Continuous LOD (Level Of Detail)

How is it implemented ?

Does anyone has a paper on it or some code ?

I’m doing a massive strategy game and those far units doesn’t need as much polygons as closer ones…

I know jME has continuous LOD implemented, could we port it ?

Any infos welcome.

For now I have the feeling it should be implemented with a cached version of the initial geometry, then being called each step (this part should be a Behavior), computing the geometry which will actually be displayed then update it via a GeometryUpdater.

Of course we should enable LOD nodes to share initial geometry.

Is it needed to compute several versions of the model at the beginning so we can smoothly interpolate from one to another (but how ?) (just what is done with MipMapping, or am I wrong ?)

You simply cannot interpolate between geometries with different polycounts.

Continuous LOD is old-tech though, as it is slow because it requires new geometry to be sent to the GPU every frame

You should just have a couple of LODed versions of your model, and switch among them at the proper distance. You know you finetuned the poly-count/distance properly, when you can’t see the switch anymore.

The most recent GPUs can do c-lod, but you might wonder whether that’s worth the effort of developing when the alternative is so plain simple.

Ah ok.

Do you know some tools permitting to reduce the number of polygons in a model (is there a feature in Blender pemitting that ?), cause if there’s not I’ll have to ask my artists to make again their models in low-poly, then smooth/correct them gradually for higher-polygons versions.