This repository contains the code for CTopPRM: Clustering Topological PRM for Planning Multiple Distinct Paths in 3D Environments presented in our paper Novosad, Penicka, Vonasek CTopPRM'23.

If you use this code in an academic context, please cite the following publication:

M. Novosad, R. Penicka and V. Vonasek, "CTopPRM: Clustering Topological PRM for Planning Multiple Distinct Paths in 3D Environments," in IEEE Robotics and Automation Letters, vol. 8, no. 11, pp. 7336-7343, Nov. 2023, doi: 10.1109/LRA.2023.3315539. (PDF)

@ARTICLE{novosad2023ctopprm,
  author={Novosad, Matej and Penicka, Robert and Vonasek, Vojtech},
  journal={IEEE Robotics and Automation Letters}, 
  title={CTopPRM: Clustering Topological PRM for Planning Multiple Distinct Paths in 3D Environments}, 
  year={2023},
  volume={8},
  number={11},
  pages={7336-7343},
  doi={10.1109/LRA.2023.3315539}
}

GPL-3.0 License. Copyright (C) 2022 M. Novosad, R. Pěnička, V. Vonásek (Faculty of Electrical Engineering, Czech Technical University in Prague).

This is a research code, expect that it changes often and any fitness for a particular purpose is disclaimed.

The code has been developed and tested on Ubuntu 22.04.

Clone the repository and update the submodules

git clone [email protected]:ctu-mrs/CTopPRM.git

cd CTopPRM/

git submodule update --init --recursive

gitman update

You might have to install gitman first pip install gitman

Install the following dependencies

sudo apt-get install build-essential cmake pkg-config ccache zlib1g-dev libomp-dev libyaml-cpp-dev libhdf5-dev libgtest-dev liblz4-dev liblog4cxx-dev libeigen3-dev python3 python3-venv python3-dev python3-wheel python3-opengl

Compile the dependencies present in submodules

make dependencies

Finally, compile the code

make

To create the ESDF maps from the mesh .obj fles use the map.py script in the python folder. To have all the dependencies we suggest using the python environment. Start the environment and activate it using:

python3 -m venv env source env/bin/activate

Afterwards install the python dependencies using pip (or pip3 if python3 is not default):

pip install setuptools~=57.5.0

pip install scikit-learn

pip install wheel

pip install pyopengl==3.1.0

pip install numpy trimesh matplotlib mesh_to_sdf python-csv

When the dependencies are installed run the following command in the blender folder to create the ESDF maps (replace MESH_NAME with the name og the mesh file you wish to use):

./map.py MESH_NAME.obj

After compilation you should see the main binary. The prepared cofiguration files for exisitng maps are stored in config_files/ where the desired parameters and map can be set. Desired cofiguration file has to be defined in main.cpp. Before the code is run for the first time a directory for storing results has to be created:

mkdir prints/

Finally the code can simply be run using the command:

./main

To visualize results of the last run in blender follow the instructions:

• import .obj file of the mesh of the evironments used to blender
• open columns_2.py script
• select folder in Line 147 to be 'prm', 'clusters', 'connections' or 'shortened' - depending on which step of the algorithm you want visualized
• is_cl = true in line 142 for 'prm', 'clusters' and 'connections' AND main has to be run with PRINT (true) macro in topological_prm_clustering.hpp
• run script

Some of the prepared .obj files are missing a cluster object required to visualize 'prm', 'clusters', 'connections' - to fix it just import prepared cluster.obj object into blender as well