SIMPLE-C: Uncertainty-aware Multi-LiDAR Extrinsic Calibration using a Simple Dynamic Target in Moving Feature-Sparse Environments

Docker Repository for Easy Online Capability

Installation

• Install Docker. (Or an alternative such as podman. However, all following commands are given for Docker.)
• Change directory into the Docker repo

Building and running the docker container

Automatic (re)build and run using the script rebuild.sh

• The script will first copy the online calibration repository, then build the container and finally start it.
• Set the mount environment for the online calibration repository in the docker container: export REPO_DIR=<path to your repository directory>
• Mount a data directory to the docker container: export DATA_DIR=<path to your data directory> (Log Files wil be stored)
• For troubleshooting, see the details in the following sections about building/running manually.
• To avoid starting the default launch file in the docker container, execute ./rebuild.sh.
• You can start another terminal in the docker container using sudo docker exec -it calibration_tool_1 bash
• Everytime you start a new terminal, you have to source ROS: source /opt/ros/humle/setup.bash

Building manually

• Ensure you have access to the internet.
• Put/copy/link the online calibration repository into the directory ./build/files/ros_pkgs. (See also section Adding more ROS packages.)
• Run sudo docker build --tag calibration_tool ./build/ (This will build an image based on the instructions in the Dockerfile.)
• If APT errors (especially resolving URLs) occur during build, try to update & upgrade APT on the host system.
• The convention used in the rebuild.sh script is to call the image calibration_tool, and the container calibration_tool_1.

Running the container manually

• The following is also automatically done in
• Ensure all configurations in the ./config/ directory are correct, see section above.
• Run the container with the following settings:
  • --name calibration_tool_1 (convention for the container name, see section Build)
  • --net host to use the host's network. Avois tedious port-forwarding.
  • --volume ./config:/CONFIG:ro mount the config directory. (Linking the rslidar subdirectory is performed in the Dockerfile.)
  • A full docker run command could look like this: sudo docker run --rm --name calibration_tool_1 --net host --volume ./config:/CONFIG:ro -it calibration_tool
• To check whether packets are correctly forwarded to the container, or to debug the UDP ports, run sudo docker exec -it calibration_tool_1 tcpdump -nc20 (in a different terminal). You should see lines such as 12:43:36.683623 IP 192.168.1.203.6699 > 172.17.0.2.4499: UDP, length 1210 immediately.
• By default (see ./build/files/entrypoint.sh), ./config/launch.py is launched. You can overwrite this by building/starting the container manually or running ./rebuild.sh bash and then running single nodes from different terminals. Find sample commands for this in ./commands.txt.

Configuration

• The directory ./config/ contains configurations for all ROS nodes, as well as the ROS2 launch file. It is mounted as a Docker volume to /CONFIG in the container, which means that updating its contents is possible without rebuilding the container. (Note that ROS nodes (usually) have to be restarted for changes to take effect!)
• The subdirectory ./config/rslidar is linked to the RSLidar SDK's configuration directory inside the docker container. Therefore, find the RSLidar config in this directory, where Lidar network ports and topic names have to be set.
• All ROS parameters for the calibration node are set in the ROS2 launch file ./config/launch.py. There is no automatic fallback or connection to the default_parameters.yaml file from the calibration repo!
• If you do not want to run the calibrator live but on ROS-bag data, please comment the Robosense Lidar SDK out
• If you want to use different LiDAR sensors, comment the Robosense Lidar SKD out or replace by your LiDAR SDK/Driver or start your LiDAR nodes separately

Host Connection Setup

• Before starting the docker container to read data from sensors, make sure that wired connection is correct and turned on
• The host pc's ethernet properties should be changed to:
  • static IPv4 adress: 192.168.1.102
  • subnet mask: 255.255.255.0
  • gateway: 192.168.1.1
• Ensure that your host system is receiving Lidar data from the network. You can verify this using sudo tcpdump -n -c30 -i <ADAPTER>. You should see lines such as 14:34:58.823814 IP 192.168.1.203.6699 > 192.168.1.102.4499: UDP, length 1210 immediately.
• On Ubuntu, go to the network settings and make sure to have the respective Wired network enabled. If this is disabled, the Docker container won't receive packets, even if they appear in tcpdump!
• Keep in mind to check the correct sensor network ports! The MSOP_TIMEOUT error does not always trigger, even if you have sensors with the wrong ports connected! (This will result in no frames inside the calbration node.)

Adding more ROS packages

• The RSLidar SDK is installed by commands in the Dockerfile, as it requires several setup steps.
• It is of course possible to add further packages this way as well.
• If you want to include packages whose source code can be built with a simple colcon build command, then move them into the ./build/files/ros_pkgs/ directory. The contents of it are copied into /ROS_WORKSPACE/src/ before running colcon build.
• (This is how the online calibration package is built. Its source is copied to ./build/files/ros_pkgs/ in the script ./rebuild.sh.)
• Remember to add new packages to ./config/launch.py, the default ROS2 launch file, if they should be run automatically.

Web visualization

The standard launch file will also start an HTTP server hosting the web visualization interface on port 8000. It is copied into the Docker image from ./build/files/index.html. See the respective web-visualization git repository for build instructions.