AI for Game Developers
Rating: 4 out of 7.
An interesting book about artificial intelligence for game developers describing algorithms for weak AI. Interesting, but don't expect too much from it. (Not yet finnished)
The book's title is correct though misleading. The book doesn't cover all aspects of AI, but only weak AI. Neural networks are only a side chapter. Besides, the examples are for MS Windows and many of them won't compile on other operating systems. This is completely unexpected for a book from O'Reilly and partly responsible for the poor rating that I give this book. Besides the source code for the examples is pretty much undocumented and written in a very antique style of C++. It would've been good if the authors would've visited some OO programming and design pattern courses before using C++.
Content
The book has 15 chapters with the following topics:
- Introduction to Game AI
- Chasing and Evading
- Pattern Movement
- Flocking
- Potential Function-Based Movement
- Basic Pathfinding and Waypoints
- A* Pathfinding
- Scripted AI and Scripting Engines
- Finite State Machines
- Fuzzy Logic
- Rule-Based AI
- Basic Probability
- Decisions Under Uncertainty — Bayesian Techniques
- Neural Networks
- Genetic Algorithms
All except one chapter have been written in a way that you can read the chapter without having read the previous ones. This increases the book's quality for looking up AI algorithms and solutions for specific tasks without having to read the book from its beginning to its end. I wouldn't say the style of writing is great but it's sure more than good enough to not leave a negative impression.
Comments on the individual chapters
Neural Networks
Since this is only a single chapter, readers cannot expect more than an example for one single implementation. More than just showing the source code examples and explaining them, the authors also tried to present the formulae and explain them. But with the latter, they horribly failed. The formulae are presented without a proper explanation of the symbols used, which gives the impression of poor copy-paste from other books. I was only able to understand the formulae by applying the knowledge that I already had about neural networks.
Conclusion
I wouldn't say "don't buy this book". But be prepared that the examples are unprofessionally poor and non-portable and that you can't expect too much from this book.
View this book at
