OpenSpiel: A Framework for Reinforcement Learning in Games

• DOI: 10.48550/arxiv.1908.09453
• Publication Date: 2019-01-01
• Authors: Lanctot, Marc; Lockhart, Edward; Lespiau, Jean-Bap...; Zambaldi, Vinicius; Upadhyay, Satyaki; Pérolat, Julien; Srinivasan, Sriram; Timbers, Finbarr; Tuyls, Karl; Omidshafiei, Shay...; Hennes, Daniel; Morrill, Dustin; Muller, Paul; Ewalds, Timo; Faulkner, Ryan; Kramár, János; De Vylder, Bart; Saeta, Brennan; Bradbury, James; Ding, David; Borgeaud, Sebastian; Lai, Matthew; Schrittwieser, Ju...; Anthony, Thomas; Hughes, Edward; Danihelka, Ivo; Ryan-Davis, Jonah
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; Computer Science - Computer Science and Game Theory; 0202 electrical engineering, electronic engineering, information engineering; Computer Science - Multiagent Systems; 02 engineering and technology; Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)

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.