Reinforcement Learning: Jelly Drift

AI Playing Games - Racing Time

Driving a car within a video game is a relatively easy and relaxing process for a human. Here we try and train an AI to race a car around a race track as fast as possible!

This project uses the environment created here!

Project Structure

To install and use environment.yml do the following:

1. Install Anaconda or Miniconda
2. Create a new environemtn from the file conda env create -f environment.yml
3. When packages need to be added or modified update the environment.yml file and run conda env update -f environment.yml --prune
4. Once satisfied activate the new environment through conda activate jd_rl and (optionally) deactivate/close afterwards with conda deactivate

For an update the above isn't necessary - simply run conda update --all for packages and conda update -n base -c defaults conda for conda itself!

Config

This project has been set up using Hydra. To start a run please create a run config file under config/runs/ and specify it as a command line argument - python src/main.py +runs=test_run. For reproducibility avoid modifying the default configuration and instead override it within the run config file.