Additional vehicles

[fch-aa]

Hello, I'm interested in doing some simulations for GT cars on some local tracks.
I managed to create the track, but I'm having a hard time figuring out what each parameter means in the vehicle definition file (and where to find them).

Do you have any pointer on how to create one for say a Porsche 911 C2S (997.1 / 991.1)?
The one I created is roughly 30secs (approx. two minutes lap time) too quick compared to IRL.

Thanks for this amazing project!

EDIT: Managed to lower the gap to ~10secs.
I think the most "wrong" parameters for my use case are the pacejka model values (c/p from the F1 model).

porsche-997.1-c2s.txt

[juanmanzanero]

Hello! Your project sounds very cool. You can probably dig on internet to find proper values for GT cars, I have seen some projects in the past while creating Fastest-lap.

I have not created any documentation yet since the project is still under very heavy development, although there are some people using it already. You can look for the meaning of the parameters in the references I have linked in the welcome page.

Regarding your mismodelling, I would be cool to know if you are faster in the straights (too much power) or faster in the corners (too much lateral grip from the tires). Tweaking these parameters you can probably get a better match.

Question: what circuit are you running in?

Cheers!

[juanmanzanero]

BTW just saw in your XML file that you are using an engine power curve. This is not currently supported, only constant power (via maximum-power) works.

[fch-aa]

The track is called "La Ferté Gaucher" (LFG), it's a small track near Paris: https://www.circuitslfg.fr/
You will find attached to this message the KMZ / KML & corresponding XML file.

Understood for the power curve, I found those in the kart model XML file; this may explain the lap time difference!

There is definitively a difference in top speed, the simulation shows ~180 km/h at the end of the straight, but the real telemetry has a top speed of 168 km/h.

lfg.zip

[juanmanzanero]

Nice work! Feel free to drop a pull request with your files into the database/tracks folder. Surely you are overpowered, maybe reduce the engine maximum power, or increase the Cd

[juanmanzanero]

Hello @fch-aa , I am curious: have you arrived to anything interesting with the GT car model? Cheers

[fch-aa]

I tried to do just that, but I'm missing something. What you see on the Google Earth screenshot is actually my real lap.
Using the following code inspired by the unit tests I can convert correctly the track boundaries, but when I try to do the same with the simulated lap "x, y" values I get a distorted path.

I guess the Cartesian to long/lat conversion is not enough?

lon = sim_df['x'] / (R_earth * np.cos(phi0 * DEG)) * RAD + theta0
lat = sim_df['y'] / (R_earth) * RAD + phi0

The issue is most likely between the seat and the steering wheel, that's precisely what I want to see ;)

I used the notebook you created to compare F1 laps (examples/python/f1/optimal-laptime/Optimal_laptime.ipynb), in this code you did not process the distance like you mentioned above, did I miss something?

fig, ax = plt.subplots(figsize=(20,5))
ax.plot(ver_tel['Distance'], ver_tel['Speed'], color=rbr_color, label='VER')
ax.plot(ham_tel['Distance'], ham_tel['Speed'], color=mer_color, label='HAM')
ax.plot(np.array(s)-168.0,np.array(u)*3.6, label="Fastest-lap")

[juanmanzanero]

In the GPS plot, the casual 90º of error you get seems that lon/lat are exchanged or something like that. But the formulas you wrote look OK.

Exactly, the plot in my Notebook is quite relaxed and I did not took that into account. But this operations I wrote earlier represent the 'correct' traveled distance

[juanmanzanero]

Ah, one thing, also you have to flip the 'y', namely do y=-y. The circuit preprocessor has the Z normal pointing to the sky while the car simulator has the Z normal pointing to the road. I have to better document this, but the reasoning behind is that different references use these convention.

[fch-aa]

Thanks @juanmanzanero, that was it!

Blue = simulated
Gray = real

[juanmanzanero]

Well done! It looks super good