The book Real Time Rendering 3rd ed. outlines the system Flagship Studios used for Hellgate: London. It says the system uses 3 shadow maps: 1. One adjusts to the view's focus and frustum and includes just the animated objects. 2. The second is associated with a grid section of the level and includes only static objects, and is selected by the view position. 3. The third covers the entire level's static objects. The last two make sense to me. I do not quite understand how the first will work. The shadow map storing animations cannot cover the entire frustum, otherwise you won't have enough resolution. So if this shadow map focuses on covering the front of the frustum near the eye, then what happens to animated objects far away, but visible? They don't get shadowed?

Quote:
It says that "Nearby animated objects are usually the focus of attention, and by this scheme, these will cast the highest quality shadows."

For me it seems that only nearby animated objects are shadowed by the 3rd shadow map.

That is sort of how I was interpreting it. But animated objects cannot go in the other two shadow maps since the idea is that those shadow maps will not be regenerated very often. So to me it sounded like there was a problem because far away animated objects wouldn't cast shadows at all.

Quote:Original post by Quat
The shadow map storing animations cannot cover the entire frustum, otherwise you won't have enough resolution.

Sure it can. Many many games do this. You'll just want to use some sort of technique that optimizes the distribution of the shadow map across the view frustum. The most commonly-used technique is cascaded shadow maps, and there's also trapezoidal shadow maps.

For FPS games, where the view is newer high above the ground, and the light comes from a relatively high angle, then trapezoidal SM is the best/simplest method. If you can go high above the ground (airplane for example), and you can see extremely far, then cascade shadowmaps are commonly used (for example Armed Assault I).