[ODE] Re: Bendy wheel problem

[Jonathan Langridge]

I'll add my name to the list... is there a membership certificate? "I've 
experienced the freaky rubber band wheel feature" or something? :)

I'm actually getting the "feature" in a motorcycle sim during high-speed 
turns. It's very bizarre, and sometimes the meant-to-be-fixed rear wheel 
remains bent for several seconds after coming out of a turn or even turning 
the other way. Until someone finds a solution I'll just cap speed.

On another subject, in order to simulate the wheels I'm using ellipsoids, 
and have written my own code to clip them. When driving across an edge 
between 2 polys with a very small change in angle of slope betwen them, two 
clip points can be generated very close together. Sometimes this wreaks 
havoc with the physics, as I think Russ mentionned might happen in this 
case. At present I'm tempted to just check for this case and combine close 
contact points into a single average one, but I'm wondering if anyone has 
found a more physically accurate alternative? I don't really know how to 
access CVS so I haven't seen the TriCollider code - how does that code 
manage it? (the same problem must exist for sphere/tri)

AYBABTU,
Jon

>From: "McEvoy, Nick" <nick.mcevoy@dsto.defence.gov.au>
>To: "'ode@q12.org'" <ode@q12.org>
>Subject: [ODE] RE: Car wheels
>Date: Thu, 5 Dec 2002 09:38:12 +1030
>
>Robert Milesz wrote:
><snip>
> >The rear wheels tend to rotate, which is unexpected, as I set the
> >HiStop and LoStop parameters to 0.0 (all four wheels have hinge2
> >joints). This effect is visible especially when the car is sliding
> >in a turn. In this case, the rear wheels are turning towards the
> >direction of the sliding movement, which is what supposed to happen,
> >but only if there wasn't a 'lock' on the Hi and LoStop parameters.
> >I've tried quite a lot of things to fix this.
><snip>
>
>There is a growing list of people who have this problem ... I call it the 
>'bendy wheel problem' ... or the 'freaky rubber band wheel feature' ... and 
>I have *not* found any 'good' solution to the problem.
>
>I've tried all the tweaks I can think of to get the wheels to stay 
>straight:
>- reduce world step size
>- call world step more often per frame
>- increase diameter of wheel & decrease mass of wheel
>- fiddled with CFM, etc.
>- call appropriate ode methods to that should 'fix' this issue (you know 
>the ones dBodySetFiniteRotationAxis, etc.)
>
>All of these reduced the problem ... but as my vehicle reaches top speed it 
>can get nasty !
>
>Suggestions like visually drawing the wheels correctly aligned do not help 
>because the handling is affected by the wheels moving out of alignment 
>(thus the car 'crabs' sideways and ultimately spins out of control).
>
>In a previous email (http://q12.org/pipermail/ode/2002-October/002006.html) 
>Russ talked about the problem.  I wish I had some clues on how to fix it 
>myself ... but I don't look under the hood of the ODE engine much ... or if 
>I did I'd probably just scratch my head and say 'erm ... looks like you've 
>got a wheel alignment problem there ... that's gonna take some fix'n' :)
>
>I hope this issue can be sorted sometime soon ... (Russ?)
>
>Nick
>_______________________________________________
>ODE mailing list
>ODE@q12.org
>http://q12.org/mailman/listinfo/ode


_________________________________________________________________
MSN 8 with e-mail virus protection service: 2 months FREE* 
http://join.msn.com/?page=features/virus