SelComm_Webots

This repository contains the codes for our paper, which has been accepted by RA-L and IROS 2021.

This is the implementation based on Webots. For Stage_ROS version, please refer to another repository.

Requirement

• Python 3.6

• ROS Melodic

• mpi4py

• Webots 2020b

• PyTorch 1.4

Setup

Set up your Pycharm according to https://cyberbotics.com/doc/guide/using-your-ide?tab-language=python#pycharm.

How to train

Run the following command::

webots worlds/6robots_360lidar_pedestrian.wbt
cd ./controllers/pioneer_controller/
python run.py

Video Demo

The global view

The local view

References

The authors thank Liu for the open sourced code.

@misc{Tianyu2018,
    author = {Tianyu Liu},
    title = {Robot Collision Avoidance via Deep Reinforcement Learning},
    year = {2018},
    publisher = {GitHub},
    journal = {GitHub repository},
    howpublished = {\url{https://github.com/Acmece/rl-collision-avoidance.git}},
    commit = {7bc682403cb9a327377481be1f110debc16babbd}
}