Skip navigation
Hardcover | $42.00 Short | £28.95 | ISBN: 9780262122962 | 400 pp. | 7 x 9 in | 149 b&w illus., 14 tables| April 2012
 

The Soar Cognitive Architecture

Overview

In development for thirty years, Soar is a general cognitive architecture that integrates knowledge-intensive reasoning, reactive execution, hierarchical reasoning, planning, and learning from experience, with the goal of creating a general computational system that has the same cognitive abilities as humans. In contrast, most AI systems are designed to solve only one type of problem, such as playing chess, searching the Internet, or scheduling aircraft departures. Soar is both a software system for agent development and a theory of what computational structures are necessary to support human-level agents. Over the years, both software system and theory have evolved. This book offers the definitive presentation of Soar from theoretical and practical perspectives, providing comprehensive descriptions of fundamental aspects and new components.

The current version of Soar features major extensions, adding reinforcement learning, semantic memory, episodic memory, mental imagery, and an appraisal-based model of emotion. This book describes details of Soar’s component memories and processes and offers demonstrations of individual components, components working in combination, and real-world applications. Beyond these functional considerations, the book also proposes requirements for general cognitive architectures and explicitly evaluates how well Soar meets those requirements.

About the Author

John E. Laird is John L. Tishman Professor of Engineering in the Computer Science Division at the University of Michigan.

