Sign in to follow this  
KDSBest

Physic Engine

Recommended Posts

Guest Anonymous Poster   
Guest Anonymous Poster
There isn't any physics engine out there supporting general tri mesh to tri mesh collision.
Expect for ODE but we all know what that is.
The best you can do is a course approximation of the geometry by placing convex hulls.

General triangular mesh intersection is a well-understood problem in computational geometry, however this is not sufficient for contact generation.

For example using intersection routines it is possible to calculate contacts but is also possible that those contacts are positions is such way that the generates an ill condition formulization of the contact constraint matrix. That is when two or more contacts are different but each one can be expressed as a linear combination of the others.

Example of that is a box hitting a plane edge on, in general you only need two contacts one each extreme, but if for example you get a contact in between, then the dynamics phase will have a very difficult time figuring out that contact is not essential to the calculation of the resting forces.

I guess this is why not physics engine maker take such risk. In general the implementer only go for what is safe.

Share this post


Link to post
Share on other sites
DirectXFreak    166
Quote:
Original post by MrRowl
Quote:
Original post by Anonymous Poster
There isn't any physics engine out there supporting general tri mesh to tri mesh collision.


Of course there is.


Novodex does...

Share this post


Link to post
Share on other sites
leoptimus    106
And what about your physics engine MrRowl?
It's pretty good, and supports trimeshes. Can you share it to KDSBest?
Can be downloaded from:
http://www.rowlhouse.co.uk/jiggle/

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
Quote:
Original post by MrRowl
Of course there is.

Would that be your engine MrRowl? I do not remember any other capable of that.

Actually Novodex does not do triangular mesh to triangular mesh. Last time I checked they were using a Voxel space technique to create a normal map for a air tight close geometry. ( I think it is called p map)
The method does not scale well with geometry size and it is very memory hungry.




Share this post


Link to post
Share on other sites
MrRowl    2490
Quote:
Original post by Anonymous Poster
Would that be your engine MrRowl? I do not remember any other capable of that.


No of course not :)

Quote:

Actually Novodex does not do triangular mesh to triangular mesh. Last time I checked they were using a Voxel space technique to create a normal map for a air tight close geometry. ( I think it is called p map)
The method does not scale well with geometry size and it is very memory hungry.


Well I would say that's tri-mesh -> tri-mesh. Yes, it's only approximate, but then _everything_ about a physics engine is an approximation.

As for accurate tri-mesh -> tri-mesh I believe the CryEngine physics library does it, but I haven't looked in detail. Actually, if I had looked in detail, I wouldn't be able to comment :)

Share this post


Link to post
Share on other sites
MrRowl    2490
Quote:
Original post by Rompa
TrueAxis does trimesh-trimesh (swept, of course), though its not open source.


Are you sure? I only looked quickly, but the current version seems to support:

CollisionObjectComplex: Provides an representation of large static arbitrary meshes.

The following classes inherit from TA::CollisionObjectSimple:

TA::CollisionObjectConvex;
TA::CollisionObjectCapsule;
TA::CollisionObjectCylinder;
TA::CollisionObjectSphere; or
TA::CollisionObjectLineList.

And CollisionObjectCombo can contain any of CollisionObjectSimple.

So based on this it doesn't support dynamic arbitrary tri-mesh -> dynamic arbitrary tri-mesh, unless the user manually splits the dynamic collision shapes into convex hulls (or other simple shapes).

Bit of a diversion from the original question though :)

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
list of teh existing solutions
http://www.diku.dk/~kenny/other.html

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
Quote:
Original post by Anonymous Poster
list of teh existing solutions
http://www.diku.dk/~kenny/other.html


wow I love the description of the last one on that list.
http://81.57.26.38/
Quote:
Next generation physics engineSince 1999, we aim at developing a formalism to modelize any physical phenomena. Our first implementation of this formalism, PGNX Physics, is probably the first fully unified physics engine and can simulate any phenomena from quantum physics to realistic video-game interactions.

That is what I call the ultimate physics solutions. Maybe they discovered the unified fid theory or something :-)

The problem with open source solutions is that in general that leave a lot to be desired.




Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this