LINS---LiDAR-inertial-SLAM
This repository contains code for a tightly-coupled lidar-inertial odometry and mapping system for ROS compatible UGVs. The reason of fusing IMU and Lidar in a tightly-couple scheme is to handle feature-less environments where previous methods may fail. This work is built upon LIO-mapping, LeGO-LOAM and LOAM. The main contribution of this work is the lightweight lidar-inertial lidar odometry which produces robust and accurate odometry in real time using a 400-Hz IMU and a Velodyne VLP-16 Lidar. Extensive experiments show that with a robust lidar-inertial odometry, the localization and mapping performances are greatly improved in challenging environments. A demonstration of the system can be found here -> https://www.youtube.com/watch?v=Nmr1blC09qw&t=8s
Dependency
- ROS (tested with kinetic)
- gtsam (Georgia Tech Smoothing and Mapping library)
- OpenCV (tested with OpenCV 3.4)
Compile
You can use the following commands to download and compile the package.
cd ~/catkin_ws/src
git clone https://github.com/ChaoqinRobotics/LINS---LiDAR-inertial-SLAM.git
cd ..
catkin_make -j1
When you compile the code for the first time, you need to add "-j1" behind "catkin_make" for generating some message types. "-j1" is not needed for future compiling.
The system
Like LeGO-LOAM, LINS is speficifally optimized for a horizontally placed VLP-16 on a ground vehicle with a high-speed IMU (>=100Hz). It assumes there is always a ground plane in the scan.
Run the package
- Run the launch file:
roslaunch lins run_port_exp.launch
Notes: The parameter "/use_sim_time" is set to "true" for simulation, "false" to real robot usage.
- Play existing bag files:
rosbag play *.bag --clock --topic /velodyne_points /imu/data
Dataset
We provide a short piece of dataset for testing. The dataset can be founded here or here
source devel/setup.bash
roslaunch lins run_port_exp.launch
.rosbag play lidar_imu_dataset.bag --clock
.
Run Your Own Dataset
Before you run your own dataset, please make sure that the extrinsic parameters between LiDAR and IMU are estimated off-line and set in the config file (see exp_port.yaml). In this version of LINS, we assume the roll and pitch angles between LiDAR and IMU are zero, and users can set the yaw angles between them by
imu_misalign_angle
The IMU noice parameters should be estimted and set in this file, too.
# noice parameters
acc_n: 70000
gyr_n: 0.1
acc_w: 500
gyr_w: 0.05
# initial IMU biases
init_ba: !!opencv-matrix
rows: 3
cols: 1
dt: d
data: [-0.015774,0.143237,-0.0263845]
init_bw: !!opencv-matrix
rows: 3
cols: 1
dt: d
data: [-0.00275058,-0.000165954,0.00262913]
Below is the sensor we use in data collection.
Because this sensor set has small translation and rotation between LiDAR and IMU, the extrinsic parameters in the config file can be set to zero. But this may not be the case in your sensor sets.
Also, in this version of LINS, please make sure that the vehicle starts at the flat road, i.e., the roll and pitch angles are close to zero.
Cite LINS
Thank you for citing our LINS paper if you use any of this code:
@inproceedings{lins2020,
title={LINS: A Lidar-Inertial State Estimator for Robust and Efficient Navigation},
author={Qin, Chao and Ye, Haoyang and Pranata, Christian E and Han, Jun and Zhang, Shuyang and Liu, Ming},
booktitle={2020 International Conference on Robotics and Automation (ICRA)},
year={2020},
organization={IEEE}
}