Skip to content

Introduction

Introduction to Coopetition Gym

Section titled “Introduction to Coopetition Gym”

Coopetition Gym is a research-grade Python library that provides OpenAI Gym-compatible environments for studying coopetitive dynamics in multi-agent systems.

What is Coopetition?

Section titled “What is Coopetition?”

Coopetition is the phenomenon where agents simultaneously cooperate and compete. This captures real-world scenarios better than pure cooperation or pure competition models:

  • Businesses compete for market share while cooperating on industry standards
  • Nations compete economically while cooperating on global challenges
  • Researchers compete for funding while collaborating on publications

Core Features

Section titled “Core Features”

Gym-Compatible Interface

Section titled “Gym-Compatible Interface”

All environments follow the OpenAI Gym interface, making them compatible with popular RL frameworks:

env = cg.make('PrisonersDilemma-v0')
obs = env.reset()
obs, rewards, done, info = env.step(actions)

Multi-Agent Support

Section titled “Multi-Agent Support”

Environments support configurable numbers of agents:

env = cg.make('PublicGoods-v0', n_agents=5)

Game-Theoretic Foundations

Section titled “Game-Theoretic Foundations”

Each environment is based on established game theory, with known theoretical equilibria for validation.

Available Environments

Section titled “Available Environments”

Coopetition Gym includes 10 environments across different categories:

CategoryEnvironments
Classic GamesPrisoner’s Dilemma, Chicken, Stag Hunt
EconomicPublic Goods, Oligopoly, Bargaining
NetworkNetwork Formation, Coalition
CustomTragedy of Commons, Resource Allocation

Research Validation

Section titled “Research Validation”

The library has been validated against theoretical predictions, achieving a 58/60 validation score across all environments. This ensures that agent behavior aligns with game-theoretic expectations.

Next Steps

Section titled “Next Steps”