Tire model woes

[NaN]

My recent experimentation with stability control has exposed an issue with current tire model. The used force combining method (beckman) seems to be wrong, not sure whether it is the implementation or the method itself.

I've noticed the behavior before to some degree(with F1-02 and karts) but never looked into it. With the stability control where wheel brakes are engaged independently the problem is quite noticeable.

Beckman combining as we have it favors heavily the longitudinal friction, to the degree that the total force is acting against the expected direction. I think we will have to fall back to no combining, or go with one of the simpler combining methods.

beckman: braking with steering to the left, longitudinal braking force rotates the car to the right against the expected direction


no combining

[NaN]

We need to pick a combining method. No combining means no lateral grip control using throttle/brakes (lat force only depends on wheel slip angle) and too much grip in general.

[NaN]

Thanks to internet gods I've stumbled across a full set of Hankook tire coefficients for "Magic Formula Delft Tyre
96" including the combined slip coefficients. Unfortunately the paper doesn't include the magic formulas.
http://www.biblioteca.uma.es/bbldoc/tesi...602675.pdf

The 96 MF-Tyre has 4 coefficients (rbx1, rbx2, rcx1, rhx1) for longitudinal combined slip factor Gx. The versions I've found so far use 7 coefficients. If anyone has an idea how the formula looks like post here or drop a link.

[NaN]

Good news. I think I've got the correct formulas, validating them against the plots in the paper atm. If nothing weird comes up, we will be driving Hankook tires with proper force combining very soon.

I'll have to write a new pacejka editor though. I've got a simple python/tkinter script to plot the curves already, need to add parameter sliders to support real time curve editing.

[NaN]

It works.

XS with Hankook tires, forced oversteer:


I've been wrong about the steering with braking(full wheel lock). The car will first slightly move into steering direction but will then(at lower speed) rotate against it. Beckman combining is correct here.

[joevenzon]

oooo

is that checked in? I've gotta give that a drive.

[NaN]

oooo

is that checked in? I've gotta give that a drive.

Not yet. I've been short on free time recently.

The next release will be delayed too I fear, hoped to have it out this month...

[NewLife]

Good news. I think I've got the correct formulas, validating them against the plots in the paper atm. If nothing weird comes up, we will be driving Hankook tires with proper force combining very soon.

I'll have to write a new pacejka editor though. I've got a simple python/tkinter script to plot the curves already, need to add parameter sliders to support real time curve editing.

very cool NaN! can't wait to test the new tire model!
where did you find the new formulas?

PS: UnityCar 2.0 Pro has been released, you can check it at http://unitypackages.net/unitycar/

[NaN]

Finally forced myself to sit down and implement the python(tkinter) tire editor, so that we can switch to MF-Tire 4.0.

This is the old tire model editor variant(the 4.0 will follow very soon):



I know ugly as hell, but very portable. The editor allows to load a reference tire config(lighter curves in pic) to make editing more intuitive.

[NaN]

And here the new version, now with 70% more parameters, lol (Hankook tire data):


The difference to the vdrift touring tire (light colored curves in the prev post) is quite visible. Please notice the different values for braking and accelerating (Fx, red), which is confirmed in the literature. Also the lateral force (Fy, blue) is lower than the longitudinal (Fx, red).

[ghiboz]

any news about this tire model?

[NaN]

The tire code is pretty much ready. I've derived 3 types from the hankook data (touring, sports(~20% more grip), racing(~40% more grip)).

touring vs sports:


Just fighting some asymmetric behavior of the tires atm. All cars are pulling slightly to the left, due to load dependent lateral force curve shift, if I am reading it correctly.

I haven't yet decided how to deal with this. I guess the real tire will have a "THIS SIDE OUT" marker. In the sim we have the same tire on all four wheels. Will try it with shift coefficients set to zero.


Oh, and the F1-02 is broken of course. That thing is driving me insane, has to be fine-tuned to any change in the simulation...