Skip to content

Latest commit

 

History

History
executable file
·
129 lines (83 loc) · 4.13 KB

README.md

File metadata and controls

executable file
·
129 lines (83 loc) · 4.13 KB
Raw

LI-Calib

Overview

LI-Calib is a toolkit for calibrating the 6DoF rigid transformation and the time offset between a 3D LiDAR and an IMU. It's based on continuous-time batch optimization. IMU-based cost and LiDAR point-to-surfel distance are minimized jointly, which renders the calibration problem well-constrained in general scenarios.

Prerequisites

  • ROS (tested with Kinetic and Melodic)

    sudo apt-get install ros-melodic-pcl-ros ros-melodic-velodyne-msgs

  • Ceres (tested with version 1.14.0)

  • Kontiki (Continuous-Time Toolkit)

  • Pangolin (for visualization and user interface)

  • ndt_omp

Note that Kontiki and Pangolin are included in the thirdparty folder.

Install

Clone the source code for the project and build it.

# init ROS workspace
mkdir -p ~/catkin_li_calib/src
cd ~/catkin_li_calib/src
catkin_init_workspace

# Clone the source code for the project and build it. 
git clone https://github.com/APRIL-ZJU/lidar_IMU_calib

# ndt_omp
wstool init
wstool merge lidar_IMU_calib/depend_pack.rosinstall
wstool update
# Pangolin
cd lidar_imu_calib_beta
./build_submodules.sh
## build
cd ../..
catkin_make
source ./devel/setup.bash

Examples

Currently the LI-Calib toolkit only support VLP-16 but it is easy to expanded for other LiDARs.

Run the calibration:

./src/lidar_IMU_calib/calib.sh

The options in calib.sh the have the following meaning:

  • bag_path path to the dataset.
  • imu_topic IMU topic.
  • bag_start the relative start time of the rosbag [s].
  • bag_durr the duration for data association [s].
  • scan4map the duration for NDT mapping [s].
  • timeOffsetPadding maximum range in which the timeoffset may change during estimation [s].
  • ndtResolution resolution for NDT [m].

UI

Following the step:

  1. Initialization

  2. DataAssociation

    (The users are encouraged to toggle the show_lidar_frame for checking the odometry result. )

  3. BatchOptimization

  4. Refinement

  5. Refinement

  6. ...

  7. (you cloud try to optimize the time offset by choose optimize_time_offset then run Refinement)

  8. SaveMap

All the cache results are saved in the location of the dataset.

Note that the toolkit is implemented with only one thread, it would response slowly while processing data. Please be patient

Dataset

3imu

Dataset for evaluating LI_Calib are available at here.

We utilize an MCU (stm32f1) to simulate the synchronization Pulse Per Second (PPS) signal. The LiDAR's timestamps are synchronizing to UTC, and each IMU captures the rising edge of the PPS signal and outputs the latest data with a sync signal. Considering the jitter of the internal clock of MCU, the external synchronization method has some error (within a few microseconds).

Each rosbag contains 7 topics:

/imu1/data          : sensor_msgs/Imu           
/imu1/data_sync     : sensor_msgs/Imu           
/imu2/data          : sensor_msgs/Imu           
/imu2/data_sync     : sensor_msgs/Imu           
/imu3/data          : sensor_msgs/Imu           
/imu3/data_sync     : sensor_msgs/Imu           
/velodyne_packets   : velodyne_msgs/VelodyneScan

/imu*/data are raw data and the timestamps are coincide with the received time.

/imu*/data_sync are the sync data, so do /velodyne_packets .

Credits

This code was developed by the APRIL Lab in Zhejiang University.

For researchers that have leveraged or compared to this work, please cite the following:

Jiajun Lv, Jinhong Xu, Kewei Hu, Yong Liu, Xingxing Zuo. Targetless Calibration of LiDAR-IMU System Based on Continuous-time Batch Estimation. IROS 2020. [arxiv]

License

The code is provided under the GNU General Public License v3 (GPL-3).