ADBSCAN is an advanced unsupervised clustering algorithm that groups high-dimensional points based on their distribution density. It is an improvement over the classic DBSCAN algorithm where the clustering parameters are adaptive based on the range, making it especially suitable for processing LiDAR data. ADBSCAN is an Intel-patented algorithm which improves the object detection range by 20-30% on average.
This repository contains several AMR (Autonomous Mobile Robot) algorithm implementations based on ADBSCAN, including:
- ROS2_node - ROS2 package with a node that subscribes to pointcloud sensors (LIDAR/RealSense) and publishes a list of objects in ObstacleArray message format
- Standalone - Standalone C++ source code for the ADBSCAN algorithm with sample input files
- Visualization - Python scripts to visualize the bounding boxes of detected object clusters in pointcloud data
- Follow_me_RS_2D - ROS2 package implementing a person-following algorithm with gesture and voice audio control
- package/tutorial_follow_me - ROS2 tutorial for running the follow-me application on a custom AAEON robot
- package/tutorial_follow_me_w_gesture - ROS2 tutorial for running the gesture-based follow-me application on a custom AAEON robot
- package/tutorial_aaeon_adbscan - ROS2 tutorial for running the ADBSCAN algorithm on a custom AAEON robot
All ROS2 packages support the following platforms:
- OS: Ubuntu 22.04 (ROS2 Humble) and Ubuntu 24.04 (ROS2 Jazzy)
Prepare the target system following the official documentation.
To build Debian packages, export the ROS_DISTRO environment variable to the desired platform and run the make package command. After the build process successfully finishes, built packages will be available in the root directory. The following command is an example for the Jazzy distribution:
ROS_DISTRO=jazzy make packageYou can list all built packages:
$ ls | grep -i .deb
ros-jazzy-adbscan-ros2_2.3-1_amd64.deb
ros-jazzy-follow-me-interfaces_2.3-1_amd64.deb
ros-jazzy-adbscan-follow-me-rs2d_2.3-1_amd64.deb
...The *build-deps*.deb packages are generated during the build process and installation of such packages can be skipped on the target platform.
To clean all build artifacts:
make cleanIf Ubuntu 22.04 with Humble is used, then run:
source /opt/ros/humble/setup.bashIf Ubuntu 24.04 with Jazzy is used, then run:
source /opt/ros/jazzy/setup.bashFinally, install the Debian package that was built via make package:
sudo apt update
find . -type f -name '*.deb' -not -name '*-build-deps_*' -not -name '*-dbgsym_*' -not -name '*.changes' -exec sudo apt-get install -f -y {} \;To run unit tests (implemented with colcon) execute the below command with target ROS_DISTRO (example for Jazzy):
ROS_DISTRO=jazzy make testThere is a set of prepared Makefile targets to speed up the development.
In particular, use the following Makefile target to run code linters:
make lintAlternatively, you can run linters individually:
make lint-bash
make lint-clang
make lint-githubactions
make lint-json
make lint-markdown
make lint-python
make lint-yamlTo run license compliance validation:
make license-checkTo see a full list of available Makefile targets:
$ make help
Target Description
------ -----------
clean Clean up all build artifacts
clean-colcon Clean up Colcon build and test artifacts
clean-debian Clean up Debian packaging artifacts
help Display this help message
license-check Perform a REUSE license check using docker container
lint Run all sub-linters using super-linter
lint-all Run super-linter over entire repository
lint-bash Run Bash linter using super-linter
lint-clang Run clang linter using super-linter
lint-githubactions Run Github Actions linter using super-linter
lint-json Run JSON linter using super-linter
lint-markdown Run Markdown linter using super-linter
lint-python Run Python linter using super-linter
lint-yaml Run YAML linter using super-linter
package Build Debian packages
test Test code using colconThis ROS2 package for the ADBSCAN algorithm contains a ROS2 node which subscribes to pointcloud sensors (LIDAR/RealSense camera) and publishes a list of objects in ObstacleArray message format.
Find instructions for this package: package instructions
This directory contains standalone C++ source code for the ADBSCAN algorithm. Sample pointcloud input files to run the executable are available in the input files directory.
This directory contains necessary Python scripts to visualize the bounding boxes of the object clusters in the pointcloud data.
This is a ROS2 package for an AMR algorithm where a robot follows a target person. It contains a ROS2 node which subscribes to pointcloud sensors (LIDAR/RealSense camera), uses the ADBSCAN algorithm to cluster the data and detect the location of the target person, and subsequently publishes the velocity commands for a differential drive robot.
This package supports four demo modes combining two sensor types (2D LiDAR or Intel RealSense depth camera) with optional hand-gesture control and voice audio control via OpenVINO speech recognition. Find instructions for this package: Follow_me_RS_2D package instructions
This package provides a ROS2 tutorial for the basic follow-me application on a custom AAEON robot. It demonstrates how to deploy the person-following pipeline based on ADBSCAN clustering, without gesture control.
This package provides a ROS2 tutorial for the gesture-based follow-me application on a custom AAEON robot. It extends the basic follow-me pipeline with a gesture recognition model so that the robot motion can be controlled using hand gestures of the target person.
This package provides a ROS2 tutorial focused on running the ADBSCAN algorithm itself on a custom AAEON robot. It shows how to deploy and configure the ADBSCAN ROS2 node for obstacle clustering and object detection, without the full follow-me behavior.
Comprehensive documentation on this component is available here: dev guide
ADBSCAN is licensed under Apache 2.0 License.