Not raycast wheels

This doesn't answer your question, but: raycast wheels don't have to be simple (or indeed fast!). You could cast a volume (e.g. sphere or cylinder) instead of using a ray to give you better interaction with the ground (going over kerbs), or multiple rays. You can use the four (for a car) raycasts to generate suspension forces at each wheel, but apply drive forces centrally to the car, or you can apply individual drive/friction forces at each wheel. You can calculate the wheel forces independently (which can result in slipping artefacts) or solve them simultaneously (similar to what you'd do with simultaneous rigid body contacts).

I.e. there's many options about what you do - I'm not sure there are any significant effects that you can't represent with raycast-like wheels, if you put the effort in.

Using separate bodies for the wheels may intuitively seem more realistic, but most physics engines will have a hard time dealing with (a) objects that are rotating very fast, and (b) interactions between objects with large mass differences (compare the wheel to car mass ratio).

I think a common approach is to cast the Minkowski sum of a disc and sphere instead of rays.

Quote:

Original post by MrRowl I.e. there's many options about what you do - I'm not sure there are any significant effects that you can't represent with raycast-like wheels, if you put the effort in.

I think the same. But my boss make the final decisions and he wants to know the _names_ of released games which use sincere wheels - it is just to compare.

itten: if you didn't read it yet - i sent you a private message, i hope it helps finding for what you are looking for.

Something to take into account is that many physics engines have a maximum angular velocity per simulation step, that is too low to simulate racing cars. At least Newton Dynamics and Havok have them, so other open source and commercial engines probably have them too. In a racegame I was working on this meant you couldnt go faster than about 60 MpH at the average simulation framerate (about 60 fps).

Of course there are ways to improve on this such as using bigger wheels for the torque part, or having multiple simulation steps per graphics frame (which doesn't work very well if you're already cpu-bound because of the physics and other parts of the game). Anyway, these methods only give you more problems and inconsistencies.

[Edited by - Tree Penguin on June 29, 2008 11:28:18 AM]

Quote:

Original post by Tree Penguin Something to take into account is that many physics engines have a maximum angular velocity per simulation step, that is too low to simulate racing cars. At least Newton Dynamics and Havok have them

What's your source on that newton has limit for max angular velocity per simulation step? that information is totally new to me, perhaps you can tell me where you got it?

If what you are saying is true, then why does newton have such great multibody vehicle simulation?

An example:
http://newtondynamics.com/forum/viewtopic.php?t=4498 (HIT DEMO #5)

Quote:

Original post by Delfi An example: http://newtondynamics.com/forum/viewtopic.php?t=4498 (HIT DEMO #5)

The insane fov in that demo makes it difficult to tell the speed limit for those cars (it makes it look faster than it really is). Also, the framerate is quite high, and as the limit is per frame (simulation step), that makes the speed limit higher also. It looks like that demo is limited by it, but it's hard to tell without sourcecode.

Here is a thread about what I mentioned. It seems however, that if you do use Newton's special car joints, the problem is fixed.

Anyway, that thread also states that it will be fixed in the next release, which is possibly out already, so I'll keep my mouth shut ;).