AI Game Programming Wisdom 2
Author: edited by Steve Rabin
Publisher: Charles Rivers Media
ISBN: 1-58450-289-4
Purchasing: [Amazon.Com] - US$69.95, [Amazon.co.uk] - UK£32.87, [TransAtlantic Publishers] - UK£46.95.
Reviewed: 30th January 2004

Front Cover Shot:

Overview

This review is for the second in a popular series, unlike several other books recently reviewed that are second editions, this is all new material - continuing the trend set by the first.

The first book was a very popular resource for AI programmers everywhere - and given it was simply a dedicated spin off from the larger "Game Programming Gems" series, it has done remarkably well. Such that it would appear, and in no bad way, that this series will become a regular (if not annual) publication.

There is a lot to be said about AI - especially game related AI. We've pretty much cracked advanced graphics, end users aren't noticing the new techniques as much as they did a few years ago (they've gotten used to flash graphics!), whereas now AI is getting more attention. Quite simply, its a big field - more than enough to justify a second release in this series.

For reference, the first book in the series was reviewed here.

Familiar Tale

As said, this is an extension to a popular series, and for this reason it is a fairly familiar tale - the content might be new, but the design, goals and general style of the book has not changed significantly.

One of the most immediately obvious things about this book compared with its predecessor is that it is bigger. The first book was 672 pages, compared with 732 in this new release; there's only 60 pages difference - but physically it is surprisingly noticable. This release has 4 less articles in it, but an extra section - which can be seen as slightly lengthier articles covering a slightly wider field.

The general overview of game AI, represented by the various chapter headings, is refreshingly varied by comparison to the first release. It was distinctly noticeable in the first book that large sections were dedicated to pathfinding (A* in particular); granted, pathfinding is an immensely important and suitably big part of game AI - but for this type of book variety should be king. With this installment each chapter heading is obviously different from the rest - knowing that whilst you may get a few less articles, you do get 12 chapters covering 12 different areas of AI.

For reference, the 12 areas are:
1. General Wisdom
2. Pathfinding and Movement
3. Group Movement, Tactics, and Planning
4. AI Animation Control
5. State Machines
6. Architecture
7. FPS, RTS, and Strategy AI
8. Racing and Sports AI
9. Scripting
10. Learning
11. Genetic Algorithms and Neural Networks
12. Speech Recognition and Dialog

Some articles obviously overlap in scope or content, and no chapter has a definite/complete coverage of the entire field; but it does fit together well given the nature of this book (and series).

Consistently Mixed

One of the trends that is common in these collection like books is that with many different authors, you get many different styles of writing. As such, on my part as reviewer, it is quite difficult to rate the writing style as a whole. But in the case of these books, it would be safe to say that the writing style is good to excellent.

Steve Rabin, as editor of both books now, deserves credit for keeping the content diverse yet maintaining a high level of quality. The individual authors are to be respected for their views - as this book very much has an air of being "by the industry, for the industry"; many people who write for this series also work on the games we play in our spare time. Have a look at this webpage - part of the supporting website for this series; on the page it lists all of the games that the authors have been involved in - whilst not all of them are industry veterans, there is certainly some high-profile writers.

More importantly, its far better than the ramblings of the theoretical researcher type authors - not to undermine the quality of their work, but it is refreshing to have a book like this that is built mostly on the real-world experiences of talented people.

Pre-Requisites

The one big negative against this book, or more generally at the series as a whole is it's level of pre-requisite. It is a collection of hints, tips and ideas - but none of it could really be described as beginner level; it is firmly aimed at those who have at least a rough idea of what they're doing.

If you've never done AI programming, or don't really have a significant interest in AI, then this book is probably not for you. A more traditional AI programming book is probably more suitable to learn from - this book is for when you've learnt AI, and are ready to build in some of the more impressive features of award winning game AI.

Also, you do need to be a competent C++ programmer to make the most of this book - almost all of the source code both in the book and on the CD is in C++. If you're not too keen on C++, or just plain not that good with it yet you'll probably find it difficult to get much out of the applied sections of this code. However, I have successfully converted several articles from the first book into VB6 code with only minimal understanding of the C++ language.

In Conclusion

This book is without doubt a worthy sequel to the first installment; and if the editor can keep up the high level of diversity and quality then I look forward to the third in the series.

The unfortunate thing about this book is that it comes across too much like a reference text, rather than a traditional computer/game book with  a beginning-middle-end structure. Even if you do relegate it to being a reference book, it will be one of the best reference books you can have.