Many other books have covered similar material, for physics engines, rendering engines and for AI. I have read none that have really done the subject justice, looking in depth at the techniques and providing background information to make sensible decisions. This book changes that: it is a superb, detailed, accurate (as far as I have checked so far) and timely account of one of the most awkward bits of modern game construction.The biggest strength of this book is its broad range. Ericson steps through the whole gamut of collision detection tasks: from simple collision tests for bullet intersections, through to accurate collision manifold generation for moving objects (the kind of collision data needed for high end physics simulation).
The book includes a wealth of information on geometrical primitive collision detection (working out when a sphere and plane collide, and so on), but so do many other books. He includes information about spatial data structures and how to work with them, covered well in only a handful of other books. He also looks at combining them all together, with really powerful optimisation, and how to write code to automatically generate the collision geometry and high-level data structures from the underlying geometry. I've never seen this covered in more than an off-hand paragraph before, and it makes this book worth double its price tag for serious game developers.
I've worked in the games industry for a long time, and I learned a lot from this book. Already some of his suggestions have made it into the game engine my company owns, and have made a performance difference. I can't think of higher praise than that.
Its not aimed, I don't think, at hobbiest game programmers, because it really handles the subject in great depth (although the simpler algorithms would be useful). It is a serious book for serious developers, and I wouldn't hesitate to recommend it to anyone working at the sharp end of games.
