An Introduction to Multi-agent Systems
by Michael Wooldridge, ISBN 047149691X, paperback, £20.99
Also available in an Adobe e-Book edition for £32.75.
I've just finished reading Michael Wooldridge's book on multi-agent systems. As a physicist I've found a great deal of the literature in this area very hard to penetrate, there is a lot of overlap with sociology and several other of the fuzzier disciplines, which means that the hand waving can get a bit much.
The publisher's blurb says that, "This is the first textbook to be explicitly designed for use as a course text for an undergraduate &graduate courses on multi-agent systems. Assuming only a basic understanding of computer science, this text provides an introduction to all the main issues in the theory and practice of intelligent agents and multi-agent systems."
Personally I think that the assertion that the book assumes "only a basic understanding of computer science" is a bit on the far fetched side, you'll need a decent grasp of a high level language, several would be better, a nodding acquaintance with distributed systems, some familiarity with the basic concepts behind artificial intelligence and (rather crucially) a familiarlity with set and logic notation which is heavily used throughout the book to explain a lot of the concepts discussed in the text.
However, if do have the background knowledge required I'd certainly recommend An Introduction to Multi-agent Systems as just that, and excellent formal introduction to agent architectures.
After borrowing it from my graduate student's desk, I recently read the book cover to cover during two tedious train journeys. I picked up a couple of new ideas, and got some formal grounding using set and logic notation on a whole bunch more stuff about which I'm already fairly familiar.
If you're moving into this area and need a book to make the rest of the literature a bit more transparent, you could do a lot worse than this one, if a poor beffudled physcist like me can understand it, anyone can.