Active steering centering?

[Gauss Guzzler]

I've noticed that the Model 3 drives much straighter than my previous cars and I've also sensed a little extra heaviness in the steering around center that almost feels artificial...

... then one day I was adjusting front toe, and mistakenly turned both tie rods the same direction instead of opposites. Doh!
I took it for a drive and the steering wheel was a bit crooked and a steering/autopilot warning soon came on screen - of course. It was also now pulling hard left but I had places to go so I kept driving for ~50 miles and the left pull seemed to get a little better along the way but was still very bad by the time I got home. The next day I corrected my mistake and then it pulled slightly to the right for the next ~20 miles before eventually settling back into perfectly straight.

How is this possible? I never touched anything other than the tie rods. No camber, caster, rear, anything. I never even jacked the car up, I just reached under the front bumper and spun the tie rods.

So it sure seems like the car is using GPS to gradually determine what "straight" is and then adding a little bit of steering assist to make it track perfectly, right? I'm curious to know what the algorithm might be and what other cars may be using it. Does anyone here know about such a technology?

 

[Candleflame]

i sometimes also feel that the car seems to drive unnaturally straight when you let go of the steering wheel. but i presume thats just because the wheels are properly balanced and theres less vibration....

 

[XPsionic]

Even if the electric power steering have a self centering feature (which I haven't heard in a Tesla. Some cars have it for crosswind compensation), if you turn the tie rods so both front tires are pointing one way more than the other, you should see the car tracking straight with the steering wheel slightly turned.

Is that what you're seeing? Centering force causes the car to go straight with steering wheel slightly off center?

 

[Gauss Guzzler]

No, it was the opposite. It would veer off the road if I let go of the wheel. The steering wheel was crooked, and it would veer in the direction of straightening the wheel.

I assume this is because it has "learned" over many miles of driving that for perfect tracking, the steering wheel should be at some angle, say -0.2 degrees. But now that I'm driving along with the wheel at +3.7 degrees it's nudging the wheel left.

 

[Lindenwood]

Technically, front wheel alignment doesn’t cause a “pull”—I.e. a constant force on the wheel to keep straight. As mentioned above, if the front wheels are misaligned, the steering wheel will simply be crooked when driving straight (even if there is a tendency to feel unstable (severe toe-out) or heavy (severe toe-in). What causes a pull is when the rear wheels have imbalanced toe, creating a thrust angle that does not run through the center of the longitudinal axis of the car (I.e. directly between the front wheels).

That said, it still doesn’t refute your hypothesis. It is theoretically possible the car has been automatically compensating for an existing off-center thrust angle, and is having to “re-learn” since the steering wheel alignment (and thus the car’s definition of “centered” steering) changed. However, as mentioned, I am not sure this is supported by any specific documentation.

 

[Welbach1]

"adding a little bit of steering assist to make it track perfectly, right?"

the simple answer is yes.

 

[brkaus]

It would likely use the cameras to establish/calibrate straight.

 

[dbeiler]

Modern cars use a column mounted steering angle sensor to ascertain "straight". When the steering wheel is held in a specific position for extended periods of time; it is assumed (and, most likely, taught) to be the straight ahead position. This is why most modern cars require a steering angle sensor reset after a wheel alignment. According to Hunter, over 40 million vehicles on the road today require a reset/recalibration of the steering angle sensor (SAS) following a wheel alignment. Link: Steering Angle Sensor Reset | Hunter Engineering Company®
Tesla vehicles are no exception.
Steps 10-13 of the Model 3 alignment procedure copy and pasted from the Tesla service manual:

10. Connect a laptop with Toolbox 3 to the vehicle.
11. In Toolbox 3, click Actions, type "angle" in the search field, click Play to the left of "PROC_EPAS_ESP_CLEAR-ANGLE-OFFSETS", and then select RUN.
12. Click Actions, type "angle" in the search field, click Play to the left of "TEST_EPAS3P_X_CHECK-APPLIED-ANGLE-OFFSET", and then select RUN.
13. Disconnect the laptop from the vehicle.

This angle reset has been a problem for many Tesla owners seeking alignments from independent repair shops who did not have easy or free access to Tesla reset software. The Snapon, Matco, Autel, etc tools used by independent shops don't have Tesla reset capability. My personal $1600 Autel MaxiSys can reset steering angle on pretty much any car..........except Teslas.

Fortunately, as of 2022.20.6+, Tesla service mode is no longer geofenced to service centers. Link:

We now have access to DIY reset EPAS angle offsets. Hooray!

 

[Gauss Guzzler]

Modern stability control systems rely upon the steering angle sensor to determine the driver's intention and guide the car accordingly - hence the need to ensure reasonable accuracy and reset the zero calibration as needed. That's unrelated to any "active centering assist" however.

But you're right, it appears that we can reset this angle calibration without Toolbox3 now. I didn't know that!

 

[dbeiler]

to determine the driver's intention and guide the car accordingly ............... That's unrelated to any "active centering assist" however.

Why do you consider these unrelated? Are you under the impression that SAS (steering angle sensor) data is used ONLY by stability control systems? SAS data is used by the full suite of modern car ADAS. SAS data is also used to mimic and even improve upon basic function of the EPS racks that have long been replacing hydraulic systems.

The fine details and description of operation for anything regarding ADAS, EPS, etc aren't really available to us or anyone outside the engineering department. This will be Exhibit A for any future Right-To-Repair movements concerning ADAS cars. We can reverse engineer. Disconnect the SAS connector or cut the SAS communication wires and see what doesn't work on your car anymore.

 

[Gauss Guzzler]

Because traditional stability control systems use the brakes on a wheel-by-wheel basis to attempt to match the yaw rate measured by the gyro sensor with the calculated yaw rate that should be expected for the particular steering angle measurement at that speed. This has nothing to do with power steering. It simply reads the steering wheel position and the gyroscopic yaw yate and slams the brakes accordingly.

Some stability control systems might improve upon the above algorithm by intervening with the driver's steering input in an effort to prevent a rollover or spin in extreme circumstances. This does involve power steering but it's an even more extreme case.

Those cases involve an extreme loss of traction, major computer intervention, and I'm not referring to either. I'm only referring to casually driving to the market on a sunny day and having the computer just slightly lean on the steering to make the car drive a little straighter. It's a different algorithm for a different circumstance and totally unrelated to the safety system that kicks in when the car has lost traction and is sliding dangerously out of control.

 

[tm1v2]

My last ICE car had stability control. It had a clockspring steering angle sensor which dumped readings onto the CAN bus that got used by stability control among other systems on the car. After I upgrade to a quick steering rack (going from 15:1 stock to 11.5:1), if stability control was on it would freak out and grab brakes literally every turn.

That was expected of course. I kept stability control disabled and everything else worked fine. No issues from racetrack to blizzards...so long as I kept stability control off. The other systems that nominally used the steering angle sensor seemed not so sensitive, no issues with them.

M3P is a different story though. There's no official way to disable stability control besides entering Track Mode. I bet the Handling Bias relies heavily on the steering angle sensor too. On the M3P I wouldn't do anything that messes with steering angle readings. Thankfully it comes with nice quick steering from the factory.