AI-IMU Dead-Reckoning

Martin Brossard; Axel Barrau; Silvère Bonnabel

doi:10.1109/TIV.2020.2980758

Article Dans Une Revue IEEE Transactions on Intelligent Vehicles Année : 2020

AI-IMU Dead-Reckoning

(1) , (2) , (1)

1
2

Martin Brossard

Fonction : Auteur
PersonId : 1016519

Centre de Robotique

Axel Barrau

Fonction : Auteur
PersonId : 776804
IdRef : 194303462

Safran Tech

Silvère Bonnabel

Fonction : Auteur
PersonId : 866182
IdHAL : silvere-bonnabel
ORCID : 0000-0002-6001-7766

Centre de Robotique

Résumé

In this paper we propose a novel accurate method for dead-reckoning of wheeled vehicles based only on an Inertial Measurement Unit (IMU). In the context of intelligent vehicles, robust and accurate dead-reckoning based on the IMU may prove useful to correlate feeds from imaging sensors, to safely navigate through obstructions, or for safe emergency stops in the extreme case of exteroceptive sensors failure. The key components of the method are the Kalman filter and the use of deep neural networks to dynamically adapt the noise parameters of the filter. The method is tested on the KITTI odometry dataset, and our dead-reckoning inertial method based only on the IMU accurately estimates 3D position, velocity, orientation of the vehicle and self-calibrates the IMU biases. We achieve on average a 1.10% translational error and the algorithm competes with top-ranked methods which, by contrast, use LiDAR or stereo vision. We make our implementation open-source at: https://github.com/mbrossar/ai-imu-dr

Mots clés

deep learning invariant extended Kalman filter KITTI dataset inertial navigation inertial measurement unit localization

Domaines

Robotique [cs.RO] Automatique / Robotique Machine Learning [stat.ML]

Fichier principal

main.pdf (1.06 Mo)

01data.pdf (453.82 Ko)

01gt.pdf (152.75 Ko)

01p_imu.pdf (161.15 Ko)

01p_rinsw.pdf (160.84 Ko)

01p_rinsw_sans_alignment.pdf (161.31 Ko)

01p_rinsw_sans_deep.pdf (161.23 Ko)

02data.pdf (107.08 Ko)

02gt.pdf (145.14 Ko)

02p.pdf (178.24 Ko)

02p_imls.pdf (153.83 Ko)

02p_imu.pdf (155.79 Ko)

02p_orb.pdf (153.65 Ko)

02p_rinsw.pdf (153.58 Ko)

07gt.pdf (158.89 Ko)

07p_imls.pdf (164.15 Ko)

07p_imu.pdf (161.1 Ko)

07p_orb.pdf (163.2 Ko)

07p_rinsw.pdf (165.4 Ko)

08gt.pdf (168.4 Ko)

08gt2.pdf (40.39 Ko)

08imu.pdf (175.86 Ko)

08nous.pdf (169.75 Ko)

09gt.pdf (154.15 Ko)

09p_imls.pdf (161.79 Ko)

09p_imu.pdf (161.81 Ko)

09p_orb.pdf (162.13 Ko)

09p_rinsw.pdf (161.81 Ko)

car2.png (71.17 Ko)

main-figure0.dep (112 B)

main-figure0.dpth (0 B)

main-figure0.log (94.37 Ko)

main-figure0.md5 (62 B)

main-figure0.pdf (77.1 Ko)

main-figure1.dpth (0 B)

main-figure1.log (94.57 Ko)

main-figure1.md5 (62 B)

main-figure1.pdf (100.47 Ko)

main-figure10.dpth (0 B)

main-figure10.log (95.63 Ko)

main-figure10.md5 (62 B)

main-figure10.pdf (3.57 Ko)

main-figure11.dpth (0 B)

main-figure11.log (95.63 Ko)

main-figure11.md5 (62 B)

main-figure11.pdf (3.57 Ko)

main-figure12.dpth (0 B)

main-figure12.log (95.63 Ko)

main-figure12.md5 (62 B)

main-figure12.pdf (3.57 Ko)

main-figure13.dpth (0 B)

main-figure13.log (95.63 Ko)

main-figure13.md5 (62 B)

main-figure13.pdf (3.57 Ko)

main-figure14.dpth (0 B)

main-figure14.log (95.63 Ko)

main-figure14.md5 (62 B)

main-figure14.pdf (3.57 Ko)

main-figure15.dpth (0 B)

main-figure15.log (95.63 Ko)

main-figure15.md5 (62 B)

main-figure15.pdf (3.57 Ko)

main-figure16.dpth (0 B)

main-figure16.log (95.63 Ko)

main-figure16.md5 (62 B)

main-figure16.pdf (3.57 Ko)

main-figure17.dpth (0 B)

main-figure17.log (95.63 Ko)

main-figure17.md5 (62 B)

main-figure18.dep (116 B)

