This repository provides the environments for the paper "Emergent Complexity via Multi-agent Competition," enabling researchers and practitioners to explore emergent behaviors in competitive multi-agent systems. The code is offered as-is, with no further updates planned.

How It Works

The project implements competitive multi-agent environments based on the OpenAI Gym framework. It leverages MuJoCo for physics simulation, allowing for complex agent interactions and emergent strategies within these simulated worlds.

Quick Start & Requirements

• Install dependencies: pip install -r requirements.txt
• Requires Python 3.6, OpenAI Gym 0.9.1 with MuJoCo 1.31 support (mujoco-py 0.5.7), Tensorflow 1.1.0, and Numpy 1.12.1.
• Install the package: cd gym-compete && pip install -e .
• Demo all environments: bash demo_tasks.sh all
• Demo a single environment: bash demo_tasks.sh <task>

Highlighted Details

• Provides environments for tasks like run-to-goal-humans, run-to-goal-ants, you-shall-not-pass, sumo-ants, sumo-humans, and kick-and-defend.
• Includes an agent-zoo folder with pre-trained agent policies.

Maintenance & Community

The project is archived and no updates are expected.

Licensing & Compatibility

The repository does not explicitly state a license.

Limitations & Caveats

The code is provided as-is and is archived, indicating no ongoing development or support. The dependencies are for older versions of Python, Gym, MuJoCo, and TensorFlow, which may pose compatibility challenges with modern systems.