LVGS-SLAM: LiDAR-Visual-Supervised Gaussian Splatting SLAM with Dense Depth Rendering for Unstructured Environments

Installation

1. Installation of localization module

Prerequisites:

• Ubuntu 20.04 (ROS Noetic)
• ROS Core System
• ceres-solver

1. Install system & ROS dependencies:

   sudo apt-get update && sudo apt-get install -y git libpng++-dev 
   sudo apt-get install -y python3-catkin-tools ros-noetic-opencv-apps
   

2. Create a catkin workspace :

   mkdir -p ./catkin_ws/src
   cd ./catkin_ws
   catkin init
   cd ..
   mv ./SLAM/src/* ./catkin_ws/src
   

3. Install dependencies:

   cd ./catkin_ws/src/limo
   bash install_repos.sh
   

4. Build the workspace:

   cd ../
   catkin_make
   

2. Installation of mapping and refinement module

1. Create a conda environment:

   conda create -n LVGS python=3.10
   conda activate LVGS
   

2. Install CUDA 11.8:

   pip install torch==2.0.1+cu118 torchvision==0.15.2+cu118 --extra-index-url https://download.pytorch.org/whl/cu118
   conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit
   pip install dill --upgrade
   pip install --upgrade pip "setuptools<70.0"
   pip install ninja git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch
   

3. Install dependencies:

   
   pip install waymo-open-dataset-tf-2-11-0==1.6.1
   
   cd neurad-studio
   #The `-e` flag installs the package in editable mode.
   pip install -e .
   pip install submodules/gsplat
   cd src
   pip install -e .
   

---

Run

This project uses the BotanicGarden Dataset.

1. Download the Rosbag: Obtain the dataset from its official source and place it in the directory specified in botanic.launch.

2. Run SLAM

   roslaunch test_ape test.launch
   

3. Run only the localization module.

   roslaunch demo_keyframe_bundle_adjustment_meta botanic.launch
   

4. Run only the mapping and refinement module.

   python nerfstudio/scripts/train.py splatad-wild --vis viewer+tensorboard botanic-data
   

Acknowledgements

Our work is built upon the following projects:

• nerfstudio
• neurad-studio
• limo
• Difix3D
• img2img-turbo
• GS-LIVM

We thank the authors and contributors of these repositories for making their work publicly available.