Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Akaguma]

Your engine uses different LOD for terrain patch depending on distance from player. Is it right?
So, how do you calculate the height (coordinate along upward axis) of objects (trees, cars, etc), stationed on the ground?
I mean, do you need to recalculate the height of trees for example when terrain LOD became more or less detailed?
I think its critical for multiplayer game, if you plan it.

[cameni]

LOD for terrain patches is determined by a metrics that computes a required texel/pixel quality for a given view. Broadly speaking the LOD depends mainly on the distance, but it's also greatly affected by other factors like the terrain shape and current quality settings and the limiter state and such.

Using it dynamically for interactions would cause a great havoc, but of course physics and positioning are not dependent on this. Fixed detail levels are used for physics as well as for building/tree placement, and the objects won't move as the detail changes.

[knackered]

so when you generate a tile far from the viewer, at a coarse LOD, say 100m per pixel, and then you instantiate trees on the tile....you're saying that the vertical displacement of the tree doesn't change as the viewer approaches it? Surely as the terrain tile gets subdivided and refined this will result in the tree sitting in the sky?
also, for physics objects, are you saying you pick say a 1m per pixel resolution, then generate 1mpp tiles around each physics object no matter how far from the viewer?

[cameni]

That's a bad example, you don't instantiate trees on tiles that are 100m per pixel. Tree placement is determined on tiles that are roughly 10m per pixel, so there's one instance of tree in a 10x10m cell. While the terrain elevation at the position of the tree certainly changes as terrain gets refined, unless it was on a cliff edge it stays within bounds. It is a task of the probability function to avoid some corner cases where such errors could appear.

How the physics is handled will depend on the game. I was only saying that it is reasonable to simulate vehicles using a constant resolution mesh so that the physics is predictable. In this regard, if a game still wants to simulate a vehicle even if it's gone far, the engine will have to generate the terrain for it. However, there are several levels of detail that can be used for physics and it's up to the game what level will be used when. A game can even revert to a statistical mode when the vehicle is far away and nobody sees it. In a multiplayer environment the players that have the mesh of the required level can be the authorities that provide outputs for given vehicle.

In other words, many of these things will have to be solved in practice for this huge game environments.

[Wolfhound]

Hello, I just recently discovered this project and must say it looks incredible.

Will it be possible for users to integrate complex electromagntic effects for infrared and microwave/radar with the terrain model e.g differing reflectivity of terrain and temperature effects (different objects heating and cooling at different rates etc)

Thanks

[cameni]

Um, I think something like that would require a custom development, though some related work is planned for LIDAR functionality .. but that's outside of the game engine area

[Wolfhound]

Um, I think something like that would require a custom development, though some related work is planned for LIDAR functionality .. but that's outside of the game engine area

Thanks for your reply cameni.

The effects I mentioned would be for flightsim/game development such as is seen in current commercial flightsims.  I just wondered if similar features would be possible and if those features could be expanded on for greater fidelity?

[cameni]

I would need a deeper info on what is going on and what are the requirements there, but if it's mostly about assigning a material info properties to terrain and objects and then being able to query it, it's surely doable. But I don't see what is the problematic part there, and what level of fidelity would be desirable (and what's the current state).

[Wolfhound]

Yes, I believe thats what is required, material properties such as reflectivity, smoothness density etc.

Cheers

[Spudly]

I am curious how you guys are currently doing the terrain physics.  This seems like it is difficult to do with large levels, let alone this large.  Are you using a constantly updated/tiled heightfield or a mesh or something else?

[cameni]

Yes, from the physics point of view it's a tiled heightfield, with the tiles being dynamically managed as required.