Statistical Uncertainty Learning for Robust Visual-Inertial State Estimation
Statistical Uncertainty Learning for Robust Visual-Inertial State Estimation
This is a lightweight stereo visual-inertial odometry (VIO) project designed for real-time performance. It utilizes feature tracking, IMU pre-integration, sliding window optimization with Ceres Solver, and Pangolin for visualization.
This project implements the statistical uncertainty learning method proposed in:
Statistical Uncertainty Learning for Robust Visual-Inertial State Estimation
Seungwon Choi, Donggyu Park, Seo-Yeon Hwang, Tae-Wan Kim
arXiv:2510.01648
If you use this work in your research, please cite:
@misc{choi2025statistical,
title={Statistical Uncertainty Learning for Robust Visual-Inertial State Estimation},
author={Seungwon Choi and Donggyu Park and Seo-Yeon Hwang and Tae-Wan Kim},
year={2025},
eprint={2510.01648},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2510.01648}
}This project is licensed under the MIT License - see the LICENSE file for details.
https://github.com/93won/vio_ros_wrapper
Installation Guide - Complete installation instructions for both Docker and native builds
Dataset Download Guide - Dataset download and preparation
config/d435.yamlRunning Examples - Usage examples
Performance Analysis Guide - Comprehensive performance evaluation and benchmarking
The source code is organized into the following directories:
app/: Main application entry pointssrc/:database/: Data structures for Frame (including IMU data), MapPoint, and Featureprocessing/: Core VIO modules, including Estimator, FeatureTracker, IMUHandler, and Optimizeroptimization/: Ceres Solver cost functions for visual reprojection errors and IMU pre-integration constraintsviewer/: Pangolin-based visualizationutil/: Utility functions for configuration and data loadingthirdparty/: External libraries (Ceres, Pangolin, Sophus, spdlog)config/: Configuration files for VO and VIO modesdocs/: Detailed documentation guidesContributions are welcome! Please feel free to submit issues and pull requests.
This project's IMU pre-integration implementation is based on the following paper:
@article{forster2016manifold,
author = {Forster, Christian and Carlone, Luca and Dellaert, Frank and Scaramuzza, Davide},
year = {2016},
month = {08},
title = {On-Manifold Preintegration for Real-Time Visual-Inertial Odometry},
volume = {33},
journal = {IEEE Transactions on Robotics},
doi = {10.1109/TRO.2016.2597321}
}No open issues yet, or sync has not completed.