Car Physics: stability/ lateral forces

i can't get my car sim to get stable oversteer situations. for example, if i steer in one direction, and then counter-steer at high velocitys, my car will always loose control, and do one or more 360 turns. i've been playing around with all my simulation velues for quite some time, but either i get the car to simply understeer in each curve, or i get very hard-to-handle oversteers. i guess that something is worng about my calculation of the lateral, or the way of combing these with the longitudinal forces. i basically calculate lateral forces as the dot-product of the wheels velocity, and the vector orthogonal to the wheel. i multiply this with a constant factor and mass carried by the wheel. then i sum my lateral force and the longitudinal force, and clamp the result to the maximal force the wheel can transfer to the road. somewhere on the list i also read about another way of combining these forces, where the longitudinal is clamped to the maximal force, and then the lateral force is applied as 'what is left in the traction budget' using the pythagorean theorem. i also tried this, and it would greatly raise the impact of the longitudinal force over the lateral force, letting me do much tighter donuts, and making the car react much more to the throttle in curves, and thus, unfortunatly, increasing my problem of unhandable oversteering. does anyone see something obviously wrong with my approach? or is there value i might not have considered yet? thanks for any advice.. jonas ps. is there a good reference (preferably online) for this sort of questions?

Hi Jonas, First of all you may want to look into Pacejka magic formulae for calculating the tyre forces. This seems to be the industry standard method for simulating vehicles. This also has the advantage that you don't have to cap the forces to the friction circle, since this is part of the model (if you used the combined version of the formula). The only online description I've found of this is Brian Beckman's series http://www.esbconsult.com.au/ogden/locost/phors/phors.htm I understand that G. Genta's book and Race Car Vehicle Dynamics by Milkenen and Milkenen also cover this. (I'm sure someone on this group can give you complete references for these). As for your current problem, this is one I've encountered myself and I'm still ironing out the problems. It seems to be caused by the relative forces applied in longitudinal and lateral directions on a tyre. If you apply too much longitudinal force, you'll have less grip to play around with at the rear causing oversteer (assuming rear wheel drive). Try playing around with the force you're applying in this direction and see how that affects things. Good luck, Mike "jonas echterhoff" <[email hidden]> wrote in message news:100820011805527551%[email hidden]...

Quoted text (35 lines)

> i can't get my car sim to get stable oversteer situations. for example, > if i steer in one direction, and then counter-steer at high velocitys, > my car will always loose control, and do one or more 360 turns. i've > been playing around with all my simulation velues for quite some time, > but either i get the car to simply understeer in each curve, or i get > very hard-to-handle oversteers. > > i guess that something is worng about my calculation of the lateral, or > the way of combing these with the longitudinal forces. i basically > calculate lateral forces as the dot-product of the wheels velocity, and > the vector orthogonal to the wheel. i multiply this with a constant > factor and mass carried by the wheel. > > then i sum my lateral force and the longitudinal force, and clamp the > result to the maximal force the wheel can transfer to the road. > > somewhere on the list i also read about another way of combining these > forces, where the longitudinal is clamped to the maximal force, and > then the lateral force is applied as 'what is left in the traction > budget' using the pythagorean theorem. i also tried this, and it would > greatly raise the impact of the longitudinal force over the lateral > force, letting me do much tighter donuts, and making the car react much > more to the throttle in curves, and thus, unfortunatly, increasing my > problem of unhandable oversteering. > > > does anyone see something obviously wrong with my approach? or is there > value i might not have considered yet? > > thanks for any advice.. > > jonas > > ps. is there a good reference (preferably online) for this sort of > questions?

In article <HIad7.18927$[email hidden]>, Mike Stanley <[email hidden]> wrote:

Quoted text (10 lines)

> Hi Jonas, > > First of all you may want to look into Pacejka magic formulae for > calculating the tyre forces. This seems to be the industry standard method > for simulating vehicles. This also has the advantage that you don't have to > cap the forces to the friction circle, since this is part of the model (if > you used the combined version of the formula). The only online description > I've found of this is Brian Beckman's series > http://www.esbconsult.com.au/ogden/locost/phors/phors.htm >

thanks for the link (i knew the PHOR series before, but all the web archives i found before didn't have the last chapters.)

