Not to fan the LOD flames, but I gotta stick up for my favorite algorithm here. :)
Curtis L. Olson wrote: > Recently there has been a lot of work on continuous level of detail > schemes. The stuff I've seen however has been great for demos and > certain games, but there are serious issues in using this sort of > scheme for a flight simulator that needs accurate terrain mapping. > > We need to cut holes for airports. That sounds like an assumption to me. :) Why not drape them over the LOD triangles with a polygon offset? This isn't terribly cheap, but the complexity goes as the number of triangles in the final tesselation. So long as the "draped" complexity isn't too high, this seems doable to me. Even cutting holes isn't too bad if you think about it -- I bet a nifty stencil buffer hack could be effective. > We need lots of objects/buildings on top of the scenery. Just draw them where they should be and forget the differences in the local ground polygons. See below for a discussion of why the "LOD error" is a myth. The LOD error is measured in "real" coordinates, not screen coordinates. In pixels, the error is (by definition) smaller than you care about. > We need to be able to fly seamlessly across the entire world and be > able to store the data for the entire world on a single computer. Now, this is a problem. Not the storing of data -- heightfields are going to be smaller than the current tile format, not larger. But the gridding and stitching of multiple LODs into a single all-the-way-to-the-horizon mesh is difficult, as is the generation of textures to draw on those polygons. I've gotten the first part to work acceptably in the past. I've never attacked the texture generation problem. > If you are doing some sort of combat, what happens if your opponent > flies behind a hill in the distance, but your renderer has removed or > simplified the hill in the distance because of the LOD scheme. This one is basically a myth. If your CLOD scheme has an error measurable in the 5-10 pixel range, then it's not doing its job. For earth-like terrain, you can get that accuracy with a 1000 triangle tesselation; current hardware beats that by at least an order of magnitude. This idea is an impression left over from the days when polygon budgets mattered; it's no longer an issue. Sub-pixel LOD accuracy is acheivable. I'm not saying FlightGear should jump on the LOD bandwagon right now. It would involve *big* code changes and lots of effort that can more productively be spent elsewhere. But it's not nearly so hard to do as its reputation implies. Andy -- Andrew J. Ross NextBus Information Systems Senior Software Engineer Emeryville, CA [EMAIL PROTECTED] http://www.nextbus.com "Men go crazy in conflagrations. They only get better one by one." - Sting (misquoted) _______________________________________________ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
