INS_ESKF_KITTI

I barely found GPS-IMU fusion localization algorithm using real world dataset on github,most of them are using data generated from gnss-imu-sim.So I developed ins_eskf_kitti,a GPS-IMU fusion localization algorithm using error-state kalman filter based on kitti dataset.I Use the formula that shared By Dr.Gao Xiang in Zhihu

Menu

• INS_ESKF_KITTI
  • Menu
  • System architecture (Improtant!)
  • Dependency
  • Install
  • Sample datasets
  • Run the package

System architecture (Improtant!)

I create a ROS-wrapper(for ROS-noetic),so the core algorithm is dependent from ROS,and easy for other developers to integrate into other platforms or ROS2. And I use /kitti/oxts/gps/vel and orientation in /kitti/oxts/imu/extract so I can initialize the whole system.BTW,I also plot the position in /kitti/oxts/imu/extract & oirentation in /kitti/oxts/imu/extract in rviz as a baseline th show the result of fusion algorithm.

Here in the whole system arcgitecture:

Dependency

ROS-notic

Eigen:

sudo apt-get install libeigen3-dev 

geographiclib :

sudo apt-get install libgeographic-dev

glog :

git clone https://github.com/google/glog.git
cd glog
mkdir build 
cd build
cmake ..
make

yaml-cpp:

sudo apt-get install libyaml-cpp-dev

Install

Use the following commands to download and compile the package. Before compiling the project , you have to edit the include/global_definition.h change the PROJECT_PATH to yours. Then

cd ~/catkin_ws/src
git clone https://github.com/leo6862/ins_eskf_kitti.git
cd ..
catkin_make

Sample datasets

Using a Kitti Dataset which has imu data frequency higher than 100HZ. here's a dataset from mine.You can download it from baidu net disk 提取码: wtu9.

Run the package

1. Run the launch file:

roslaunch ins_eskf ins_eskf.launch

2. Play existing bag files:

rosbag play your-bag.bag