REVIEW - AI Techniques for Game Programming

Title:

AI Techniques for Game Programming

Author:

Mat Buckland

ISBN:

193184108

Publisher:

Premier Press (2002)

Pages:

448pp

Reviewer:

Daire Stockdale

Reviewed:

June 2003

Rating:

★★☆☆☆

The title of the book is slightly misleading and I think it would be better titled something like 'An introduction to Neural Networks and Genetic Algorithms'. The reader will not find mentioned many techniques used by most game engines, such as node based path-finding or A*. I would have hoped a thorough title on programming game AI would also mention 'bot' programming, state machines, virtual machines, aides to debugging AI, etc. none of which makes it into this book.

However, as an introductory book to the difficult topic of neural networks and genetic algorithms, I think this book is very useful. The author knows his subject well and explains it helpfully, accompanied by many examples on the book's CD. He codes in clean, well-written and nicely commented C++. This book would definitely be useful to anyone without experience of neural networks and genetic algorithms and who was thinking about using them in an application. These topics are covered in sufficient depth that even those who feel competent in this area might learn a thing or two. I would like to see the author write something a little more serious and definitive, perhaps aimed at the harder-core coding audience.

The book suffers from feeling it must cater even for readers with absolutely no experience of Win32 or 3D math programming by covering topics such as vectors, GDI and the Windows message loop. The first 20% of this book has nothing to do with AI whatsoever. This seems to be a common flaw in all the books I have seen in Premier Press' game development series. It would be far better if they devoted one title to this subject rather then force their customers to repeatedly buy the same material in every book.


Book cover image courtesy of Open Library.