Vio based on VINS-Mono

Desicription： This is a code base developed for reference purposes. It is based on the VINS-Mono framework but does not rely on ROS, Ceres, or G2o. The code is quite basic and aims to demonstrate common algorithms in SLAM optimization, including backend LM algorithms, sliding window algorithms, robust kernel functions, etc., using only the Eigen library.

Installation Dependencies：

1. pangolin: https://github.com/stevenlovegrove/Pangolin

2. opencv

3. Eigen

4. Ceres(2.1.0): The VINS initialization part uses Ceres for Structure from Motion (SfM), so we still need to rely on Ceres.

Compiling the Code

mkdir vislam
cd vislam
mkdir build 
cd build
cmake ..
make

Running the Code

1. CurveFitting Example to Verify Our Solver.

cd build
../bin/testCurveFitting 

2. VINS-Mono on Euroc Dataset

cd build
../bin/run_euroc /home/dataset/EuRoC/MH-05/mav0/ ../config/

3. Validation Results

evo package

evo_ape euroc euroc_mh05_groundtruth.csv pose_output.txt -a -p

Results from Euroc-MH05