[ODE] degrees of freedom in a slider - other joint types
STenyaK (Bruno Gonzalez)
stenyak at gmx.net
Thu Aug 31 13:35:48 MST 2006
On Thu, 31 Aug 2006 21:04:35 +0200, Deak Szabolcs <ancient at ludens.elte.hu>
> I'm trying to use ODE for vehicle dynamics and currently I see two
> methods of doing suspensions: the overly-simplified approach of using
> the specialized Hinge2 joint (which is actually one example of an
> "extra" joint I'm talking about) which is incorrect compared to a real
> suspension in both kinetics and kinematics or model the whole suspension
> with all the arms as separate bodies and connect them by the basic
> joints, which is theoretically correct but a computational overkill
> while most probably introducing severe numerical errors in the process
> of solving the system due to bad mass ratios of the bodies involved. As
> opposed to this, I believe we could use a so-called "distance
> constraint" (a 1d joint) to represent a "massless" link attaching two
> bodies with "ball" joints at its ends (essentially 1 control arm in a
> suspension). This would not be completely "correct" as there is no such
> thing as a massless joint, but the result would probably be much faster
> to compute while hardly any less accurate..
A properly modelled front race car suspension usually has 12 joints and 10
bodies. That's just one front suspension. Put 4 suspensions in each car,
put 15 cars in the race.
That would a total of 600 joints and 495 bodies, if i'm not mistaken.
That's a huge bottleneck in the cpu.
Using 1d joints, that would be reduced to 420 joints and 285 bodies. I'd
love to have that joint available in ODE, since my cpu is already
struggling with just one car on screen, and it doesn't even have the whole
suspensions modelled (just 5 joints and 5 bodies per suspension).
So yes, i'd be very interested in this new "distance constraint" joint.
(On a side note, some professional sim devs use IK in order to solve the
suspension system. It seems to be faster than ODE approach, and it always
gets solved in one physics step. I think the lack of masses is not a
problem when compared with the dangerous mass ratios and unaccuracy of
solving too many joints. I don't know if this (IK) would be feasible in
ODE or not though... )
Site: http://1ksurvivor.homeip.net <1kSurvivor>
Mail: stenyak AT gmail DOT net
More information about the ODE