rec.autos.simulators

Hi there,

I'm trying to deal with sliding in my sim at this moment.
For skid sounds, it seems looking at slip angles and slip ratios
getting large are a nice way to go.
For sliding, I'm not too sure. I use Pacejka to get long/lat forces
based on slipRatio/Angle/normal force and camber (camber is still 0 at
the moment).
So I get an Fx (longitudinal) and Fy. For sliding, I should check if
the resultant force (Fx+Fy) doesn't exceed the maximum force of the
tire. ( |Fx+Fy|<=maxForce )
However, how do I extract a friction coefficient (or similar, the
maximum total tire force) for a given situation? Could the Pacejka
constants help here? I think so, since the peak of the Fx force for a
given set of Pacejka input parameters will be the maximum force that
the tire can generate. However, this means I would have to search for
the optimal slipRatio (which delivers the highest Fx), which delivers
the highest output.
Or perhaps, it's just D+Sv, because the formula is like D*sin(...)+Sv
(or was it Sh, haven't got the book at hand right now).

Second, an effect that I'm still missing is the car switching ends
when flooring the throttle from a standstill. Given a large wheel spin
velocity, at the rear, the car won't move forward quickly, because of
the reduced Fx force. However, since slipAngle is calculated as the
ratio of longitudinal and lateral wheel speed (wrt the road), the
slipAngle builds up as the car wants to drift left or right, I just
get a Pacejka Fy which stabilizes the car in yaw.
So, where do I go wrong? It seems that because of the large spinning
velocity the lateral friction coefficient becomes almost 0 (like with
a motor where you can let the rear wheel drift along the track,
keeping the front wheel locked on the tarmac).

Thanks for any ideas,

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.racesimcentral.net/
Free car sim  : http://www.racesimcentral.net/

Hi Ruud,

I do skid sounds by checking for the size of the slip vector (ratio and
angle combined), and slightly below the optimal value I turn the volume
up to full when past the optimum, and then also change the pitch when
the value of the slip vector increases.

About Pacejka, I have absolutely no experience with that, but it seems
to be you are using the lateral and longitudinal forces separately based
on slip angle and ratio. For illustration, as an approximation (which I
still use :) ) you can combine them in a slip vector, calculate the
force that corespods to the magnitude of this slip vector, and point it
in the opposite direction to it.

The (complete, if such a thing exists) Pacejka formula should do
something similar and enable you to plug in both the ratio and angle
values simultaneously and getting both the lateral and longitudinal
forces from that. The calculations should not be done independently for
both axes. You really don't need to perform any special sliding checks
if this is done correctly. This will also take care of the twitching
problem, by the way.

-Gregor

On Fri, 06 Apr 2001 14:36:32 +0200, Gregor Veble

I wondered about that, if pitch increases. It seems though you only do
it slightly, couldn't hear much variation from a testrun with your
SimEngine. I now use the slipvector for lateral skid sounds, and the
slip ratio to detected skidding in the long. direction. Somehow I
think the two sounds sound different (because of other spring rates
long vs lat), but it will do for now.

I used the slipvector in v0.1 (not to be funny, lol), and it worked
great for lowspeed; no low-speed instability whatsoever. Perhaps it's
something useful after all when doing lowspeed. :) Got the idea from
RARS, which targets more to robot simulation actually.
I can't really point it in the opposite direction, because I take the
resultant force vector as the sum of the lateral and longitudinal
force. I will try capping it (am now only reducing the lateral force
if the longitudinal force gets large, so this changes the direction of
the force, which doesn't feel right intuitively).
Pacejka indeed inputs: SR, SA, camber and Fz. From that you get Fx,
and Fy (and Mz) from 3 separate formulae. I don't have the book from
Genta here, so I can't check whether the lateral force looks at the
slip ratio. In my carsim, it doesn't seem to do so though. I'll check
it out when I get home. (first have to get v0.4.1 going tonight)

You would think so, yes, but it doesn't seem to. Likewise, for wheel
locking, I have to interpolate the total force vector to move to
-slipVector as the locking becomes greater (SR nearing -1 or worse).
The twitching is almost gone by the way, and thanks for your replies!
I got it working in the end because of a stupid bug where I used the
suspension force on the wheel as a normal force, instead of taking the
REAL road force up (which is tireRate*(ySurface-yContactPath) in my
sim). Having fixed that, suddenly it all started working, although
with a damping coefficient of ~30, really magical values here. ;-)

I suspect Pacejka doesn't handle fast-spinning wheels correctly for
calculating Fy (lateral force). The tire will go stiff and the lateral
spring effect would reduce greatly, if I picture it. Perhaps that
should be reflected in the Pacejka constants (where at a certain point
you have BCD which is the cornerning stiffness; I use that in the
SAE950311 damping method, just like BCD for longitudinal stiffness).

I'll go read some more RCVD 'Simultaneous longitudinal and lateral
forces' to see what it says.
v0.4.1 btw displays lateral and longitudinal forces (and SA/SR for
each wheel) so it might be interesting to watch. I'll tweak my force
capping algorithm just a little more now.

Thanks,

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Free car sim  : http://www.marketgraph.nl/gallery/racer/

Ruud,

What you ideally need is the 'combined slip' Pacejka formula.  This deals
with the interactions between long and lat slip.  It is based on the
seperate formulae for 'pure' lat and 'pure' long slip, but then applies a
new additional calculation to give the 'combined slip' Fx and Fy.  This is
what I use in my simulation work within my PhD.  I have the documentation if
you're desperate.

