Forums
New sim code - Printable Version

+- Forums (https://www.vdrift.net/Forum)
+-- Forum: Project (https://www.vdrift.net/Forum/forumdisplay.php?fid=4)
+--- Forum: Development (https://www.vdrift.net/Forum/forumdisplay.php?fid=9)
+--- Thread: New sim code (/showthread.php?tid=1589)

Pages: 1 2 3 4 5


New sim code - NaN - 01-17-2012

I've pushed new simulation code into logzero/vdrift and broke a bunch of things. Consider this fork and the accompanying vdrift-data-dev as experimental till the regressions are fixed.

The new code models all components(engine, clutch, diff, brake, tire, suspension) as constraints and tries to solve them(iteratively a la bullet).

Features:
- More accurate simulation.
- Configurable driveline (2-n wheels, 0-m differentials).

Regressions:
- Simple hinge suspension geometry(camber effects disabled, anti-roll not implemented).
- Due to car config, suspension changes some cars might be broken.
- TCS/ABS not-implemented/broken.
- Tire squeal not-implemented/broken.
- Replay not-implemented/broken.
- AI most probably broken, not tested.
- And maybe some more...

[Image: 94561566.jpg]


- NaN - 02-13-2012

I've ported the driveline code into vdrift master(suspension and tire code unchanged). Not committed yet.

Here is a simulation check pic:
[Image: simz.jpg]

In detail explanation:

The loads are the power in kW applied by the component. We have 576kW engine output(full throttle) and -553kW load from clutch, thus 23kW are accelerating the crankshaft(inertia 0.1kgm2).

The differential applies a load of -543kW on the driveshaft, 553 - 543kW = 10kW are accelerating the driveshaft(inertia = 0.25kgm2).

We have 2 * 271kW applied to the rear wheels. The rear tires are generating 2 * 260kW grip. 11kW are accelerating the rear wheel(inertia about 5kgm2), which corresponds to the 10kW front wheel acceleration load.

2 * 260kW - 2 * 10kW front wheels = 500kW of the 576kW are accelerating the car. The interesting part about the code is that component efficiency(transmission, differential) can be added rather easily to have a more realistic model.


The problem is. The cars behave different, are harder to drive(with keyboard at least). You have to be cautious with the F1-02 in the first gear with TCS enabled. On the positive side the brakes behave much better(brake distance about 50-70%). I've been thinking about some simple stability control and changes to the traction control. One could also try to tune F1-02 parameters(tires, wheels) to get the old behavior.

All in all it means the new code will affect car handling. Any suggestions how to proceed, it is a game and not a (hardcore)simulation after all?


- NaN - 02-13-2012

Hmm, I had to change the autoclutch code due to the new engine, clutch and transmission model. It is less slippy now which could explain the F1-02 first gear grip problems, will investigate it.


- NaN - 08-06-2012

I've merged the fork into sim branch with the target ship it with the next release.

For the brave who want to give it a try. Use branches/vdrift-data-dev. A bunch of things are broken atm, will be fixed:

- You are driving on the track mesh(not on the road patches), it might feel a bit rough
- Recently added cars are not converted yet, will fail to load.
- Anti-roll bars, ABS, TCS are missing.
- A lot of cars need adjustments to suspension, center of mass, brakes (after anti-roll bars are back).
- Replays won't work.
- AI will most probably not work(road patches).
- And some more minor stuff like clutch, shift logic need tuning.

Anyway, a few cars are doing quite well already 360, XS, LE(needs anti-roll bars and maybe lower center of mass though).

I rewrote tire squeal estimation to use tire thermal load(from friction) and ideal slip fraction. This allows to drive decently without ABS using keyboard. So listen to your tires. :wink:


- NaN - 08-08-2012

Road patches and ABS/TCS are back. And I've got anti-roll working with the new suspension model, will push it soon.

What remains are a few bits in AI and clutch/transmission logic, tire sound pitch has to be fixed(easy one). Replay can be fixed later.

Then comes the heavy stuff. Updating/testing the 47 cars we've got. Not sure how to approach this. Focus on a few selected ones first?


- NaN - 08-08-2012

Interesting observation is, that the 360 and LE have been setup from the real world info available online. I've used Lotus Elise manual to setup the suspension and engine of the LE. And they are doing quite well already. The LE might need a bit lower center of mass, but we have to guess here. Running with an anti-roll bar fixes any roll over tendencies already.

It would be awesome to get all cars in line with their real world specs. This would be a huge undertaking though.


- berzingue - 08-09-2012

Sorry, can you publish a link to download the complete last version of Vdrift ? stable version ?? thanks a lot !


- NaN - 08-10-2012

I've uploaded a fresh build (vdrift-2012-08-10.exe) if you are using svn(tortoisesvn) to get latest data:
https://sourceforge.net/projects/vdrift/files/vdrift/nightly%20builds/

If you are running the latest release, link to complete dataset(> 3GB):
http://vdrift.svn.sourceforge.net/viewvc/vdrift/tags/vdrift-data-2012-07-22/?view=tar

I'd prefer svn, much easier to keep up to date.


- berzingue - 08-10-2012

Sorry to come with more problems again...I'd tied to update with SVN, here are messages errors :

[Image: 263814bug.jpg]

[Image: 287269bug.jpg]

[Image: 486508bug.jpg]

Also, I'd tried to create an account on sourceforge (3 times, 3 differents accounts), but impossible to receive passwords to connect. And impossible to commit...

Any idea ?


- NaN - 08-10-2012

OK, https://sourceforge.net/projects/vdrift/files/vdrift/nightly%20builds/ contains the latest builds(packed with 7zip). Open the link in a browser and download vdrift-2012-08-10.exe. Extract it.

Create a data directory inside the extracted vdrift folder. Then checkout data from https://vdrift.svn.sourceforge.net/svnroot/vdrift/vdrift-data


- NaN - 08-10-2012

Quote:Also, I'd tried to create an account on sourceforge (3 times, 3 differents accounts), but impossible to receive passwords to connect. And impossible to commit...
What is your sourceforge user name? You should get a confirmation email from sourceforge, with a link you have to follow to confirm your registration. Make sure it is not blocked by your e-mail spam filter(check your spam bin/folder?)


- berzingue - 08-10-2012

Ok thanks !

sorry for longtime without contribution...i'm actually working on a Jetfighter simulator project (http://www.digital-fighters.net/site/), but I'd like to modelise cars and tracks again...I have to remember how all works here...

so !

- account on sourceforge is ok (you was right : anti-spam. Damn !).
- Vdrift 2012-07-22 download is ok, the game is ok. I have to say you made a really good job with shaders, opponents, menu...wow !)
- Problem : I've overwritted 2012-07-22 files with 2012-08-10 files : Vdrift can't launch...
- Download datas with SVN works good !


- joevenzon - 08-10-2012

NaN Wrote:Interesting observation is, that the 360 and LE have been setup from the real world info available online. I've used Lotus Elise manual to setup the suspension and engine of the LE. And they are doing quite well already.

That's awesome! Great achievement. If real-world specs are achieving pretty good results, that'll make it a lot easier to do the tuning than before, where one had to fiddle with a dozen variables by feel.


- berzingue - 08-10-2012

Ooops, Wackersdorf doesn't load, can't drive on it !


- NaN - 08-10-2012

Look here: http://vdrift.net/Forum/viewtopic.php?t=1778