ORB_SLAM3_ROS2

ORB_SLAM3 with ROS2 support

Environment

OS: Ubuntu 22.04 LTS ROS Distro: Humble

Dependencies

Pangolin (Optional)

Pangolin for visualization and user interface. Tested with Pangolin 0.9.1.

sudo apt install ros-humble-pangolin

Foxglove (Optional)

Foxglove for visualization.

sudo apt install ros-humble-foxglove-bridge

OpenCV

OpenCV for manipulating images and features. Tested with OpenCV 4.5.x.

sudo apt install libopencv-dev

Eigen3

Eigen3 for linear algebra computation like matrices, vectors, numerical solvers, and related algorithms. Tested with Eigen3 3.4.0.

sudo apt install libeigen3-dev

Sophus

Sophus for Lie groups computation. Tested with Sophus 1.22.x.

sudo apt install ros-humble-sophus

How to Use

We use binary vocabulary file for faster loading and lower file size. The converted vocabulary file is placed in ORB_SLAM3/vocabulary/ORBvoc.bin. We also provide a python script for convertion in asset/python/voc2bin.py, in case you want to use your own vocabulary.

Compile

Simply run ./build.sh if your're using the same ROS distro with me. If not, modify the second line source /opt/ros/humble/setup.bash to suit your ROS environment.

Visualization

Visualization is enabled by default. If you're not using this function pass -DENABLE_PREVIEW=OFF when compiling.

./build.sh -DENABLE_PREVIEW=OFF

Run

Running ORB_SLAM3 with Realsense D435i module for example.

Currently only support RGB-D mode.

source /opt/ros/humble/setup.bash
source install/setup.bash
ros2 launch orb_slam3 realsense_435i.launch.py

Foxglove Visualization

Pass parameter foxglove_bridge=True in your launch command. Follow the Foxglove official documentation.

To-do List

• Support pointcloud map publish
• Make visualization compile optional
• Support Foxglove for visualization
• Implement RGB-D with inertial mode
• Make vocabulary file binary