CURL-MAP (Paper)
We extend the continuous reconstruction and memory efficiency advantages of CURL to large-scale maps, introducing new capabilities for odometry and updatable mapping.
git cloue https://github.com/SenseRoboticsLab/CURL-MAP.git
cd CURL-MAP
- Build the image
CPU only
sudo docker build -f ./docker/curl_map_cpu.dockerfile --build-arg NUM_THREADS=12 -t curl_map:v0 .
With Nvidia GPU, can speed up rivz visualization. (NVIDIA Container Toolkit needs to be installed)
sudo docker build -f ./docker/curl_map_gpu.dockerfile --build-arg NUM_THREADS=12 -t curl_map:v0 .
- Enabling X Server Access with
xhost +
- Create a container
Please fill <directory on your host machine> in the following commands
CPU only
sudo docker run --name curl_map_container -it -e DISPLAY=$DISPLAY -e QT_X11_NO_MITSHM=1 -v /tmp/.X11-unix:/tmp/.X11-unix:rw --volume="$HOME/.Xauthority:/root/.Xauthority:rw" --network host --add-host host.docker.internal:host-gateway -v <directory on your host machine>:/home/user/data curl_map:v0 /bin/bash
With Nvidia GPU
sudo docker run --name curl_map_container -it --privileged --gpus all -e="NVIDIA_VISIBLE_DEVICES=all" -e="NVIDIA_DRIVER_CAPABILITIES=graphics,compute,utility" -e DISPLAY=$DISPLAY -e QT_X11_NO_MITSHM=1 -v /tmp/.X11-unix:/tmp/.X11-unix:rw --volume="$HOME/.Xauthority:/root/.Xauthority:rw" --network host --add-host host.docker.internal:host-gateway -v <directory on your host machine>:/home/user/data curl_map:v0 /bin/bash
- Create a new terminal inside the container
sudo docker exec -it curl_map_container /bin/bash
- Run CURL-MAP for Livox
roslaunch --sigint-timeout=60000 --sigterm-timeout=60000 curl_slam curl_slam.launch
- Publish example data
roslaunch patchworkpp run_patchwork_kitti.launch data_dir:=/home/user/catkin_ws/src/CURL-MAP/example_data/kitti/sequences seq:=05
- Publish rosbag data
Modify the LiDAR topic name inside /home/user/catkin_ws/src/patchwork-plusplus-ros-modified/launch/demo_newer_college.launch
roslaunch patchworkpp demo_newer_college.launch
rosbag play <your rosbag>
roslaunch curl_slam curl_reconstruction.launch
Click refresh button shows in the image
Then w and degree can be used to adjust density and detail of the reconstruction respectively.
- Clean the config file, remove useless parameters
- Add comments for parameters
- Clean the code
Newer College Dataset
KITTI Dataset
KITTI-05
CURL-MAP is able to achieve real-time performance on a CPU in the FAST mode.
The conformal parameterisation module in this project is a C++ reimplementation of the original MATLAB code from
garyptchoi/rectangular-conformal-map,
which is licensed under the Apache License 2.0.