positioning solution of the LSTM network is improved by more than 90% compared to INS. Further research is necessary on finding an optimal network architecture to ensure near real-time application with various vehicle dynamic motions and path, which will require more data and information to build a rigorous model.
This paper proposed using LSTM to provide more reliable navigation solutions than using INS during GNSS outages. The proposed algorithm consists of two independent LSTM networks to predict current attitude and velocity of the vehicle individually from the sequence of IMU measurements and mechanization solutions.

후속연구

The predicted LSTM closely matches with the reference trajectory and the positioning solution of the LSTM network is improved by more than 90% compared to INS. Further research is necessary on finding an optimal network architecture to ensure near real-time application with various vehicle dynamic motions and path, which will require more data and information to build a rigorous model.

참고문헌 (27)

Abdolkarimi, E. S., Abaei, G., & Mosavi, M. R. 2018, A？wavelet-extreme learning machine for low-cost INS/GPS navigation system in high-speed applications, GPS？Solutions, 22, 15. https://doi.org/10.1007/s10291-017-0682-x？

상세보기
Adusumilli, S., Bhatt, D., Wang, H., Bhattacharya, P., &？Devabhaktuni, V. 2013, A low-cost INS/GPS integration？methodology based on random forest regression,？Expert Systems with Applications, 40, 4653-4659.？https://doi.org/10.1016/j.eswa.2013.02.002？

상세보기
Belhajem, I., Ben Maissa, Y., & Tamtaoui, A. 2018, Improving？low cost sensor based vehicle positioning with machine？learning, Control Engineering Practice, 74, 168-176.？https://doi.org/10.1016/j.conengprac.2018.03.006？

상세보기
Bengio, Y., Frasconi, P., & Simard, P. 1993, The problem？of learning long-term dependencies in recurrent？networks, In Proceedings of the IEEE International？Conference on Neural Networks, San Francisco, CA,？28 March - 1 April 1997, pp.1183-1188. https://doi.org/10.1109/ICNN.1993.298725？
Bitar, N. A., Gavrilov, A., & Khalaf, W. 2020, Artificial？intelligence based methods for accuracy improvement？of integrated navigation systems during GNSS？signal outages: An analytical overview, Gyroscopy？and Navigation, 11, 41-58. https://doi.org/10.1134/S2075108720010022？

상세보기
Chen, W., Li, X., Song, X., Li, B., Song, X., et al. 2015, A？novel fusion methodology to bridge GPS outages for？land vehicle positioning, Measurement Science and？Technology, 26, 075001. https://doi.org/10.1088/0957-0233/26/7/075001？

상세보기
Cui, Y. & Ge, S. S. 2003, Autonomous vehicle positioning？with GPS in urban canyon environments, IEEE？Transactions on Robotics and Automation, 19, 15-25.？https://doi.org/10.1109/TRA.2002.807557？

상세보기
El-Sheimy, N., Hou, H., & Niu, X. 2008, Analysis and？modeling of inertial sensors using Allan variance, IEEE？Transactions on Instrumentation and Measurement,？57, 140-149. https://doi.org/10.1109/TIM.2007.908635？

상세보기
ETRI AI Nanum [Internet], cited 2021 Nov 29, available？from: https://nanum.etri.re.kr/share/kimjy/car_dataset？
Fang, W., Jiang, J., Lu, S., Gong, Y., Tao, Y., et al. 2020, A LSTM？algorithm estimating pseudo measurements for aiding？INS during GNSS signal outages, Remote Sensing, 12,？256-280. https://doi.org/10.3390/rs12020256？

상세보기
Groves, P. D. 2013, Principles of GNSS, Inertial, and？Multisensor Integrated Navigation Systems, 2nd ed.？(Boston: Artech House Inc.)？
Han, S., Meng, Z., Omisore, O., Akinyemi, T., & Yan,？Y. 2020, Random error reduction algorithms for？MEMS inertial sensor accuracy improvement-A？review, Micromachines, 11, 1021-1057. https://doi.org/10.3390/mi11111021？

