Hi,
Most of you may have heard of or read about the SAE950311 paper which
deals with low speed slip angles and ratios. The idea is to not
calculate the ratio/angle from the current velocities, but instead
make it into a differential equation which you integrate per step to
get the latest ratio/angle with the added benefit that you get
relaxation lengths for free.
I tried this method, and found I had big problems with the damping.
Needed coefficients that weren't near the 0.7 critical damping, but
required something like 35 for example, and even then it could go
horribly wrong at some points.
I then threw out the slip angle approach, as I couldn't get it right,
but left in the slip ratio diff equation, with a different damping
method, namely by scaling the slipRatio directly with 0.7 for example
when it changes sign. This works ok.
Yesterday I read in a thesis by Erik McKenzie Lowndes declaring the
above method to be used, explaining it a little, and casually
remarking that the damping is used as suggested (from SAE950311) and
that the damping force changes sign at each integration step.
This last thing, changing the damping force's sign each step, I
couldn't find this noted anywhere in the original SAE950311 document.
Still, when I think about it, this may either nullify the damping
(every 2nd step), or indeed work like expected.
Can anybody confirm, after implementing the SAE950311 method of
Bernard & Clover, that indeed this damping works if you reverse the
damping force each step? Or perhaps that it DOES need to have the same
sign after all.
Would be nice to be able to use it like that, since it uses lateral
force building because of the relaxation coefficient, something which
GPL supposedly didn't have. ;-)
Thanks, and I hope this all makes some sense to you all,
Ruud van Gaal, GPL Rank +53.25
Pencil art : http://www.marketgraph.nl/gallery/
Free car sim : http://www.marketgraph.nl/gallery/racer/
Hi Ruud,
I never thoroughly checked tha paper as I had the method in the sim
before seeing it. I can tell you how I do the damping, though.
Focusing on either of the two diretions (lat/long), we have two
variables at our disposal, the Slip and d Slip/dt, the later being
governed by the equation in the paper. When you feed the Slip into the
force equations (let's call the parameter that enters the force
equations Slip_enter), you may want to substitute it by
Slip_enter = Slip + Tau * (d Slip/dt)
where Tau is a certain time constant that should roughly correspond to
the oscillation period of the system in the given direction (lat is
different, much slower than long) to get close to critical damping. It
works for me, but I don't know whether it's this the same system that is
used in that paper or not.
-Gregor
Ruud van Gaal wrote:
>
> I tried this method, and found I had big problems with the damping.
> Needed coefficients that weren't near the 0.7 critical damping, but
> required something like 35 for example, and even then it could go
> horribly wrong at some points.