Table of Contents

  • The Soar Cognitive Architecture
  • The Soar Cognitive Architecture
  • John E. Laird
  • The MIT Press
  • Cambridge, Massachusetts
  • London, England
  • © 2012
  • Massachusetts Institute of Technology
  • All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher.
  • For information about quantity discounts, email special_sales@mitpress.mit.edu.
  • Set in Stone Sans and Stone Serif by Toppan Best-set Premedia Limited. Printed and bound in the United States of America.
  • Library of Congress Cataloging-in-Publication Data
  • Laird, John, 1954–
  • The Soar cognitive architecture / John E. Laird.
  •  p. cm.
  • “A Bradford Book.”
  • Includes bibliographical references and index.
  • ISBN 978-0-262-12296-2 (hardcover : alk. paper)
  • 1. Artificial intelligence. 2. Software architecture. I. Title.
  • Q335.L329 2012
  • 006.3—dc23
  • 2011030567
  • 10 9 8 7 6 5 4 3 2 1
  • to my parents, Margaret L. and John R. Laird; my wife, Ann Alpern; and my daughters, Emily, Jessica, and Valerie
  • The secret of success is constancy of purpose.
  • —Benjamin Disraeli
  • In the computer field, the moment of truth is a running program; all else is prophecy.
  • —Herbert Simon
  • Contents
  • Preface xiii
  • Acknowledgments xv
  • 1 Introduction 1
  • 1.1 Background 3
  • 1.2 Cognitive Architectures 5
  • 1.3 Soar 17
  • 1.4 Research Strategy 22
  • 1.5 Preview of Chapters 2–14 24
  • 2 Requirements for Cognitive Architectures 27
  • 2.1 Characteristics of Environments, Tasks, and Agents 29
  • 2.2 Architectural Requirements 32
  • 3 The Problem-Space Computational Model 43
  • 3.1 Task Environments 44
  • 3.2 The Problem-Space Framework 47
  • 3.3 Knowledge Search 53
  • 3.4 Problem-Space Computational Models 55
  • 3.5 Impasses and Substates 63
  • 3.6 Using Multiple Sources of Knowledge 66
  • 3.7 Acquiring Knowledge 67
  • 3.8 Alternative PSCMs 68
  • 4 Soar as an Implementation of the PSCM 69
  • 4.1 Production Systems 70
  • 4.2 Mapping Production Systems onto the PSCM 72
  • 4.3 The Soar Processing Cycle 78
  • 4.4 Demonstrations of Basic PSCM 97
  • 4.5 Discussion 107
  • 4.6 Analysis of Requirements 116
  • 5 Impasses and Substates: The Basis for Complex Reasoning 119
  • 5.1 Impasses 120
  • 5.2 Substates 121
  • 5.3 Problem Solving in Substates 122
  • 5.4 Substate Results 125
  • 5.5 Maintaining Consistency 129
  • 5.6 Demonstrations of Impasses and Substates 130
  • 5.7 Discussion 155
  • 5.8 Analysis of Requirements 156
  • 6 Chunking 159
  • 6.1 Chunking in Soar 160
  • 6.2 Implications of Chunking in Soar 164
  • 6.3 Demonstrations of Chunking 166
  • 6.4 Assumptions Inherent to Chunking 175
  • 6.5 Analysis of Requirements 179
  • 7 Tuning Procedural Knowledge: Reinforcement Learning 181
  • 7.1 Reinforcement Learning in Soar 183
  • 7.2 Learning over Large State Spaces 189
  • 7.3 Demonstrations of Reinforcement Learning 191
  • 7.4 Analysis of Requirements 202
  • 8 Semantic Memory 203
  • 8.1 Semantic Memory in Soar 207
  • 8.2 Encoding and Storage 209
  • 8.3 Retrieval 213
  • 8.4 Demonstrations of Semantic Memory 215
  • 8.5 Analysis of Requirements 223
  • 9 Episodic Memory 225
  • 9.1 Episodic Memory in Soar 227
  • 9.2 Encoding and Storage 229
  • 9.3 Retrieval 230
  • 9.4 Use of Episodic Memory 232
  • 9.5 Demonstrations of Episodic Memory 234
  • 9.6 Comparison of Episodic Memory and Semantic Memory 243
  • 9.7 Analysis of Requirements 245
  • 10 Visuospatial Processing with Mental Imagery 247
  • 10.1 Visual and Spatial Representations 249
  • 10.2 Visuospatial Domains 252
  • 10.3 SVS 254
  • 10.4 Demonstrations of Spatial and Visual Imagery 265
  • 10.5 Analysis of Requirements 268
  • 11 Emotion 271
  • 11.1 Appraisal Theories of Emotion 272
  • 11.2 Abstract Functional Cognitive Operations 274
  • 11.3 Unifying Cognitive Control and Appraisal 277
  • 11.4 Emotion, Mood, and Feeling 278
  • 11.5 Emotion and Reinforcement Learning 279
  • 11.6 Demonstrations of Emotion Processing 280
  • 11.7 Analysis of Requirements 284
  • 12 Demonstrations of Multiple Architectural Capabilities 287
  • 12.1 Learning to Use Episodic Memory with Reinforcement Learning 287
  • 12.2 Using Mental Imagery with Reinforcement Learning 294
  • 12.3 Diverse Forms of Action Modeling 299
  • 12.4 Analysis of Requirements 304
  • 13 Soar Applications 307
  • 13.1 Applications 307
  • 13.2 TacAir-Soar 313
  • 13.3 Imagining TacAir-Soar 2.0 317
  • 14 Conclusion 325
  • 14.1 Soar from a Structural Perspective 326
  • 14.2 Soar from a Functional Perspective 328
  • 14.3 Evaluating Soar on Architectural Requirements 331
  • References 347
  • Index 367

Endorsements

"John Laird's book gives a complete account of the momentous developments that have occurred in the Soar Cognitive Architecture. This book is a must-read for researchers and students who are interested in the grand goals of Cognitive Science and AI. "
John R. Anderson, Carnegie Mellon University

"John Laird has been at the forefront of research on cognitive architectures since the early 1980s and this book is a culmination of nearly 30 years of work. The book is a substantial achievement and a fine synthesis of the author's work. It provides both a study of how integrated computational mechanisms can generate intelligent behavior and a renewed opportunity for cognitive science to pursue integrated theories. It is essential reading."
Andrew Howes, School of Computer Science, University of Birmingham

"A clear and comprehensive account of decades of effort aimed at understanding intelligence and building intelligent systems. The detailed discussion of cognitive architectures, the enumeration of criteria for judging cognitive architectures, and the description of SOAR make this a book that belongs in the library of everyone seriously interested in AI and its applications."
Patrick Henry Winston, Ford Professor of Artificial Intelligence and Computer Science, Massachusetts Institute of Technology

"The SOAR enterprise is one of the most interesting big bets in Artificial Intelligence and Cognitive Science. John Laird's impressive book provides an excellent synopsis of the important ideas, results, and new directions in SOAR research. Anyone interested in the computational modeling of minds should read this book."
Ken Forbus, Northwestern University