Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

2018.10.4 Problems

This site may earn commission on affiliate links.

Padelford

Member
Supporting Member
Jul 1, 2017
683
637
Seattle
Here are a couple of 2018.10.4 problems I discovered today.


The first problem, failure to complete an S-turn, could be a problem with the left side lane-line tracking algorithm. The vehicle seems to loose lane control when the yellow left line turns white.

The second problem is an extreme case of the on-ramp vehicle offset. Something about this on-ramp causes AP to make an extreme, full-lane shift to the right, then the vehicle guides back onto the freeway off the right white line.
 
...failure to complete an S-turn...

Thanks for the video. That's a good catch. It's now pretty good on same direction curve like a circle, like sharp turn to the right only but once the curve changes to the other side like sharp curve to the left, like the way an S looks, it's a pretty bet that it's not yet competent yet for an S curve.

...full-lane shift to the right...

This has been a consistent behavior since the beginning. It's pretty good if the right lane has the same width.

But once the lane is widened from 1-lane width to 2 or more lane width, it has to make a decision:

1oYUaoA.jpg


1) stay close to the left lane marking
2) straddle in the middle of now gigantic lane
3) or in your case, stay close to the right side marking.

Some one has to make a decision how to code your Autopilot in this scenario.

None of those is illegal in California where 1 lane only allows 1 car at a time. It's illegal to fit 2 cars in a gigantic lane.
 
Last edited:
Thanks for the video. That's a good catch. It's now pretty good on same direction curve like a circle, like sharp turn to the right only but once the curve changes to the other side like sharp curve to the left, like the way an S looks, it's a pretty bet that it's not yet competent yet for an S curve.



This has been a consistent behavior since the beginning. It's pretty good if the right lane has the same width.

But once the lane is widened from 1-lane width to 2 or more lane width, it has to make a decision:

1oYUaoA.jpg


1) stay close to the left lane marking
2) straddle in the middle of now gigantic lane
3) or in your case, stay close to the right side marking.

Some one has to make a decision how to code your Autopilot in this scenario.

None of those is illegal in California where 1 lane only allows 1 car at a time. It's illegal to fit 2 cars in a gigantic lane.

I think the vehicle would have finished the S-turn and stayed in the exit lane if the left lane marking/color hadn’t changed.
 
I think the vehicle would have finished the S-turn and stayed in the exit lane if the left lane marking/color hadn’t changed.

It's an interesting theory.

I myself don't think so. The competency of Autopilot is incremental and it takes time to achieve a simple competency after another.

If you do a search in your Owner's Manual for the term "winding roads", you would see that's one of many limitations of the current Tesla driver assistance system.

Some day it will handle "winding roads" regardless of lane colors but not now.
 
I’m not sure where to send this. I’ve sent in videos of AP problems before and never heard anything back.

I’ve heard that Tesla folks watch these forums. Maybe they’ll find this.
Send it to [email protected]. Subject line: Bug Report. It will get to the engineering team. And no, you probably won't hear anything back, but it will get to them and be reviewed.

Hoping they'll see it here is hit and miss. Someone might see it. Someone might let the right people know. But you really can't count on the forum as a feedback tool.
 
Last edited: