[odejava] Weird GeomTriMesh problems

Game Mechanics
#61

Anybody having the same problem?

DP

#62

I have TriMesh problems too… and it’s very annoying ( you really can’t do a game engine with that ).
I’m dropping a sphere on a trimesh terrain, and the sphere react normally, until it falls through the ground.
I’m using the newest Odejava CVS version, and the latest binaries.

#63

I’ve that also, but it gets better, if you test more often for collisions I think. It happens more often when things are moving fast, right?

#64

Yeah we really should have something like a GeomTerrain, which automatically is able to correct that.

#65

Maybe its about time Odejava changed its policy about ODE modules, e.g. the GeomTerrain module and incoorporate it in the build…

DP

#66

What would be the difference between GeomTrimesh and GeomTerrain ??
I solved my problem by smoothing the 3D model ( not good for performances… ), but it isn’t a good solution.

#67

If you have a GeomTerrain, you know that no elements can be beneath it, so ode can always correct that. As an example I’ve never seen anything get stuck in a GeomSphere. (Ok I haven’t tested it)

#68

Oh I see.
And you say in ODE there is a GeomTerrain ? And not in OdeJava ?

#69

No - it’s the other way round there is no such GeomTerrain in ODE, so we can’t have a GeomTerrain either.

#70

I believe ODE does have a GeomTerrain, but its a contributed module.

Ode.dCreateTerrain is the ODE method.

DP

#71

Yep I also heard about that, but I think it doesn’t also work for the ODE guys. Maybe we’ll just have to wait some time :wink:

[quote]Ode.dCreateTerrain is the ODE method.
[/quote]
There is no such function.

#72

There is no such function…
[/quote]
…In Odejava

#73

Any one solve any of this yet?

#74

I think there won’t be any fix for ODEJava.
But we at Joode are trying to recode these things right.
@t_larkworthy : will your arbitrary polygon detection collision handle the terrains cases well ?

#75

Yeah I am hoping for that. By partially defining the Gauss representation for an object then the potential anti-podal(opposite) featurse can be limited to only some directions, thus only in some directions can a penetration be measured. It should be trivial to implement a plane like that, but because the system I am working on is only for convex polyhedra, creating a terrain from it is not staightforward. (but obviously there is still a break things up into convex peices algorithm that could be done in the future)

#76

So we need to have an algorithm for dividing concave polyhedras to convex ones… That won’t be a piece of cake…

#77

Its failry easy doing a poor one, just keep cutting the polyhedra until every part is convex. I am hoping the effeciency of the ConvexConvex collision algorithm will of make up for it.