DISCLAIMER: This library is still under development. Be warned that some interfaces will be changed and/or extended in the future.

The VDB Mapping ROS Package is a ROS wrapper around VDB Mapping

This library requires OpenVDB as it is build around it. This library was initially developed using Version 5.0 and should work with all versions above.
Either use the apt package which will be automatically installed via rosdep or compile the package from source using the provided build instructions

Since the required VDB Mapping library is a plain C++ package, you cannot use catkin_make directly. Instead you have to use catkin build or catkin_make_isolated to build the workspace.

# source global ros
source /opt/ros/<your_ros_version>/setup.bash

# create a catkin workspace
mkdir -p ~/catkin_ws/src && cd ~/catkin_ws/src

# clone packages
git clone https://github.com/fzi-forschungszentrum-informatik/vdb_mapping
git clone https://github.com/fzi-forschungszentrum-informatik/vdb_mapping_ros

# install dependencies
sudo apt update
rosdep update
rosdep install --from-paths src --ignore-src -y

# build the workspace.  
catkin build

# source the workspace
source devel/setup.bash

The remote mapping module is at the moment mainly designed to provide an external access to the map, without saving and loading it in a different node. In our case we use it to visualize the map data on a remote base station which is then used to control the robots from far away and give the user an overview of the area the robot operates in. Basically it is a full additional mapping node which can access the data chunks generated by the main mapping on the robot. These can the be integrated into the remote mapping instance to generate the same or a similar map depending on the chosen parameter settings. Diving deeper into the remote mapping functionality you might stumble over the term of updates, overwrites and sections. These are different mechanics to update a remote mapping instance. The first two are just a byproduct of the normal mapping process. Therefore there is no additional computation needed to generate the data for a remote mapping instance. However all three mechanics use OpenVDB grids which are compressed and serialized as a bitstream before sending them out over the network.

Updates are grids generated during the raycasting process and are also used for the internal map update. Here all sensor data is previously accumulated into an update grid which specifies whether the occupancy probability in the map should be updated as occupied or free. This two step process helps to prevent discretization issues due to the larger resolution of the grid compared to the raw sensor data. It is also possible to accumulate multiple sensor sources and measurements into one update grid (see accumulation parameters). The update grid can then be applied to the local or remote instance to generate the same map on both sides.

Overwrite grids are similar to the update grid but only contain the voxel that have changed their occupancy state in the last iteration. Therefore they are more lightweight in terms of necessary bandwith. This can be useful in cases where the network is rather limited but has the downside, that the probabilistic framework is lost on the remote side.

In contrast to updates and overwrites, sections are not bound to the sensor data rate but are generated with a defined rate and create additional processing load. Here the binary occupancy state of a section around a specified frame is copied out of the map and transformed into an update grid that can be applied to a remote instance. Although this created additional overhead it is still highly useful in certain scenarios. In our case we had to handle transmission delay and package loss where both the update and overwrites would result in incomplete maps since grids were lost. Sections on the other hand contain the complete chunk of the map while still being small with respect to the bandwidth.

VDB Mapping is highly configurable using ROS parameters. Below is a complete list of all available parameters. Internally we store them in a yaml configuration file. An example can be found here.

Listed below are the general parameters to configure the basic behavior of vdb_mapping

VDB Mapping is able to integrate an arbitrary amount of different sensor sources (given they are available as PointCloud2 msg). To add a new sensor source the user has to add it to the list of sources in the basic parameters. Below are listed the parameters that can be used to configure the individual sensor source. These have to be put into the corresponding namespace in the yaml file. For an example see this configuration file.

VDB Mapping provides a remote operation mode. Here the map data is shared between multiple instances to generate the same or similar maps (depending on the chosen parameter settings). For this remote sources can be specified over the parameter server.

Below are listed the parameters that can be used to configure the individual remote source. These have to be put into the corresponding namespace in the yaml file. An example config file can be found here.

Thanks that you read until here and please let us know if you run into any issues or have suggestions for improvements. If you use our work in your publications please feel free to cite our manuscript:

  @inproceedings{besselmann2021vdb,
  title={VDB-Mapping: a high resolution and real-time capable 3D mapping framework for versatile mobile robots},
  author={Besselmann, M Grosse and Puck, Lennart and Steffen, Lea and Roennau, Arne and Dillmann, R{\"u}diger},
  booktitle={2021 IEEE 17th International Conference on Automation Science and Engineering (CASE)},
  pages={448--454},
  year={2021},
  organization={IEEE}
  doi={10.1109/CASE49439.2021.9551430}}
}

or for the remote mapping case:

@incollection{besselmann2022remote,
  title={Remote VDB-Mapping: A Level-Based Data Reduction Framework for Distributed Mapping},
  author={Besselmann, Marvin Grosse and R{\"o}nnau, Arne and Dillmann, R{\"u}diger},
  booktitle={Robotics in Natural Settings: CLAWAR 2022},
  pages={448--459},
  year={2022},
  publisher={Springer}
  doi={10.1007/978-3-031-15226-9_42}
}

The research leading to this package has received funding from the German Federal Ministry of Education and Research under grant agreement No. 13N14679:

ROBDEKON - Robotic systems for decontamination in hazardous environments
More information: robdekon.de