DB-TSDF: Directional Bitmask-based Truncated Signed Distance Fields for Efficient Volumetric Mapping

Accepted at the IEEE International Conference on Robotics and Automation (ICRA) 2026

DB-TSDF presents a high-efficiency, CPU-only framework for volumetric mapping. It utilizes a novel directional bitmask-based integration scheme to incrementally fuse LiDAR data into a dense voxel grid.

Key features include:

• Directional Kernels: Efficiently model beam geometry and occlusion in 3D.
• Bitmask Encoding: Ensures constant-time updates per scan, independent of grid resolution.
• High Performance: Multi-threaded C++ implementation fully integrated with ROS 2.

The design prioritizes predictable runtime and high-resolution reconstruction, making it an ideal solution for robotic platforms with limited GPU resources.

1. Prerequisites

Before you begin, make sure you have ROS 2 Humble and Ubuntu 22.04 (or higher) installed on your system. These are the core requirements for the project to run smoothly. If you haven't installed ROS 2 Humble yet, follow the official installation guide for your platform. This guide will walk you through all the necessary steps to set up the core ROS 2 environment on your system.

2. Installation

Option A — Docker (recommended)

The Dockerfile is self-contained: it installs ROS 2 Humble and every dependency, clones the repo and builds it with colcon. Fastest way to a clean, reproducible setup.

git clone https://github.com/robotics-upo/DB-TSDF.git
cd DB-TSDF
docker build -t db_tsdf_ros2:humble \
  --build-arg USER_UID=$(id -u) \
  --build-arg USER_GID=$(id -g) .
xhost +local:docker   # allow GUI apps like RViz
docker run -it \
  --env="DISPLAY" \
  --env="QT_X11_NO_MITSHM=1" \
  --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" \
  --name db_tsdf_container \
  db_tsdf_ros2:humble

To resume the same container later: docker start -ai db_tsdf_container.

Option B — Local ROS 2 workspace

Clone into the src folder of a ROS 2 workspace and build with colcon:

mkdir -p ~/ros2_ws/src
cd ~/ros2_ws/src
git clone https://github.com/robotics-upo/DB-TSDF.git
cd ~/ros2_ws
rosdep install --from-paths src --ignore-src -r -y
colcon build
source install/setup.bash

3. Running the Code

The launch system uses a single main launch file (mapper_launch.py) and a config argument that selects which dataset configuration to load (e.g., mai loads mai.yaml and mai.rviz).

Quick start with a test dataset

1. Download a ready-to-play MaiCity sequence. The script fetches the official archive, converts it from ROS 1 to rosbag2 and removes every intermediate file, leaving the dataset at datasets/mai_city/<sequence>/ (path printed on completion):

   ./src/db_tsdf/scripts/download_test.sh        # sequence 01 only — quick test (~215 MB)
   ./src/db_tsdf/scripts/download_mai_city.sh    # all sequences (~3.4 GB)

2. Launch DB-TSDF with the matching config. RViz will open and the node will wait for data:

   ros2 launch db_tsdf mapper_launch.py config:=mai

3. In a second terminal, play the dataset back:

   For the Docker container, open one with:

   docker exec -it db_tsdf_container bash

   and run it with:

   ros2 bag play ~/ros2_ws/datasets/mai_city/01

4. Configuration

The system is highly configurable via YAML parameters. (e.g., config/college.yaml).

Core Parameters

Parameter	Type	Description	Default
in_cloud	string	Input PointCloud2 topic	/os_cloud_node/points
odom_frame_id	string	Fixed frame for TF lookup	odom
use_tf	bool	Enable/Disable TF transformations	True
verbose_init	bool	Full parameter dump + kernel preview at startup	False

Grid Definition

Parameter	Type	Description	Default
tdf_grid_res	float	Voxel side length in meters	0.05
tdf_max_cells	int	Max active cells in hash table	75000
tdfGridSizeX/Y*_low/high	float	Horizontal volume boundaries	+/-100
tdfGridSizeZ_low/high	float	Vertical volume boundaries	-10 / +50

Integration Kernel

Parameter	Type	Description	Default
kernel_size	int	Kernel size (odd number)	7
bins_az / bins_el	int	Angular discretization	60
occ_min_hits	int	Min measurements to mark occupied	50

5. Output Data and Services

The node provides four std_srvs/srv/Trigger services to export the reconstructed map. Each one runs in the background and writes its output relative to the directory the node was launched from:

Service	Output	Description
/save_grid_pcd	grid_data.pcd	Occupied-voxel point cloud (PCD)
/save_grid_ply	grid_data.ply	Occupied-voxel point cloud (PLY)
/save_grid_csv	grid_data_csv/	Per-cell voxel data (CSV + PLY), one file pair per allocated subgrid cell
/save_grid_mesh	mesh.stl	Surface mesh extracted with Marching Cubes

ros2 service call /save_grid_pcd  std_srvs/srv/Trigger "{}"
ros2 service call /save_grid_ply  std_srvs/srv/Trigger "{}"
ros2 service call /save_grid_csv  std_srvs/srv/Trigger "{}"
ros2 service call /save_grid_mesh std_srvs/srv/Trigger "{}"

(back to top)

## Citation If you use DB-TSDF in your research, please cite our paper:

@inproceedings{maese2026dbtsdf,
  title={DB-TSDF: Directional Bitmask-based Truncated Signed Distance Fields for Efficient Volumetric Mapping},
  author={Maese, Jose E. and Caballero, Fernando and Merino, Luis},
  booktitle={2026 IEEE International Conference on Robotics and Automation (ICRA)},
  year={2026}
}

Acknowledgements

This work was supported by the grants PICRA 4.0 (PLEC2023-010353), funded by the Spanish Ministry of Science and Innovation and the Spanish Research Agency (MCIN/AEI/10.13039/501100011033); and INSERTION (PID2021-127648OB-C31), funded by the "Agencia Estatal de Investigación - Ministerio de Ciencia, Innovación y Universidades" and the "European Union NextGenerationEU/PRTR".