상세보기
Hochreiter, S. & Schmidhuber, J. 1997, Long Short-Term？Memory, Neural Computation, 9, 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735？

상세보기
InvenSense Inc. 2016, MPU-9250: Product Specification？Revision 1.1. (San Jose, CA: InvenSense Inc.). https://invensense.tdk.com/wp-content/uploads/2015/02/PSMPU-9250A-01-v1.1.pdf？
Jaradat, M. & Abdel-Hafez, M. 2017, Non-linear？autoregressive delay-dependent INS/GPS navigation？system using neural networks, IEEE Sensors Journal, 17,？1105-1115. https://doi.org/10.1109/JSEN.2016.2642040？

상세보기
Li, D., Jia, X., & Zhao, J. 2020, A novel hybrid fusion？algorithm for low-cost GPS/INS integrated navigation？system during GPS outages, IEEE Access, 8, 53984-53996. https://doi.org/10.1109/ACCESS.2020.2981015？

상세보기
Meiling, W., Guoqiang, F., Huachao, Y., Yafeng, L., Yi, Y., et al.？2017, A loosely coupled MEMS-SINS/GNSS integrated？system for land vehicle navigation in urban areas, In？Proceedings of the IEEE International Conference on？Vehicular Electronics and Safety, Vienna, Austria, 27-28 June, 2017, pp.103-108. https://doi.org/10.1109/ICVES.2017.7991909？
Noureldin, A., El-Shafie, A., & Bayoumi, M. 2011, GPS/INS integration utilizing dynamic neural networks for？vehicular navigation, Information Fusion, 12, 48-57.？https://doi.org/10.1016/j.inffus.2010.01.003？

상세보기
Noureldin, A., Karamat, T. B., & Georgy, J. 2012,？Fundamentals of Inertial Navigation, Satellite-based？Positioning and Their Integration (Berlin: Springer AG.)
Saadeddin, K., Abdel-Hafez, M. F., Jaradat, M. A., & Jarrah, M.？A. 2014, Optimization of intelligent approach for low-cost INS/GPS navigation system, Journal of Intelligent &？Robotic Systems, 73, 325-348. https://doi.org/10.1007/s10846-013-9943-2？

상세보기
U-blox AG. 2021, High precision GNSS module Professional？grade, ZED-F9P-02B - Data sheet. https://content.u-blox.com/sites/default/files/documents/ZED-F9P02B_DataSheet_UBX-21023276.pdf？
Wang, D., Xu, X., & Zhu, Y. 2018, A novel hybrid of a fading？filter and an extreme learning machine for GPS/INS？during GPS outages, Sensors, 18, 3863-3885. https://doi.org/10.3390/s18113863？

상세보기
Wang, R., Hou, X., Liu, F., & Yu, Y. 2020, GPS/INS integrated？navigation for quadrotor UAV considering lever arm, In？Proceedings of the 2020 35th Youth Academic Annual？Conference of Chinese Association of Automation,？Zhanjiang, China, 16-18 October, 2020, pp.132-136.？https://doi.org/10.1109/YAC51587.2020.9337634？
Wei, X., Li, J., Feng, K., Zhang, D., Li, P., et al. 2021, A mixed？optimization method based on adaptive Kalman filter？and wavelet neural network for INS/GPS during GPS？outages, IEEE Access, 9, 47875-47886. https://doi.org/10.1109/ACCESS.2021.3068744？

상세보기
Yao, Y., Xu, X., Zhu, C., & Chan, C. 2017, A hybrid fusion？algorithm for GPS/INS integration during GPS outages,？Measurement, 103, 42-51. https://doi.org/10.1016/j.measurement.2017.01.053？

상세보기
Zhang, Y. 2019, A fusion methodology to bridge GPS？outages for INS/GPS integrated navigation system,？IEEE Access, 7, 61296-61306. https://doi.org/10.1109/ACCESS.2019.2911025？

상세보기
Zhao, L., Li, N., Li, L., Zhang, Y., & Cheng, C. 2017, Real-time GNSS-based attitude determination in the？measurement domain, Sensors, 17, 296-311. https://doi.org/10.3390/s17020296？

상세보기

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증