Agar.io

Agar is a popular multi-player online game. Players control one or more cells in a Petri dish. The goal is to gain as much mass as possible by eating cells smaller than the player's cell while avoiding being eaten by larger ones. Larger cells move slower. Each player starts with one cell but can split a sufficiently large cell into two, allowing them to control multiple cells. The control is performed by mouse motion: all the cells of a player move towards the mouse position. We transform the Free-For-All (FFA) mode of Agar (https://agar.io/) into an Reinforcement Learning (RL) environment and we believe it can be utilized as a new Multi-agent RL testbed for a wide range of problems, such as cooperation, team formation, intention modeling, etc.

This is the source code of Agar.io, which is used in the paper "Discovering Diverse Multi-agent Strategic Behavior via Reward Randomization" https://arxiv.org/abs/2103.04564. This repository is developed from the original environment code (https://github.com/buoyancy99/PyAgar), but has been slightly modified to fit the algorithms used in RPG (Reward-Randomized Policy Gradient).

1. Installation

Agar.io is easy to install, see the following command. If you encounter visualization problems with pyglet, we would suggest you to seek help from official pyglet URL.

pip install pillow gym pyglet

We also use baselines from openai, but we changed it a little, so you should install it by running

pip baselines/setup.py install

2. User Guide

1. The Reinforcement Learning (RL) environment is defined in Env.py. This implementation supports multi-agent tasks, so you can specify any number of players (i.e. hyperparameter num_controlled_agent), which means you are allowed to control multiple agents by passing a [N, 3] 2D array, where N is numbers of controlled agents, 3 is the dimension of action space. The action space for each agent can be described as [mouse.x, mouse.y, Split/Feed/Non-operation], where mouse.x, mouse.y are continuous, while Split/Feed/Non-operation is discrete. There are two options for reward setting: std and agg, which correspond to the standard setting and aggressive setting of the paper. Note that this implementation only supports FFA mode (i.e. hyperparametergamemode=0)

2. The game configuration is defined in Config.py, we would suggest you to keep the default parameters.

3. Enjoy

We provide HumanControl.py for quick check and sample use. You can use the following command to play with mouse and keyboard. There are 2 agents by default, and you can change num_agents to any number you want. Hope you enjoy this game:)

python HumanControl.py

4. Citation

If you use Agar.io in your research, you can cite us with:

@misc{tang2021discovering,
      title={Discovering Diverse Multi-Agent Strategic Behavior via Reward Randomization}, 
      author={Zhenggang Tang and Chao Yu and Boyuan Chen and Huazhe Xu and Xiaolong Wang and Fei Fang and Simon Du and Yu Wang and Yi Wu},
      year={2021},
      eprint={2103.04564},
      archivePrefix={arXiv},
      primaryClass={cs.AI}
}

5. License

Agar.io is released under the MIT license.