[ODE] Faking a wheel joint

[Nate W]

On Wed, 2 Jul 2003, Max wrote:

> 14.04.2003 17:27:15, "David Whittaker" <david@csworkbench.com> wrote:
> >You take a hinge joint with a vertical axis, attach it to the chassis and
> >an extra body, at the same position as the wheel.  Then you attach a
> >slider from that extra body to the wheel body, set both its stops to 0,
> >and adjust the StopCFM and StopERP parameters as you would for the hinge2
> >suspension parameters.
> 
> Has anyone tried this with success ?
> 
> Btw, Nate, do you experience the "bendy wheels" problem with your
> double-wishbone suspension, too ?

Yes, the double-wishbone buggy will show the bendy wheel problem above a
certain speed.  The rest of the car holds together well, suspension and
all, but the axle joints get very erratic.

However, if I was to start work on a racing game (the thought has crossed
my mind), I'd try using regular hinge-2 joints, with the wheel-rotation
axis locked.  Friction between the wheel and environment would have to be
calculated 'manually,' using FDir1 and FDir2 to get the sorts of forces
you'd expect from rolling tires.  The advantage is that the bendy wheel
problem won't happen, because the wheels would never spin at all.  The
drawback is that you'd have to compute your own wheel-ground friction
force (which you would probably want anyhow for a racing game), and you'd
have to compute a 'wheel rotation matrix' for the rendered mesh (so you
can draw spokes rotating realistically at lower speeds).

I think that is basically what David was describing, but, using a Hinge-2
joint (rather than a hinge, slider, and intermediate body) should require
a bit less computation per world step.

It might be possible to use unlocked Hinge-2 joints and normally
calculated friction at low speeds, then transition to the above scheme
when wheel rotation speed approaches the bendy-wheel threshold.  That
would save you the trouble of deriving the wheel orientation at low
speeds.  When the speed hits the limit, you could lock the wheel
orientation, compute your own friction, and draw some sort of blurred
wheel texture.  Then again, it would probably be a lot more complicated to
handle both cases, and the transitions back and forth, than to just derive
the wheel rotation matrix for rendering purposes - especially considering
that half of that code would probably already be in place for the same of
drawing blurred wheel spokes...

Just thinking out loud. :-)

-- 

Nate Waddoups
Redmond WA USA
http://www.natew.com