oversteer measurement formula in Motec i2

[nado59]

Hi,

I had a look of the Motec i2 pro software. There is a maths function delivered with it, which calculates oversteer. According to my (very limited) knowledge, there is no low-cost way to get a reliable measure of oversteer/understeer. The formula in the software uses only the vehicle speed, LatG, Steered angle and the vehicle wheel base to come up with a figure. I am wondering if someone of you uses this and would be willing to comment on the measure vs driver feedback.

The complete formula is:

smooth(choose('Corr Speed'[km/h] < 50, 0, sgn('G Force Lat'[m/s/s]) * (('Vehicle Wheelbase'[m] * 'G Force Lat'[m/s/s] / sqr('Corr Speed'[m/s])) - sgn(stat_mean('Steered Angle'[rad] * 'G Force Lat'[m/s/s])) * 'Steered Angle'[rad])), 0.2)

Ignoring the filtering (smooth) and the condition to evaluate above 50 km/h (choose), this is just:

+/-(wheel base x LatG / speed^2) x (-/+ steer angle)

The +/- and -/+ represent change of sign (sgn in the function) in the first case in agreement with to the sign of the LatG. In the second case as the inverse of the stat-mean of the product of the LatG and Steered angle.

Does anyone have any comment about the deduction of this formula?

Thanks!

Nado

 

[cibachrome]

Looks like they are computing the steer required to negotiate a curvature (based on steady state speed, wheelbase and Ay m/sec2) minus the steer in the car at the position. Thus its an "attenuated steered" signal, not any sort of gradient which is the terminology used in "new school" engineered cars. The units are degrees (road wheel ? unless the ratio has been ass-exhumed in the code somewhere). Its a TooMuch/Insufficient steer calculation: Most cars would understeer so the sign for most of such a signal on a track would be negative. As long as you're a 'constant speed in the turns cornering driver' this could be useful. I.E. constant yawrate and speed while cornering.

Accelerate in a turn with some sideslip acceleration (the other major component of total lateral acceleration) and this gets bad/wrong very quickly.

I'd dispute your sorrow about a cheap way to compute understeer (gradient of rw_steer by Ayg minus Ackermann gradient). With just yaw velocity (r) and speed (u), you can compute understeer from r/u vs. u*r if you have enough room to run a speed sweep at several fixed steer angles (curvatures). The slope is 'understeer' (also need wheelbase). If you can sneak into a shopping mall parking lot, a constant radius test (100m is ISO4138) gets pretty simple. Use a protractor for Steer angle, your speedometer for speed, estimate your steer overall ratio from a protractor and a lock to lock test. Normalize steer angle to get road wheel angle. Fit data at speeds 10 up to max speed in 10 kph increments, use nice function, derivative is understeer. No Ackermann gradient to mess with in a const. radius test. If you have a sideslip transducer, it will read zero at some magic speed. This speed has as much value as understeer when used in computing the rear axle sideslip gain of the vehicle. All you need to have to complete this process is a measuring tape.

BTW: I used the first method (speed & yawrate) to compute the understeer of my bass boat. Its 6 deg/g (linear range) with a 3 bladed prop, 3.5 deg/g with a 4 bladed prop, and 2 deg/g with a 5 bladed stainless steel propeller. Had to estimate the 'wheelbase' though. You get the idea I hope.

 

[drwhitewater]

Depending on your definition of "low cost", you could use 2 accelerometers (one at the front of the vehicle and one at the rear). If front accel is greater than rear you have understeer, if rear accel is greater than front you have oversteer.

 

[GregLocock]

Well, that wouldn't really fit with the SAE definition of understeer.



Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[murpia]

If you subtract the theoretical steer angle required to drive round the measured curvature (which is what I think that MoTeC formula does) from the actual steering angle, you get a very useful trace that clearly indicates oversteer snaps and trends in car balance over several overlayed laps.

Using that trace as an objective measure of understeer is probably taking it too far though.

Regards, Ian

 

[GregLocock]

The excess steer angle (per g) is as useful a definition of understeer as any.



Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[EfiOz]

I have used it and it's pretty rough and ready. This formula is almost unworkable in a V8 Supercar touring car (low aero, big slip angles) but almost works in a reasonably responsive open wheeler.

It will give a very rough guide to US/OS balance but it's so rought that any driver will be feeling it anyway. Magneti Marelli have had a similar thing in there software for years that they just call KUS. Details of the equation are hidden from view though.

 

[murpia]

I don't doubt that a driver can give you better feedback than such a data trace. However it's still very useful once your driver's gone home, for detecting snaps and to overlay laps during a run.

What do you mean by unworkable in a V8 context? Too noisy? Inconsistent answers?

Regards, Ian