Quoted text (12 lines)

> I understand that G. Genta's book and Race Car Vehicle Dynamics by Milkenen > and Milkenen also cover this. (I'm sure someone on this group can give you > complete references for these). > > As for your current problem, this is one I've encountered myself and I'm > still ironing out the problems. It seems to be caused by the relative forces > applied in longitudinal and lateral directions on a tyre. If you apply too > much longitudinal force, you'll have less grip to play around with at the > rear causing oversteer (assuming rear wheel drive). Try > playing around with the force you're applying in this direction and see how > that affects things. >

no, this is not the problem in my case (i have implemented a kind of traction control to avoid this) - it also happens with FWD cars or when i'm giving zero throttle. jonas

On Sat, 11 Aug 2001 19:40:55 +0200, jonas echterhoff <[email hidden]> wrote:

Quoted text (9 lines)

>In article <HIad7.18927$[email hidden]>, Mike >Stanley <[email hidden]> wrote: > >> you used the combined version of the formula). The only online description >> I've found of this is Brian Beckman's series >> http://www.esbconsult.com.au/ogden/locost/phors/phors.htm >> >thanks for the link (i knew the PHOR series before, but all the web >archives i found before didn't have the last chapters.)

Note that that one uses just the separate slip functions, not combined slip. Brian does an attempt at combined slip (which you can feel in my sim btw, amongst other methods). The Pacejka formula was expanded in '97 with weight functions, called G() which combines the Fx and Fy forces using cos()-like hill-like curves. (I don't think it changes the input, if you're reading this, Gregor). It adds some constants to define the influence, and lamba-factors to quickly test out variations (sounds familiar to Gregor's Pacejka factors hehe). And no, I don't have a URL for that info. Brian's method of combined slip seems to do very well. I'll have to read, read and reread the '97 methods to understand it a bit. :) Even more cos() and such, but all based on the same Pacejka base formula. 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, Brian's method is the simplest method that satisfies some basic requirements (linearity at small slips, max force reached for both components at combined slip simultaneously, upper and lower bounds for maximum slip within those for independent ones etc.), but that's just one way of doing it, and real life data may vary from this a bit. But if actual combined slip data is unknown, Brian's method is as good as any, and probably the simplest possible one at that. Genta's method probably takes more combined slip info into account, and if you have that info it's probably a good idea to use it. If you don't however, there's no reason to switch to it. Cheers, -Gregor Ruud van Gaal wrote:

Quoted text (18 lines)

> > Note that that one uses just the separate slip functions, not combined > slip. Brian does an attempt at combined slip (which you can feel in my > sim btw, amongst other methods). > The Pacejka formula was expanded in '97 with weight functions, called > G() which combines the Fx and Fy forces using cos()-like hill-like > curves. (I don't think it changes the input, if you're reading this, > Gregor). It adds some constants to define the influence, and > lamba-factors to quickly test out variations (sounds familiar to > Gregor's Pacejka factors hehe). > And no, I don't have a URL for that info. Brian's method of combined > slip seems to do very well. I'll have to read, read and reread the '97 > methods to understand it a bit. :) Even more cos() and such, but all > based on the same Pacejka base formula. > > Ruud van Gaal, GPL Rank +53.25 > Pencil art : http://www.marketgraph.nl/gallery/ > Free car sim : http://www.marketgraph.nl/gallery/racer/

Eh, that should be "... upper and lower bounds for maximum grip... " -Gregor Gregor Veble wrote: > upper and lower bounds for maximum slip within those for independent ones

On Mon, 13 Aug 2001 14:16:13 +0200, Gregor Veble <[email hidden]> wrote:

Quoted text (9 lines)

>Hi Ruud, > >Brian's method is the simplest method that satisfies some basic >requirements (linearity at small slips, max force reached for both >components at combined slip simultaneously, upper and lower bounds for >maximum slip within those for independent ones etc.), but that's just >one way of doing it, and real life data may vary from this a bit. But if >actual combined slip data is unknown, Brian's method is as >good as any, and probably the simplest possible one at that.

