Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Engine friction model
08-05-2012, 01:13 PM,
Engine friction model
I've been working on the new simulation implementation again, trying to understand a very snappy off-throttle oversteer that rwd cars exhibit.

Current vdrift implementation also has this issue, although not as pronounced due to slippy clutch(at low rpm especially). The clutch in the new implementation does lock properly and exaggerates the effect.

It seems to be caused by engine friction(braking). The torque is strong enough to kill rear wheels grip when coasting along a straight line(clutch locked, throttle off).

Thus I think the used equation/approximation might be the problem. I've been wondering if there are some "plausible" simple engine friction models to look at? Even a few numbers would help just to get an idea, fix current equation.
08-05-2012, 01:56 PM,
As a hack/fix reducing the hardcoded friction factor -1300 to about -600 seems to help.
08-17-2012, 11:56 AM,
The issue is still there unfortunately. Can be observed with rwd cars(350Z, CO). Just brake the car down into first gear release the brakes and watch it doing roulette. TCS, ABS can suppress this behavior (help stabilizing the car) though.

I am not sure how to address it. One would need a clever auto clutch controller (a la TCS/ABS) to make sure that it doesn't overload the driven wheels.

How do real world semi-automatic/automatic transmissions deal with this? Or is our engine friction model just nuts?

Maybe I should test it with a real world car(rwd): Accelerate in the first gear and lift off the throttle abruptly just before reaching redline. Nah, just kidding. Don't try this at home.
08-17-2012, 12:56 PM,
As a sanity check I've increased simulation rate from 90Hz to 270Hz. The issue is still there, must be a model problem. It is easier to counter-react manually though. The higher rate gives in general more control, transitions are less abrupt, which was to be expected of course.
08-17-2012, 04:56 PM,
I've run the simulation with engine friction torque disabled(-1NM). This helped to get a more precise idea of the issue. It is caused by locked up rear wheels, or to be more precise the transition from locked to rolling wheels. Engine friction enhances the effect by increasing the load transition on the driven wheels.

Immediately after releasing the brakes we have rear wheels still at full slip and 0 RPM accelerated by the engine. One of the wheels will get more grip than the other at some moment (due to load, contact velocity, engine torque, rear left wheel for left hand drive cars for example) and initiate oversteer as the overall grip is still very low. This is the point where TCS helps by braking this wheel, almost acting like stability control. ABS helps to prevent to get there in the first place by avoiding wheel lock.

I think one could try to add some auto clutch traction control functionality to make cars more controllable without ABS/TCS, keep clutch open on overrun if the wheels are beyond the ideal slip value. The TCS/ABS is more effective here though by acting directly on the affected wheel.
08-17-2012, 06:55 PM,
It gets more interesting. Running with an open differential (anti-slip=0) reduces the effect considerably.

The new implementation is essentially an ideal 2-way differential. Changing the code to simulate a 1-way version (only lock on acceleration) reduces the effect as with an open diff.

I also have been experimenting with replacing current engine friction model by -0.2 * combustion_torque which seems to work a bit better, not causing any traction problems on throttle lift off(CO).
08-18-2012, 04:08 AM,
I've pushed the modifications to make rwd cars more stable on rear wheel lock to roll transition.

The only problem child is the F1-02. It will require an auto clutch traction control system. But then I am not sure if anyone is racing the F1 without ABS/TCS.
08-18-2012, 10:13 AM,
I've implemented a simple auto clutch traction control. It takes current wheel grip into account and allows to race F1-02 without ABS/TCS. Big Grin

PS: Just need to fix AI, Replay and the roll over tendency of some cars and we can merge the new sim code into master.
08-18-2012, 12:30 PM,
Awesome! Cheers for the running commentary; very interesting to read.
08-18-2012, 06:57 PM,
Thanks. I am used to have a logbook when working on a project and thought it might be of interest to post a few notes here for anyone following.
08-20-2012, 11:28 AM,
Wikipedia has some plots of a lift over oversteer event:

It just has accelerations so it's probably not super useful but it's interesting anyway. Smile
08-20-2012, 12:24 PM,
They describe it as a vertical load transfer. In our case with the old engine friction model the engine braking was the primary effect, to the extreme that the CO would completely lose rear grip and oversteer while driving on a straight(in the first gear). :lol:

Forum Jump:

Users browsing this thread: 1 Guest(s)