Fast and Robust Online Initialization of Monocular Visual-Inertial Odometry via One-Dimensional Cost Approximation
IEEE Transactions on Robotics · 2026 · 42 · 2289-2304
DOI: 10.1109/tro.2026.3686255 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Robotics, 2026. Verfasst von Jiseock Kang und 5 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Robot Industry Technology Development, Korea Institute for Advancement of Technology, Ministry of Trade.
Abstract (Originalsprache)
In this article, we present a fast and accurate initialization algorithm of monocular visual-inertial odometry (MVIO). Our algorithm enables MVIO initialization within under-second motion time (0.45 s for Q1 and 0.7 s for Q2) without divergence for 1500 random runs in the EuroC-MAV and TUM-VI dataset. Our algorithm is based on almost-exact one-dimensional (1-D) approximation of the loosely coupled MVIO (LC-MVIO) problem, which is obtained by approximately decoupling optimization of vision rotation error from the original LC-MVIO optimization. Thanks to the efficiency of 1-D domain exploration, this approximated LC-MVIO is feasible without the initial guess, therefore can be used as an initializer of the original MVIO. Based on this 1-D formulation, we design our initializer with exhaustive correctness guarantee, a novel estimator guarantee we propose in this work, that certifies the absence of multiple, ambiguous solution modes with similar likelihood cost. Our algorithm is validated through Monte-Carlo simulation, real-world dataset experiment, and real-vehicle experiment. An open-source MVIO pipeline with the suggested initialization algorithm is also provided as a supplement.
Förderung
- Robot Industry Technology Development
- Korea Institute for Advancement of Technology
- Ministry of Trade
- Industry & Energy
- Korea Institute of Marine Science & Technology Promotion
- Ministry of Oceans and Fisheries
Zitieren
APA
Kang, J., Choe, J., Kong, D., Yoon, B., Cho, J., & Lee, D. (2026). Fast and Robust Online Initialization of Monocular Visual-Inertial Odometry via One-Dimensional Cost Approximation. IEEE Transactions on Robotics, 42, 2289-2304. https://doi.org/10.1109/tro.2026.3686255
Vancouver
Kang J, Choe J, Kong D, Yoon B, Cho J, Lee D. Fast and Robust Online Initialization of Monocular Visual-Inertial Odometry via One-Dimensional Cost Approximation. IEEE Transactions on Robotics. 2026;42:2289-2304. doi:10.1109/tro.2026.3686255
BibTeX
@article{Kang2026,
author = {Kang, Jiseock and Choe, Jaeu and Kong, Doyoon and Yoon, Byoungkwon and Cho, Jaehwi and Lee, Dongjun},
title = {Fast and Robust Online Initialization of Monocular Visual-Inertial Odometry via One-Dimensional Cost Approximation},
journal = {IEEE Transactions on Robotics},
year = {2026},
volume = {42},
pages = {2289--2304},
doi = {10.1109/tro.2026.3686255}
}
EUVIMED