Deep Reinforcement Learning for Mobile Robot Path Planning

This repository contains the semester project for ROB962.

TODO: Add more description about the repository and how to get started.

Installation

System requirements

• Webots R2020a or newer: https://cyberbotics.com
• Python 3.7
  • IMPORTANT! you must use the distribution from python.org if using macOS or Windows, only this distribution is supported in Webots R2020a, else the Webots controllers must be recompiled. This also makes it difficult to use virtual environments e.g. with Anaconda.
  • This may not be an issue in Ubuntu where the system-wide Python distribution can be used
• Stable Baselines - Deep Reinforcement Learning algorithms: https://stable-baselines.readthedocs.io/en/master/guide/install.html

How to install the gym environment

The repository contains two gym environments:

• gym-webots-RLenv (Not working atm)
• P9-RL-env (Use this one)

To install the environment, go to the P9-RL-env folder and run the pip install command:

$ cd P9/P9-RL-env/

$ pip install -e .

The environment was created using the following guide: (https://medium.com/@apoddar573/making-your-own-custom-environment-in-gym-c3b65ff8cdaa)

How to run the Webots example

Assuming you already have WeBots set up with an existing project, go to the project library, and

• copy the master folder to the controller folder
• copy the contents of the world folder into your project's world folder (P9_v0.0.1.wbt is used currently)

That should be all. You should be able to open the world from WeBots.

How to modify the example

If you want to modify something (e.g. reward function, action space, etc) you need to modify the gym environment, which is at P9/P9-RL-env/P9_RL_env_v01/envs/P9RLEnv.py

The master controller uses and manages this environment, therefore the robot needs to be set as having the master controller.

Alternatively, you can set the Webots robot controller to extern and run master.py from e.g. vscode.

Authors

• Mark Adamik
• Asger Printz Madsen