VerFlex-Vio: Versatile and Flexible Visual-Inertial Odometry System

A high-performance Visual Inertial Odometry (VIO) system implemented in C++ with ROS integration, featuring real-time state estimation using camera and IMU data.

🚀 Features

• Multiple camera models: monocular and stereo camera models
• Dual operation modes: online/offline processing
• Flexible initialization: static/dynamic visual initialization
• Multiple estimators: support ESKF, Sqrt-ESKF solvers
• Multiple frontends: traditional KLT-based tracking & NN descriptor-based tracking
• Threading options: multi-threaded/single-threaded execution
• Schmidt ESKF: First clone pose anchoring to reduce trajectory drift
• Numerically stable: Joseph form covariance update for improved robustness
• Dual feature support: MSCKF features for batch updates & SLAM features for landmark-based updates
• Sequential ESKF: Sequential ESKF update for SLAM features to boost processing speed

🛠️ Installation

Prerequisites

• ROS Noetic (Ubuntu 20.04)
• CUDA 11.0+ (for neural feature extraction)
• C++17 compatible compiler
• CMake 3.15+

System Dependencies

# Install system dependencies
sudo apt update
sudo apt install -y \
    build-essential \
    cmake \
    git \
    libeigen3-dev \
    libopencv-dev \
    libgoogle-glog-dev \
    libgtest-dev \
    libceres-dev \
    libsophus-dev \
    python3-pip

Building from Source

# Clone the repository
git clone git@github.com:qqqGpe/VerFlex-Vio.git
cd catkin_ws

# Quickly build the project with script
./src/VerFlex-Vio/script/build_vio_backend.sh

# Source the workspace
source devel/setup.bash

🚀 Quick Start

1. Basic Usage with Euroc Dataset

# Launch VIO with Euroc dataset with offline rosbag
roslaunch vio euroc_serial_backend.launch

# Launch VIO with Euroc dataset with rosbag playback
roslaunch vio euroc_serial_backend_subscribe.launch

# Launch VIO with Intel Realsense Camera
roslaunch vio realsense_serial_online.launch

2. Batch simulatation with Euroc dataset

# Batch simulation and evaluation
source ./devel/setup.zsh

# Run all data
python3 src/vio_backend/script/run_vio_batch.py --align --dataset_dir ~/dataset/euroc_mav --cases all

# Run specified data
python3 src/vio_backend/script/run_vio_batch.py --align --dataset_dir ~/dataset/euroc_mav --cases MH_01_easy

# An example of running results for batch simulation on EuRoc Dataset
# ================================================================================
# VIO SIMULATION RESULTS SUMMARY
# ================================================================================
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | Dataset      |     RMSE |     Mean |   Median |      Std |      Min |      Max |      SSE |
# +==============+==========+==========+==========+==========+==========+==========+==========+
# | MH_01_easy   | 0.121111 | 0.105462 | 0.085949 | 0.059545 | 0.015374 | 0.355678 | 53.0098  |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | MH_02_easy   | 0.261663 | 0.22578  | 0.198678 | 0.132252 | 0.026258 | 0.753107 | 203.485  |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | MH_03_medium | 0.18343  | 0.163888 | 0.150588 | 0.082384 | 0.01175  | 0.571727 |  88.2884 |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | V1_01_easy   | 0.216476 | 0.18816  | 0.173854 | 0.10704  | 0.023342 | 0.41278  | 132.385  |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | V1_02_medium | 0.222966 | 0.201963 | 0.181061 | 0.094472 | 0.032893 | 0.811831 |  82.5253 |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | V2_01_easy   | 0.095293 | 0.085423 | 0.07872  | 0.042233 | 0.012368 | 0.234315 |  20.3046 |
# +--------------+----------+----------+----------+----------+----------+----------+----------+
# | V2_02_medium | 0.261529 | 0.245031 | 0.229532 | 0.091417 | 0.005754 | 0.542613 | 157.519  |
# +--------------+----------+----------+----------+----------+----------+----------+----------+

Note: This is a research-grade VIO system. For production use, additional testing and validation is recommended.