Visual SLAM Bachelor Thesis

Features

• Visual/Visual-Inertial SLAM implementation
• Can run with one or multiple¹ Stereo Cams
• GPU based
• NVIDIA Jetson & ZED 2i
• ROS2
• Created for ANT-Robotic Adir platform

Requirements

• NVIDIA GPU with CUDA
• ZED-SDK installed
• Camera with at least 30FPS

Installation

At the moment of the release the SLAM package does not support building the it from source for the ZED camera since it's currently not supported.

mkdir -p ~/ros2_ws/src/ # create your workspace if it does not exist
git clone  --recursive https://github.com/hermanndererdmann/ADIR_VSLAM.git
cd ..
sudo apt update
rosdep update
rosdep install --from-paths src --ignore-src -r -y # install dependencies
colcon build --symlink-install --cmake-args=-DCMAKE_BUILD_TYPE=Release --parallel-workers $(nproc) # build the workspace
source install/setup.bash

This guide can be help to get it working for the ISAAC ROS 3.0.1 Version.

Launch

Launch each in a seperate terminal with sources workspace.

ros2 launch zed_wrapper zed_camera.launch.py camera_model:='zed2i'

ros2 launch adir_amr_description display.launch.py

rviz2 -d src/nvidia_vslam/rviz/default.cfg.rviz

For VI-SLAM

ros2 launch nvidia_vslam visual-inertial.launch.py

or V-SLAM

ros2 launch nvidia_vslam visual.launch.py

Things to keep in mind

For visual-inertial SLAM the Allan variance of the IMU has to be calculated. This can be done with this repo (readme not up to date). It is recommended to record 15-24h of IMU topics to a ROSBAG. The IMU has to be perfectly still during that time.

These calculated values have to be put in the NVIDIA VSLAM launch files.

Links

Official NVIDIA cuVSLAM Documentation

Changelog

Version 1.0.0

Released 09/08/2024:

• Inital release (end of thesis)

Footnotes

1. Not implemented/tested for ZED yet. ↩