I think so, the Pacejka '97 paper I have here uses some extra Pacejka-like formulae (cos(...)) as weight functions. I'll hang on to your and Brian's method for the time being. Too busy on other nice things. ;-) >Genta's method probably takes more combined slip info into account, and >if you have that info it's probably a good idea to use it. If you don't >however, there's no reason to switch to it. It's not from the Genta book, it's from Pacejka and eh, Bisselink or something. Added some r1-r6 constants. I guess the list of parameters is endless. :) 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, as a late professor of ours used to say, with enough fitting parameters you can even describe an elephant :). Cheers, -Gregor Ruud van Gaal wrote:

Quoted text (8 lines)

> > It's not from the Genta book, it's from Pacejka and eh, Bisselink or > something. Added some r1-r6 constants. I guess the list of parameters > is endless. :) > > Ruud van Gaal, GPL Rank +53.25 > Pencil art : http://www.marketgraph.nl/gallery/ > Free car sim : http://www.marketgraph.nl/gallery/racer/

Ruud van Gaal wrote:

Quoted text (8 lines)

> >Genta's method probably takes more combined slip info into account, and > >if you have that info it's probably a good idea to use it. If you don't > >however, there's no reason to switch to it. > > It's not from the Genta book, it's from Pacejka and eh, Bisselink or > something. Added some r1-r6 constants. I guess the list of parameters > is endless. :) >

Oops, had a mental slip there. Of course, I was refering to the Pacejka extension as you mentioned it. -Gregor

On Mon, 13 Aug 2001 17:11:07 +0200, Gregor Veble <[email hidden]> wrote: >Hi Ruud, > >as a late professor of ours used to say, with enough fitting parameters >you can even describe an elephant :). LOL! Funny how little kids can tell an elephant is an elephant, without fitting parameters or anything. Shouldn't we just create 'class Brain' and do away with all that math. :) Ruud van Gaal, GPL Rank +53.25 Pencil art : http://www.marketgraph.nl/gallery/ Free car sim : http://www.marketgraph.nl/gallery/racer/

There may have been some good responses already but I've not seen any (my news server is not getting all the posts)... Anyway, if an oversteer car is driven above the "critical speed", then it is "fixed-control un-stable". In this case, there is no such thing as "stable oversteer"! In other words, it will take some control action (wiggling of the steering wheel) to get the car to follow a set path. The more oversteer is it, the quicker will be the divergence(spinout) -- so you may have just chosen a setup that is too extreme for you to control with the limited driver feedback of a sim (compared to the seat-of-the-pants feel in a real car). <shameless plug> This is covered in great detail in Chap 5 of "Race Car Vehicle Dynamics", order directly from the bookstore at <http://www.sae.org>. </shameless plug> --Doug Milliken Milliken Research Associates Inc. On Fri, 10 Aug 2001, jonas echterhoff wrote:

Quoted text (37 lines)

> i can't get my car sim to get stable oversteer situations. for example, > if i steer in one direction, and then counter-steer at high velocitys, > my car will always loose control, and do one or more 360 turns. i've > been playing around with all my simulation velues for quite some time, > but either i get the car to simply understeer in each curve, or i get > very hard-to-handle oversteers. > > i guess that something is worng about my calculation of the lateral, or > the way of combing these with the longitudinal forces. i basically > calculate lateral forces as the dot-product of the wheels velocity, and > the vector orthogonal to the wheel. i multiply this with a constant > factor and mass carried by the wheel. > > then i sum my lateral force and the longitudinal force, and clamp the > result to the maximal force the wheel can transfer to the road. > > somewhere on the list i also read about another way of combining these > forces, where the longitudinal is clamped to the maximal force, and > then the lateral force is applied as 'what is left in the traction > budget' using the pythagorean theorem. i also tried this, and it would > greatly raise the impact of the longitudinal force over the lateral > force, letting me do much tighter donuts, and making the car react much > more to the throttle in curves, and thus, unfortunatly, increasing my > problem of unhandable oversteering. > > > does anyone see something obviously wrong with my approach? or is there > value i might not have considered yet? > > thanks for any advice.. > > jonas > > ps. is there a good reference (preferably online) for this sort of > questions? > >