main-figure18.dpth (0 B)

main-figure18.log (96.19 Ko)

main-figure18.md5 (62 B)

main-figure18.pdf (77.1 Ko)

main-figure19.dpth (0 B)

main-figure19.log (96.38 Ko)

main-figure19.md5 (62 B)

main-figure19.pdf (100.47 Ko)

main-figure2.dpth (0 B)

main-figure2.log (94.56 Ko)

main-figure2.md5 (62 B)

main-figure2.pdf (51.82 Ko)

main-figure20.dpth (0 B)

main-figure20.log (96.37 Ko)

main-figure20.md5 (62 B)

main-figure20.pdf (51.82 Ko)

main-figure21.dpth (0 B)

main-figure21.log (96.27 Ko)

main-figure21.md5 (62 B)

main-figure21.pdf (53.19 Ko)

main-figure22.dep (155 B)

main-figure22.dpth (0 B)

main-figure22.log (96.23 Ko)

main-figure22.md5 (62 B)

main-figure22.pdf (169.9 Ko)

main-figure23.dep (155 B)

main-figure23.dpth (0 B)

main-figure23.log (96.23 Ko)

main-figure23.md5 (62 B)

main-figure23.pdf (170.97 Ko)

main-figure24.dep (152 B)

main-figure24.dpth (0 B)

main-figure24.log (96.22 Ko)

main-figure24.md5 (62 B)

main-figure24.pdf (85.24 Ko)

main-figure25.dep (150 B)

main-figure25.dpth (0 B)

main-figure25.log (96.75 Ko)

main-figure25.md5 (62 B)

main-figure25.pdf (230.45 Ko)

main-figure26.dep (149 B)

main-figure26.dpth (0 B)

main-figure26.log (96.22 Ko)

main-figure26.md5 (62 B)

main-figure26.pdf (182.2 Ko)

main-figure27.dep (180 B)

main-figure27.dpth (0 B)

main-figure27.log (96.69 Ko)

main-figure27.md5 (62 B)

main-figure27.pdf (104.33 Ko)

main-figure3.dpth (0 B)

main-figure3.log (94.46 Ko)

main-figure3.md5 (62 B)

main-figure3.pdf (53.19 Ko)

main-figure4.dep (150 B)

main-figure4.dpth (0 B)

main-figure4.log (94.41 Ko)

main-figure4.md5 (62 B)

main-figure4.pdf (169.9 Ko)

main-figure5.dep (150 B)

main-figure5.dpth (0 B)

main-figure5.log (94.41 Ko)

main-figure5.md5 (62 B)

main-figure5.pdf (170.97 Ko)

main-figure6.dep (148 B)

main-figure6.dpth (0 B)

main-figure6.log (94.41 Ko)

main-figure6.md5 (62 B)

main-figure6.pdf (85.24 Ko)

main-figure7.dep (145 B)

main-figure7.dpth (0 B)

main-figure7.log (94.94 Ko)

main-figure7.md5 (62 B)

main-figure7.pdf (230.45 Ko)

main-figure8.dep (144 B)

main-figure8.dpth (0 B)

main-figure8.log (94.41 Ko)

main-figure8.md5 (62 B)

main-figure8.pdf (182.21 Ko)

main-figure9.dep (174 B)

main-figure9.dpth (0 B)

main-figure9.log (94.88 Ko)

main-figure9.md5 (62 B)

main-figure9.pdf (104.33 Ko)

main.auxlock (29 B)

main.blg (1.94 Ko)

main.log (76.4 Ko)

main.out (1.91 Ko)

main.synctex.gz (194.01 Ko)

orcidlogo.pdf (3.57 Ko)

Origine	Fichiers produits par l'(les) auteur(s)

Martin Brossard : Connectez-vous pour contacter le contributeur

https://hal.science/hal-02097099

Soumis le : samedi 20 avril 2019-16:57:04

Dernière modification le : mardi 27 août 2024-12:50:21

Dates et versions

hal-02097099 , version 1 (11-04-2019)

hal-02097099 , version 2 (20-04-2019)

Identifiants

HAL Id : hal-02097099 , version 2
ARXIV : 1904.06064
DOI : 10.1109/TIV.2020.2980758

Citer

Martin Brossard, Axel Barrau, Silvère Bonnabel. AI-IMU Dead-Reckoning. IEEE Transactions on Intelligent Vehicles, 2020, ⟨10.1109/TIV.2020.2980758⟩. ⟨hal-02097099v2⟩

Exporter

BibTeX XML-TEI Dublin Core DC Terms EndNote DataCite

Collections

ENSMP ENSMP_CAOR PARISTECH TDS-MACS PSL ENSMP_DR

1036 Consultations

2538 Téléchargements

AI-IMU Dead-Reckoning

Résumé

Mots clés

Domaines

Dates et versions

Identifiants

Citer

Exporter

Collections

Altmetric

Partager