AUGUST 2006 (Vol. 7, No. 8) 3
1541-4922/06/$31.00 © 2006 IEEE
Published by the IEEE Computer Society
Published by the IEEE Computer Society
Book Reviews: A Primer on Computational Fields
PDFs Require Adobe Acrobat
Field-Based Coordination for Pervasive Multiagent Systems
Marco Mamei and Franco Zambonelli
Field-Based Coordination for Pervasive Multiagent Systems by Marco Mamei and Franco Zambonelli is a highly specialized research book. It's certainly not an introductory text, such as, say, Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence (Gerhard Weiss, ed., MIT Press, 2000). However, the mathematics, pseudocode, and code are easy to follow, requiring only intermediate knowledge of vector field theory and programming—particularly, Java programming.
Mamei and Zambonelli begin by explaining the concept of virtual computational fields, which generally can create information gradients that agents can exploit by being aware of the local gradient. Within this virtual field, agents can self-organize without resorting to a central communication system. Applications that use those information gradients include modular robots, amorphous computing, artificial worlds, and PDA networks.
Mamei and Zambonelli compare the field-based approach with direct-coordination, shared-data-space, and event-based models of multiagent system development, concerning their effectiveness (or lack thereof) for attacking the problems of multiagent adaptive self-organization and context awareness.
The authors use examples to explain the way a computational field is created. Computational fields can be described through the mathematical concept of a gradient. Once the gradient is established, you can solve the equations of motion for each agent. This eventually leads Mamei and Zambonelli to ask a general question: What kind of coordination field is required to let an agent move along a specific trajectory? According to the authors, the answer to this question isn't easy, although they feel that solving the equations of motion—at least approximately—might hint about the kind of field to use. They describe corrections to the initial linear modeling, including how to escape from spurious local minima. Computationally speaking, field-based coordination is implemented through middleware, which Mamei and Zambonelli discuss thoroughly.
The authors mention commercial middleware, but they discuss Tuples On the Air in more detail. TOTA is the middleware the authors developed for solving the problem of implementing computational fields. In fact, half the book uses TOTA as its main paradigm. At the end, Mamei and Zambonelli describe a possible way to use TOTA for making an object invisible and the potential applications of such a device, ranging from military to fashion.
The book is much more readable than the authors' Web site (http://www.agentgroup.unimore.it ), from which you can download their programs. The site is poorly maintained, and many pages are no longer available. When you do find a page from which you can download something (for example, MARS.tar.gz), the download is pretty fast. Installation is cumbersome, however, because the instructions are in Italian. If you do speak Italian, brace yourself and be patient, because you must have some basic software installed before you can actually run the examples offered in the MARS.tar.gz package. To get things done, I think your best bet is probably to get in contact with the authors. In the instructions, the authors make a funny reference to Gödel, Escher, Bach: An Eternal Golden Braid by Douglas Hofstatder (Basic Books, 1979), and play with the words "Domini" (Lords) and "domains" (the mathematical concept).
This book is easy to follow without resorting to a pedagogical approach (such as that of Multiagent Systems, which has exercises at the end of each chapter). If you just want to understand the concept of computational fields and see examples of their application (and even understand how the "squids" in the film The Matrix work together as a single organism). If you want to continue your research in that direction, you'll find a wealth of references to additional information. But if you would like to use TOTA, I would probably go first to Mamei and Zambonelli's Web site and read whatever pages are available. Once you're convinced this is what you want to aim at in your research, go buy this book … and contact the authors immediately afterwards.
Arturo Ortiz-Tapia is a scientific researcher at the Mexican Petroleum Institute. Contact him at firstname.lastname@example.org.