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

• EuRoC MAV Dataset - Stereo + IMU (Visual-Inertial Odometry)
• TUM-VI Dataset - Stereo + IMU (Visual-Inertial Odometry)
• Intel RealSense D435/D435i - RGBD (Visual Odometry only)
  • Supports custom RGBD datasets with color/ and depth/ folders
  • Example configuration: config/d435.yaml
  • ⚠️ Note: RGBD + IMU fusion (VIO mode) is under development

🚀 Running Examples - Usage examples

📊 Performance Analysis Guide - Comprehensive performance evaluation and benchmarking

The source code is organized into the following directories:

• app/: Main application entry points
• src/:
  • database/: Data structures for Frame (including IMU data), MapPoint, and Feature
  • processing/: Core VIO modules, including Estimator, FeatureTracker, IMUHandler, and Optimizer
  • optimization/: Ceres Solver cost functions for visual reprojection errors and IMU pre-integration constraints
  • viewer/: Pangolin-based visualization
  • util/: Utility functions for configuration and data loading
• thirdparty/: External libraries (Ceres, Pangolin, Sophus, spdlog)
• config/: Configuration files for VO and VIO modes
• docs/: Detailed documentation guides

Contributions 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}
}