peterWon/D-LIOM

D-LIOM: Tightly-coupled Direct LiDAR-Inertial Odometry and Mapping

Our paper: we have corrected some typos and errors in Section III-D of the previous version of the paper, the revised version can be accessed here. When amending our paper, I would like to say thanks to Sky Shaw, who has found my errors and warmly provided his suggestions.

Related video: A running demo can be found at https://youtu.be/21J2QLUQbno or at https://www.bilibili.com/video/BV14y4y157du.

Sensor suite supported

• 6-axis IMU + Velodyne/Ouster/Robosense LiDAR (Livox LiDAR can be also supported if the data of which is tansformed to the format of velodyne's, i.e., PointXYZRT.)

Prerequisites

• Install depends follow the instructions of "cartographer" and "cartographer_ros" respectively.

• Build OpenCV 3.3.1 (other version may work fine if it includes implementation of SURF keypoint detector) from source with "opencv_contrib"

• Build GTSAM 4.0.2 with compile options as follows:

  cmake -DGTSAM_USE_SYSTEM_EIGEN=ON -DGTSAM_WITH_TBB=OFF -DGTSAM_BUILD_WITH_MARCH_NATIVE=OFF -DGTSAM_COMPILE_OPTIONS_PUBLIC="" ..

• Build PCL 1.8.0 with compile options as follows:

  cmake -DPCL_ENABLE_SSE=OFF ..

Build

• Clone this repository in your ROS workspace

• Set CMake variable "OpenCV_DIR" in the root CMakeLists.txt of "cartographer".

• Return to the root folder of your ROS workspace (which has the same directory as this "README" file) and run:

  catkin_make_isolated
  

Run an online demo

TONGJI Dataset

• Some demo sequences can be downloaded via:

  • TONGJI dataset: link, pwd: hfrl

• roslaunch dlio demo_dlio_tongji.launch bag_filename:=BAG_FILE_PATH where BAG_FILE_PATH is the full path of the downloaded bag file in your device

• Some results on the self-collected TONGJI dataset.

  
  

NTU-VIRAL

• The dataset is available online at NTU-VIRAL

• ```roslaunch dlio demo_dlio_viral.launch bag_filename:=BAG_FILE_PATH```
  

  where BAG_FILE_PATH is the full path of the downloaded bag file in your device

• Results on VIRAL
  

  
  
  

KAIST-Complex Urban Dataset

• The dataset is available online at KAIST

• Play KAIST bag follow the instruction: file_player

• roslaunch dlio demo_dlio_kaist.launch

• Results on KAIST "Urban-09" and "Urban-10"

  

Your own data

• Write a urdf file of your device and save it to folder 'urdf'
• Write a configuration file of your data and save it to 'config'
• Write a launch file of your data and save it to 'launch'
• Run your bag as above examples

Run offline and view mapping result

Launch dlio offline

roslaunch dlio dlio_offline_tongji bag_filenames:=BAG_FILE_PATH

View mapping result

roslaunch dlio dlio_map_viewer_rectified urdf_filename:=tongji pose_graph_pb_filename:=PG_FILENAME range_data_pb_filename:=RD_FILENAME rate:=100

where PG_FILENAME and RD_FILENAME are binary pbstream files saved to disk when launch the offline script above.

Acknowledgements

• The authors of cartographer
• The authors of cartographer_ros
• The authors of LIO-SAM
• The authors of VINS-Mono
• The authors of VIO-Doc