Acting, Planning, and Learning

by Malik Ghallab, Dana Nau, Paolo Traverso, Michela Milano (Foreword by)

Overview

AI's next big challenge is to master the cognitive abilities needed by intelligent agents that perform actions. Such agents may be physical devices such as robots, or they may act in simulated or virtual environments through graphic animation or electronic web transactions. This book is about integrating and automating these essential cognitive abilities: planning what actions to undertake and under what conditions, acting (choosing what steps to execute, deciding how and when to execute them, monitoring their execution, and reacting to events), and learning about ways to act and plan. This comprehensive, coherent synthesis covers a range of state-of-the-art approaches and models –deterministic, probabilistic (including MDP and reinforcement learning), hierarchical, nondeterministic, temporal, spatial, and LLMs –and applications in robotics. The insights it provides into important techniques and research challenges will make it invaluable to researchers and practitioners in AI, robotics, cognitive science, and autonomous and interactive systems.

Product Details

ISBN-13:	9781009579384
Publisher:	Cambridge University Press
Publication date:	06/05/2025
Pages:	632
Product dimensions:	7.01(w) x 10.00(h) x 1.38(d)

About the Author

Malik Ghallab is Directeur de Recherche Emeritus at CNRS and the University of Toulouse. He has (co-)authored more than 200 scientific publications and books on AI and robotics, especially on acting, planning, and learning. He is a EurAI Fellow, and Docteur Honoris Causa of Linköping University, Sweden.

Dana Nau is Professor Emeritus at the University of Maryland in the Computer Science Department and the Institute for Systems Research. He has more than 400 refereed scientific publications, primarily on AI, game theory, and several interdisciplinary topics. He is an AAAI Fellow, ACM Fellow, and AAAS Fellow.

Paolo Traverso is Director of Strategic Planning at Fondazione Bruno Kessler (FBK), Trento, Italy. His main research interests are in automated planning and learning under uncertainty. He is the author and co-author of more than 100 scientific articles. He is a EurAI Fellow, AAIA Fellow, and AIIA Fellow.

Table of Contents

About the authors; Foreword; Preface; Acknowledgements; 1. Introduction; Part I. Deterministic State-Transition Systems: 2. Deterministic representation and acting; 3. Planning with deterministic models; 4. Learning deterministic models; Part II. Hierarchical Task Networks: 5. HTN representation and planning; 6. Acting with HTNs; 7. Learning HTN methods; Part III. Probabilistic Models: 8. Probabilistic representation and acting; 9. Planning with probabilistic models; 10. Reinforcement learning; Part IV. Nondeterministic Models: 11. Acting with nondeterministic models; 12. Planning with nondeterministic models; 13. Learning nondeterministic models; Part V. Hierarchical Refinement Models: 14. Acting with hierarchical refinement; 15. Hierarchical refinement planning; 16. Learning hierarchical refinement models; Part VI. Temporal Models: 17. Temporal representation and planning; 18. Acting with temporal controllability; 19. Learning for temporal acting and planning; Part VII. Motion and Manipulation Models in Robotics: 20. Motion and manipulation actions; 21. Task and motion planning; 22. Learning for movement actions; Part VIII. Other Topics and Perspectives: 23. Large language models for acting and planning; 24. Perceiving, monitoring and goal reasoning; A. Graphs and search; B. Other mathematical background; List of algorithms; Bibliographic abbreviations; References; Index.

Editorial Reviews

‘Every serious student of AI will benefit from reading this comprehensive text on some of the most important problems and methods underlying intelligent behavior, written by three world-leading experts.’ Stuart Russell, University of California, Berkeley

‘The book provides a comprehensive, advanced, and up-to-date introduction to the area of intelligent behaviour in AI concerned with acting, planning, and learning. It is also an original and ambitious work of synthesis that integrates classical and hierarchical planning, reinforcement learning, and robotics. An essential textbook and a key reference for anyone working in the field.’ Hector Geffner, RWTH Aachen University, Germany

‘After focusing on automated planning in their first book and its integration with acting in the second book, Malik Ghallab, Dana Nau, and Paolo Traverso close the loop by adding the last missing component, learning. With more than 600 pages of text and more than 1200 references, this massive effort is a must-read for anyone who builds autonomous agents that plan their actions and improve their behavior via learning. It is not only an excellent textbook, but it also serves as a reference book for researchers.’ Roman Barták, Charles University, Czech Republic

From the Publisher