End-to-end learning framework for IMU-based 6-DOF odometry

João Paulo Silva Do Monte Lima, Hideaki Uchiyama, Rin Ichiro Taniguchi

Research output: Contribution to journalArticle

Abstract

This paper presents an end-to-end learning framework for performing 6-DOF odometry by using only inertial data obtained from a low-cost IMU. The proposed inertial odometry method allows leveraging inertial sensors that are widely available on mobile platforms for estimating their 3D trajectories. For this purpose, neural networks based on convolutional layers combined with a two-layer stacked bidirectional LSTM are explored from the following three aspects. First, two 6-DOF relative pose representations are investigated: one based on a vector in the spherical coordinate system, and the other based on both a translation vector and an unit quaternion. Second, the loss function in the network is designed with the combination of several 6-DOF pose distance metrics: mean squared error, translation mean absolute error, quaternion multiplicative error and quaternion inner product. Third, a multi-task learning framework is integrated to automatically balance the weights of multiple metrics. In the evaluation, qualitative and quantitative analyses were conducted with publicly-available inertial odometry datasets. The best combination of the relative pose representation and the loss function was the translation and quaternion together with the translation mean absolute error and quaternion multiplicative error, which obtained more accurate results with respect to state-of-the-art inertial odometry techniques.

Original languageEnglish
Article number3777
JournalSensors (Switzerland)
Volume19
Issue number17
DOIs
Publication statusPublished - Sep 1 2019

Fingerprint

quaternions
learning
Learning
Weights and Measures
Costs and Cost Analysis
spherical coordinates
estimating
platforms
Trajectories
trajectories
Neural networks
evaluation
sensors
Sensors
products
Costs
Datasets

All Science Journal Classification (ASJC) codes

  • Analytical Chemistry
  • Biochemistry
  • Atomic and Molecular Physics, and Optics
  • Instrumentation
  • Electrical and Electronic Engineering

Cite this

End-to-end learning framework for IMU-based 6-DOF odometry. / Lima, João Paulo Silva Do Monte; Uchiyama, Hideaki; Taniguchi, Rin Ichiro.

In: Sensors (Switzerland), Vol. 19, No. 17, 3777, 01.09.2019.

Research output: Contribution to journalArticle

@article{5a092523789d4ee4bed16fea0c420963,
title = "End-to-end learning framework for IMU-based 6-DOF odometry",
abstract = "This paper presents an end-to-end learning framework for performing 6-DOF odometry by using only inertial data obtained from a low-cost IMU. The proposed inertial odometry method allows leveraging inertial sensors that are widely available on mobile platforms for estimating their 3D trajectories. For this purpose, neural networks based on convolutional layers combined with a two-layer stacked bidirectional LSTM are explored from the following three aspects. First, two 6-DOF relative pose representations are investigated: one based on a vector in the spherical coordinate system, and the other based on both a translation vector and an unit quaternion. Second, the loss function in the network is designed with the combination of several 6-DOF pose distance metrics: mean squared error, translation mean absolute error, quaternion multiplicative error and quaternion inner product. Third, a multi-task learning framework is integrated to automatically balance the weights of multiple metrics. In the evaluation, qualitative and quantitative analyses were conducted with publicly-available inertial odometry datasets. The best combination of the relative pose representation and the loss function was the translation and quaternion together with the translation mean absolute error and quaternion multiplicative error, which obtained more accurate results with respect to state-of-the-art inertial odometry techniques.",
author = "Lima, {Jo{\~a}o Paulo Silva Do Monte} and Hideaki Uchiyama and Taniguchi, {Rin Ichiro}",
year = "2019",
month = "9",
day = "1",
doi = "10.3390/s19173777",
language = "English",
volume = "19",
journal = "Sensors",
issn = "1424-3210",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "17",

}

TY - JOUR

T1 - End-to-end learning framework for IMU-based 6-DOF odometry

AU - Lima, João Paulo Silva Do Monte

AU - Uchiyama, Hideaki

AU - Taniguchi, Rin Ichiro

PY - 2019/9/1

Y1 - 2019/9/1

N2 - This paper presents an end-to-end learning framework for performing 6-DOF odometry by using only inertial data obtained from a low-cost IMU. The proposed inertial odometry method allows leveraging inertial sensors that are widely available on mobile platforms for estimating their 3D trajectories. For this purpose, neural networks based on convolutional layers combined with a two-layer stacked bidirectional LSTM are explored from the following three aspects. First, two 6-DOF relative pose representations are investigated: one based on a vector in the spherical coordinate system, and the other based on both a translation vector and an unit quaternion. Second, the loss function in the network is designed with the combination of several 6-DOF pose distance metrics: mean squared error, translation mean absolute error, quaternion multiplicative error and quaternion inner product. Third, a multi-task learning framework is integrated to automatically balance the weights of multiple metrics. In the evaluation, qualitative and quantitative analyses were conducted with publicly-available inertial odometry datasets. The best combination of the relative pose representation and the loss function was the translation and quaternion together with the translation mean absolute error and quaternion multiplicative error, which obtained more accurate results with respect to state-of-the-art inertial odometry techniques.

AB - This paper presents an end-to-end learning framework for performing 6-DOF odometry by using only inertial data obtained from a low-cost IMU. The proposed inertial odometry method allows leveraging inertial sensors that are widely available on mobile platforms for estimating their 3D trajectories. For this purpose, neural networks based on convolutional layers combined with a two-layer stacked bidirectional LSTM are explored from the following three aspects. First, two 6-DOF relative pose representations are investigated: one based on a vector in the spherical coordinate system, and the other based on both a translation vector and an unit quaternion. Second, the loss function in the network is designed with the combination of several 6-DOF pose distance metrics: mean squared error, translation mean absolute error, quaternion multiplicative error and quaternion inner product. Third, a multi-task learning framework is integrated to automatically balance the weights of multiple metrics. In the evaluation, qualitative and quantitative analyses were conducted with publicly-available inertial odometry datasets. The best combination of the relative pose representation and the loss function was the translation and quaternion together with the translation mean absolute error and quaternion multiplicative error, which obtained more accurate results with respect to state-of-the-art inertial odometry techniques.

UR - http://www.scopus.com/inward/record.url?scp=85073352667&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85073352667&partnerID=8YFLogxK

U2 - 10.3390/s19173777

DO - 10.3390/s19173777

M3 - Article

C2 - 31480413

AN - SCOPUS:85073352667

VL - 19

JO - Sensors

JF - Sensors

SN - 1424-3210

IS - 17

M1 - 3777

ER -