vins_mono_cg

Modified version of VINS-Mono (commit 9e657be on Jan 9, 2019), a Robust and Versatile Monocular Visual-Inertial State Estimator.

VINS-Mono uses an optimization-based sliding window formulation for providing high-accuracy visual-inertial odometry. It features efficient IMU pre-integration with bias correction, automatic estimator initialization, online extrinsic calibration, failure detection and recovery, loop detection, and global pose graph optimization, map merge, pose graph reuse, online temporal calibration, rolling shutter support.

[1] VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator
[2] Online Temporal Calibration for Monocular Visual-Inertial Systems

• Formula Derivation and Code Analysis of VINS-Mono (RG)

Prerequisites

tested on Ubuntu 16.04 (ROS Kinetic) and Ubuntu 18.04 (ROS Melodic)

• Eigen 3.3.3

• Ceres Solver

Build

catkin_make -j2
# or
catkin build

Run

with Dataset

• EuRoC MAV dataset MH_01_easy.bag

  roslaunch vins_estimator euroc.launch
  rosbag play <YOUR_PATH_TO_DATASET>/MH_01_easy.bag
  

with live camera

• with MYNTEYE-S1030

  roslaunch mynt_eye_ros_wrapper mynteye.launch
  roslaunch vins_estimator mynteye_s1030_mono.launch
  

  

• Ubuntu 16.04 下 VINS-Mono 的安装和使用(RealSense ZR300)

  roslaunch maplab_realsense maplab_realsense.launch
  roslaunch vins_estimator realsense_fisheye.launch
  

with Docker

• make sure ROS and docker are installed on your machine
• add your account to docker group by sudo usermod -aG docker $YOUR_USER_NAME
• run

  cd docker
  make build
  ./run.sh LAUNCH_FILE_NAME   # ./run.sh euroc.launch
  

• modified the code, simply run ./run.sh LAUNCH_FILE_NAME after your changes

ROS Graph

• rqt_graph

  

Evaluation

Evaluate the output trajectory vins_result_loop.tum with ground truth trajectory in the standard dataset (e.g. for EuRoC MAV dataset, the ground truth file is <sequence>/mav0/state_groundtruth_estimate0/data.csv ) using the evo tools.

1. copy the ground truth file data.csv to the directory as same to vins_result_loop.tum

2. evaluate (APE & RPE)

   evo_traj euroc data.csv --save_as_tum # --> data.tum
   evo_ape tum data.tum vins_result_loop.tum --align --plot
   evo_rpe tum data.tum vins_result_loop.tum --align --plot
   
   # or
   
   evo_ape euroc data.csv vins_result_loop.tum --align --plot
   evo_rpe euroc data.csv vins_result_loop.tum --align --plot
   

3. get results

   APE w.r.t. translation part (m)
   (with SE(3) Umeyama alignment)
        max	0.157368
       mean	0.081223
     median	0.076672
        min	0.021434
       rmse	0.086200
        sse	7.809322
        std	0.028865
   

   

Tutorial

• VINS-Mono代码分析总结 by Xiaobuyi
• VINS-Mono issues 14: Question about mid-point integration in integration_base.h

Related Code

• HKUST-Aerial-Robotics/VINS-Fusion: An optimization-based multi-sensor state estimator
• gaowenliang/vins_so: A Robust and Versatile Visual-Inertial State Estimator support Omnidirectional Camera and/or Stereo Camera
• castiel520/VINS-Mono: VINS-Mono中文注释
• QingSimon/VINS-Mono-code-annotation: VINS-Mono代码注释以及公式推导
• heguixiang/Android-VINS: a version of HKUST-Aerial-Robotics/VINS-Mono running on Android OS
• pjrambo/VINS-Fusion-gpu: a version of VINS-Fusion with GPU acceleration