Cheers,

Neil

Neil,

I'm also interested in the documentation you're talking about. Most of
all I'm
interested in Pacejka Coefficients for different kind of tires if you
have some.
I hope you can help me.

Best regards,

Ernesto Fina

On Sat, 7 Apr 2001 08:17:01 +0100, "Neil Everett"

Believe it, I'm desperate! :)
But really, I've bought an entire book for the Pacejka formula (G.
Genta's Modeling and Simulation), and it only contains the separate
formulae and an approach to combine them, but I'm having trouble with
the lateral force being allowed to keep big when the rears are
spinning.
Example (going the Genta route):
- I calculate Fx and Fy separately
- I use the elliptical approach to hold down Fy, this is done by:
  Fy=Fy0*sqrt(1-(Fx/Fx0)^2)
  Where:
  - Fy is the lateral force we want to know
  - Fy0 is the lateral force as calculated by Pacejka Fy=...
  - Fx is the calculated long. force (Pacejka separate)
  - Fx0 is the MAXIMUM long force (calculated from the Pacejka
separate formula as D+Sv)
- I stamp on the throttle, slipRatio goes up to say 2 or 3
- Fx at the rear becomes 1400, from a maximum (at optimal slip) of
2700 (at say, slipRatio=0.3).
- Fy is about 2000 (Pacejka Fy)
- Fy is reduced to not exceed a total force length of 2700. As Fx is
about half of it's maximum peak, the lateral force is allowed to grow
a lot.

Here's what I don't believe; when spinning, the method above keeps the
friction circle (or ellipse height/width, but let's stick to circles
for now) constant (because it takes Fx0 as the maximum force the tire
could ever generate). However, I known from doing donuts with GPL for
example, that when spinning the wheels, the friction coefficient goes
down considerably. So a car with spinning rear wheels and braked
fronts can be pushed even by someone and it will start rotating.
This is nowhere reflected in the calculations.

So the problem then becomes, how to calculate the maximum friction
coefficient, based not only on load and a generic track coefficient
(1.7 for example), but also on slip ratio? Or heat, although I don't
think that heat is the thing here that makes things slippery (at the
start).

Thanks for the help, Neil, I'd love to take a look at your simulation
work, and how you take on combined Pacejka.

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Free car sim  : http://www.marketgraph.nl/gallery/racer/

It seems these coefficients are often copyrighted or something. Not
really good for what's called a standard. :)
But well, I'm have low-priority plans on a Pacejka editor where you
can tweak the constant and see what it does (and hopefully mimic known
curves).

Ruud van Gaal, GPL Rank +53.25
Pencil art    : http://www.marketgraph.nl/gallery/
Free car sim  : http://www.marketgraph.nl/gallery/racer/

The lateral capability of the tire is reduced considerably as the tire
spins,
as is discussed in this thread.  One source of a lateral dispersion
for some suspensions is asymetry in the rear tire normal forces
caused by the driving torque. The engine torque is converted to
driving torque (turned 90 degrees) at the wheels, and this
torque on the driveshaft is reacted laterally partly in the front
and part in the rear suspension.

To summarize: One rear tire sees a higher normal force than the other.
For a locked differential, with the tires both rotating at the same slip
ratio, it's easy to see that the tire with the larger normal force will
have a larger forward force and this difference in forward force
between the two rear tires creates a yaw torque. Combined with the
reduced lateral force capability of the spinning tires (that can make the
car unstable), this yaw torque can start the car around.

- Matt

Check out Millikin & Millikin's Race Car Vehicle Dynamics, chapter on tire
data non-dimensionalization... this lists formulas for combining slip angle,
slip ratio, and camber into one unified slip variable, giving you all those
"friction circle" behaviors you're wanting...  they've also go a number of
datasets for various types of tires scattered throughout the book.  Make
sure you get the load sensitivity curves for your friction and
cornering/camber stiffness, otherwise your roll stiffness adjustments won't
produce the correct understeer/oversteer effects.  (The formulas make "mu"
and "C" look like constants, when in fact they're functions of normal
force/load--this lead me astray the first time I tried to use them!)
If you've got the rotational inertia of the wheels/drivetrain modeled, once
you plug these guys in, you'll be able to replicate just about any vehicle
behavior you can think of (donuts, brake-lockup understeer, trailing
throttle oversteer, etc)
Once you get to this point, you can get pretty far in modeling different
types of tires just by tweaking the "C" and "mu" curves, without (at first)
having to mess with the "magic numbers".

You go away for a few days and miss a whole physics thread on R.A.S. :)

Ash

Hi Doug and all

Have you seen TyreGene from Yearstretch.
http://www.yearstretch.com/yearstretch/Shop/product.asp?intProdID=1

Chris

Chris,

Thanks for the link.  It is kind of surprising that there is no
reference for Pacejka-96 (paper title and publisher).

Are there plans to complete the job?  To be useful to us, it needs aligning
torque, longitudinal force (and Fy-Fx roll-off), overturning moment, etc.

-- Doug

                Milliken Research Associates Inc.

Hi Doug

Iam not envolved with Yearstretch :O) Its something Ive come across in my
searches of the internet. I have my own little proggy for my pacejka needs
:O)

Chris

Doug,

ARe you helping Ruud with the Racer Project?  Just asking since I know the
level of respect your name garners in the circles I run around in and well
I am wondering if I need to start downloading the Racer game now.

Dave