Codebase
The book comes with a codebase written in Python, containing implementations of several MARL algorithms presented in the book. Its primary purpose is to provide algorithm code that is self-contained and easy to read.
MIT Press 2024
Multi-Agent
Reinforcement Learning
Foundations and Modern Approaches
The first comprehensive introduction to multi-agent reinforcement learning.
Errata for the second printing · report issues to issues~at~marl-book~dot~com
About the book
Multi-agent reinforcement learning (MARL) is an area of machine learning in which multiple decision-making agents learn to optimally interact in a shared environment. The book gives an introduction to the models, solution concepts, algorithmic ideas, and technical challenges in MARL, and describes modern approaches that integrate deep learning to produce scalable learning algorithms. It comes with a codebase written in Python that implements several of the algorithms discussed, along with a chapter describing how they are implemented. Readers are assumed to have an undergraduate background in basic mathematics and a basic familiarity with programming to use the codebase.
- PublisherMIT Press, 2024
- Chapters11 chapters, in two parts
- PDFFree to download
- ExtrasCode & lecture slides
Table of contents
Summary of Notation · List of Figures · Preface
- 1 Introduction
Part 1 Foundations of Multi-Agent Reinforcement Learning
- 2 Reinforcement Learning
- 3 Games: Models of Multi-Agent Interaction
- 4 Solution Concepts for Games
- 5 Multi-Agent Reinforcement Learning in Games: First Steps and Challenges
- 6 Multi-Agent Reinforcement Learning: Foundational Algorithms
Part 2 Multi-Agent Deep Reinforcement Learning: Algorithms and Practice
- 7 Deep Learning
- 8 Deep Reinforcement Learning
- 9 Multi-Agent Deep Reinforcement Learning
- 10 Multi-Agent Deep Reinforcement Learning in Practice
- 11 Multi-Agent Environments
Appendix A: Surveys on Multi-Agent Reinforcement Learning
News
Follow the authors for updates: Stefano, Filippos, Lukas
-
The Chinese translated edition of the book, published by China Machine Press, is now available (publisher book page, JD online shop).
-
The first printing of the book (released in December 2024) has sold out. A second printing is now in production, and the updated book PDF is available on this page.
-
The Artificial Intelligence Research Institute in Barcelona (IIIA-CSIC) hosted a summer course based on the book, given by Stefano V. Albrecht. The lecture recordings and coding exercises are now available.
Code and slides
Lecture slides
Lecture slides for chapters 1 to 9 are available as PDF files and LaTeX sources, so that instructors can make changes as required.
If you use the book or slides in your class, please let us know and we will add your class and institution to a class list.
Endorsements
Solicited by MIT Press
The authors meticulously bring reinforcement learning together with game theory to provide a foundation for research and application of multi-agent reinforcement learning. This book is the perfect starting point for a grounding in the field.
Multi-agent reinforcement learning is well positioned to be the next white-hot area of artificial intelligence and this book provides the essential background, concepts, and insights for understanding this exciting and important area of research.
This book is the first complete reference for the growing area of multi-agent reinforcement learning. It provides both an essential resource for newcomers to the field and a valuable perspective for established researchers.
A landmark textbook to multiagent reinforcement learning, combining game-theoretic foundations with state-of-the-art deep learning. This essential textbook delivers fundamental insights for newcomers, experts and practitioners, featuring real-world applications and advanced algorithms.
This will become the standard text of the emerging field of multiagent reinforcement learning. It builds from foundational ideas, incorporating recent breakthroughs in deep learning. This book will help accelerate theoretical and practical progress.
Citation
Stefano V. Albrecht, Filippos Christianos, and Lukas Schäfer. Multi-Agent Reinforcement Learning: Foundations and Modern Approaches. MIT Press, 2024.
@book{ marl-book,
author = {Stefano V. Albrecht and Filippos Christianos and Lukas Sch\"afer},
title = {Multi-Agent Reinforcement Learning: Foundations and Modern Approaches},
publisher = {MIT Press},
year = {2024},
url = {https://www.marl-book.com}
}Start reading today
The complete book is available as a free PDF.