OpenSpiel: A Framework for Reinforcement Learning in Games

Abstract

OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games. OpenSpiel supports n-player (single- and multi- agent) zero-sum, cooperative and general-sum, one-shot and sequential, strictly turn-taking and simultaneous-move, perfect and imperfect information games, as well as traditional multiagent environments such as (partially- and fully- observable) grid worlds and social dilemmas. OpenSpiel also includes tools to analyze learning dynamics and other common evaluation metrics. This document serves both as an overview of the code base and an introduction to the terminology, core concepts, and algorithms across the fields of reinforcement learning, computational game theory, and search.

References in corpus (10)

• Recurrent Models of Visual Attention
• DeepStack: Expert-Level Artificial Intelligence in No-Limit Poker
• The Hanabi Challenge: A New Frontier for AI Research
• Improved Memory-Bounded Dynamic Programming for Decentralized POMDPs
• Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning
• Actor-Critic Policy Optimization in Partially Observable Multiagent Environments
• Modeling Others using Oneself in Multi-Agent Reinforcement Learning
• Open-ended Learning in Symmetric Zero-sum Games
• Re-evaluating Evaluation
• Fast deep reinforcement learning using online adjustments from the past

Cited by in corpus (31)

• Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model
• PettingZoo: Gym for Multi-Agent Reinforcement Learning
• RLCard: A Toolkit for Reinforcement Learning in Card Games
• MALib: A Parallel Framework for Population-based Multi-agent Reinforcement Learning
• ABIDES-Gym: Gym Environments for Multi-Agent Discrete Event Simulation and Application to Financial Markets
• Navigating the Landscape of Multiplayer Games
• Learning to Play No-Press Diplomacy with Best Response Policy Iteration
• From Poincaré Recurrence to Convergence in Imperfect Information Games: Finding Equilibrium via Regularization
• Neural Replicator Dynamics
• Real World Games Look Like Spinning Tops
• Efficient Deviation Types and Learning for Hindsight Rationality in Extensive-Form Games
• Scaling Scaling Laws with Board Games
• Model-free Neural Counterfactual Regret Minimization with Bootstrap Learning
• Physically Embedded Planning Problems: New Challenges for Reinforcement Learning
• ${\alpha }^{\alpha }$-Rank: Practically Scaling $\alpha$-Rank through Stochastic Optimisation
• XDO: A Double Oracle Algorithm for Extensive-Form Games
• Human-Agent Cooperation in Bridge Bidding
• Multiplayer Support for the Arcade Learning Environment
• Human-Robot Collaboration via Deep Reinforcement Learning of Real-World Interactions
• Multi-Agent Training beyond Zero-Sum with Correlated Equilibrium Meta-Solvers
• Swift for TensorFlow: A portable, flexible platform for deep learning
• Approximate exploitability: Learning a best response in large games
• Alternative Function Approximation Parameterizations for Solving Games: An Analysis of $f$-Regression Counterfactual Regret Minimization
• Mixture of Public and Private Distributions in Imperfect Information Games
• Evaluating Strategy Exploration in Empirical Game-Theoretic Analysis
• Improving Fictitious Play Reinforcement Learning with Expanding Models
• Empirical Game-Theoretic Analysis for Mean Field Games
• Neural Auto-Curricula
• Perfect Information Monte Carlo with Postponing Reasoning
• Reward-Free Attacks in Multi-Agent Reinforcement Learning
• Evolutionary Dynamics and $\Phi$-Regret Minimization in Games