Supply Chain Bot Tournament implemented as an OpenAI Gym environment.
The core part of the game environment's implementation was copied from this repository.
-
Fork this repo
-
Create a new conda environment to keep things clean
conda create python=3.6 --name supply-chain-env
source activate supply-chain-env
- Clone the forked repository
git clone https://github.com/<your-user>/pydataglobal-bot-game-2020.git
- Point to root repository and install the package
cd supply-chain-env
pip install -e .
- Launch the
bot.py
file to run the game with randomly behaving agents
python bot.py --no_submit
If everything was setup properly, you'll see a bunch of messaged printed to the standard output showing agents' actions and their outcomes.
In order to implement your solution, open the bot.py
file and provide your implementation
for all given agents. You don't need to change anything except agents' classes. Also, please
note that a valid agent's strategy cannot peek into the game's environment state or communicate
internal state to other agents! (See the docstring at the very beginning of the file for
additional details).
To test your implementation, run the bot.py
script again. The better your solution works, the
smaller should be the total cost reported at the end of the game.
In order to participate in the tournament, you should follow these steps:
- Once you're happy with the developed strategy, post your changes to a new branch in this repository
to trigger the evaluation and open a PR to merge to
master
- GitHub Actions will take care of evaluating your implementation and post your results to the leaderboard
- Update your branch as often as you like, but be aware that the most recent results will be updated to the leaderboard, irrespectively of the result
Good luck and have fun!