Michael Brady of Oxford University (one of the contributors to thisbook) is usually credited with defining robotics as “theintelligent connection between perception and action.” Thisdefinition brings out the three fundamental ingredients of robots:sensing, intelligence, and the ability to act upon the world. All theseingredients are critical in mobile robots. This book is a remarkablecollection of 13 case studies of successful mobile robots, with emphasison their software architectures, vision, mapping, and navigationsystems. Autonomous mobile robots are at the heart of today’s roboticsresearch and development efforts.

It is interesting that the editors emphasize the relation betweenAI and robotics, since this has been a rocky marriage at best for manyyears. It is true that a number of the systems described in this bookwere winners in the AAAI robotics competitions, and that these contestshave helped to bring robotics back to the AI community, from which ithad literally vanished for a number of years. However, even today, manypeople in this community view robotics as a distraction from mainstreamAI, since robot builders must devote a great deal of time tonitty-gritty issues like friction or slipping wheels or the effect ofillumination. With fast mobile robots, it may even be important to lookat the effects of vehicle dynamics, which may be best described bydifferential equations. For these reasons, some people in the AIcommunity prefer to work with softbots, that is, pieces of code thatdisplay some autonomy, without having to concern themselves with thedifficulties of the real world. For these reasons, this book is evenmore remarkable, since it describes successful robots that are embeddedin the world and, moreover, are capable of performing in non-laboratoryreal-world environments. We are not quite ready to launch our mobilerobots into totally unstructured environments, but many of the systemsdescribed here function quite well in natural and even unexpectedsettings.

The authors whose papers appear in this collection are among thebest known and most imaginative roboticists in the field today. I wishthe editors had identified the institutions where the work was done.Clearly, there is a kind of leveling democracy associated withpublishing a collection of outstanding articles without authoraffiliations. Nevertheless, I found it frustrating. The issue is notsnob appeal; rather, affiliation information helps the reader to placeresearchers within a tradition and a local history. Very few of us arecompletely original in our work; most of us extend work done by others.Without an identification of this heritage, a significant part of thebook is missing.

In a short review, it is not possible to discuss each of these 13papers in detail, but they share some interesting features. Eachdescribes a successful robot, that is, one that was successful inperforming a task in a real-world setting. All these robots use sensors,primarily vision and sonar proximity sensors, to obta in informationabout the world. They use a variety of algorithms for avoidingobstacles, creating maps of their surroundings, determining theirlocation in space, and navigating toward a goal. Nearly all of therobots described here use a mixture of reactive behavior withdeliberative planning. Many use some form of machine learning to improvetheir performance with practice. Several chapters deal with fundamentalissues of uncertainty in the interpretation of data about the world,such as the use of Bayes’s theorem by Thrun and his associates (fromCarnegie Mellon University) and the use of partially observable Markovmodels by Koenig and Simmons (also from CMU). All these issues are atthe heart of contemporary research in robotics.

The book begins with an excellent intellectual history of mobilerobotics, written by the editors. The remaining chapters are organizedinto three groups: “Mapping and Navigation,” “Visionfor Mobile Robots,” and “Mobile RobotArchitectures.” Nearly all the chapters describe the robots inconsiderable detail, thus giving readers the opportunity to compare avariety of approaches to the synthesis of mobile robots. The editors areto be commended for assembling a highly readable and comprehensive guideto this rapidly developing field. All serious researchers in mobilerobotics should have this book on their shelves, and should readit.