Learning End-to-End Decentralized Formation Using Graph Neural

Networks

Core GNN Library used from: https://github.com/proroklab/gnn_pathplanning

Code for paper "Learning End-to-End Decentralized Formation Using Graph Neural Networks"

Installation

System requirement

Ubuntu 18.04LTS
Python 3.6

Install coppeliaSim

https://coppeliarobotics.com/downloads

Install Opencv

sudo apt install python3-opencv

Install Dependencies:

pip3 install -r requirements.txt

---

Testing

Start coppeliasim

./"path to coppeliasim"/coppeliaSim.sh

Always make sure the number of robots in the simulator equals to the number of robot in Demo.py

Run triangulation formation with default configuration

Open scene_5.ttt(in scene folder) in coppeliaSim

python Test.py

---

Run circle formation with default configuration

Open scene_circle.ttt(in scene folder) in coppeliaSim

python Test_cirlce.py

---

Run lin formation with default configuration

Open scene_line.ttt(in scene folder) in coppeliaSim

python Test_line.py

---

Training

Training with default configuration

Open scene_5.ttt(in scene folder) in coppeliaSim

Open scene_5.ttt(in scene folder) in coppeliaSim

python Train.py