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

Software version 17.34 2448cfc

This site may earn commission on affiliate links.
I think I finally figured out what causes the AP ping ponging.

The AP gets confused when there is a concrete paving joint within the lane. It often ping ponging trying to adjust itself. At times, the screen will read the concrete paving joint as the lane and try to center itself.

I tried to snap a photo today, but the AP was too dangerous to use.
I wonder if there's a camera alignment issue on some cars that has owners experiencing problems. I've put 3000 miles on my MX in the month I've owned it, with up to half of those on AP. It's a 2016 inventory model, so probably one of the 1st AP2 P100Ds off the line. I've always turned it off when road conditions looked unsafe, construction lane shifts, worn off lane markings w/ lots of crack sealing, etc.

I've used it extensively though in areas where recent road work has completed and I've dealt with concrete seams down the middle of the lane, angled through the lane or even driving on half asphalt/half concrete.

My wife commented on how smooth AP was driving back from dinner last night and that included multiple lane changes.

People's experiences seem so widely varied with AP, that I can't help but wonder if some other factor is it fault.

I have noticed two problems though that I watch for now. When the shadow from concrete center barriers aligns perfectly with the left lane marker, AP can't recognize it anymore. It still tracks the right lane marker, but feels a little less stable. Second, when the highway widens up from 2 to 3 and then 4 lanes, the left lane marker pulls away and they don't start the new lane marker for maybe 100-150ft. In all those cases, AP has tracked the left solid line instead of the right dashed line, so I've technically made a lane change.

In both of those cases though, it responded gracefully.
 
I wonder if there's a camera alignment issue on some cars that has owners experiencing problems. I've put 3000 miles on my MX in the month I've owned it, with up to half of those on AP. It's a 2016 inventory model, so probably one of the 1st AP2 P100Ds off the line. I've always turned it off when road conditions looked unsafe, construction lane shifts, worn off lane markings w/ lots of crack sealing, etc.

I've used it extensively though in areas where recent road work has completed and I've dealt with concrete seams down the middle of the lane, angled through the lane or even driving on half asphalt/half concrete.

My wife commented on how smooth AP was driving back from dinner last night and that included multiple lane changes.

People's experiences seem so widely varied with AP, that I can't help but wonder if some other factor is it fault.

I have noticed two problems though that I watch for now. When the shadow from concrete center barriers aligns perfectly with the left lane marker, AP can't recognize it anymore. It still tracks the right lane marker, but feels a little less stable. Second, when the highway widens up from 2 to 3 and then 4 lanes, the left lane marker pulls away and they don't start the new lane marker for maybe 100-150ft. In all those cases, AP has tracked the left solid line instead of the right dashed line, so I've technically made a lane change.

In both of those cases though, it responded gracefully.

I echo your points. I have the car for 3 months. Around $3000 miles so far. I used AP2 single day. Some versions is better than others, but since 2017.32, it is super smooth and close to perfect. However, in situation where it does not look too unsafe, I disabled it and took over.

Yes, one can argue that AP2 should handle all the situation. The reality is that it is not. We can complain all day long, but let's not take away what it can do today....as imperfect as it is. Know it limitation and use it accordingly.

For those who have problems with jerkiness or phantom brake, I have a theory. Maybe it is related to the distance between you and the car in front of you. Naturally, if you set the car too close to the front, it would brake hard if it senses slight slow down from the car in front of you. I set it to 4. I find this distance perfect for me. Not too far to invite other cars to cut in, but not too close for the car to sudden slow down
 
Yes, I see some slight difference in behavior now, but cannot quantify yet.

I had a major summon malfunction today. I've used it for a year or so to back out of my garage. Most of the time it works great, but sometimes if it senses the car is too far left or right to provide adequate clearance, it aborts. That makes sense.

Today the car was pretty well centered in the garage, and I initiated summon with the key fob, standing on the left side of the garage to get in the drivers side when it came out. It did what it usually does, e.g. comes out a little way, pauses and then comes out the rest.

Because I was standing to the side I couldn't see that it had swerved left and the next thing I knew it caught the driver's mirror, snapped it back, and then wedged the left side of the car, just in front of the driver's door, against the side of the garage. Even though it was immobilized against the garage door frame, for a short while you could see it continuing to try to move backwards--wedging it even tighter.

The driver's door couldn't open so I tried to summon forward, but that didn't work. So I climbed in through the passenger side and steered it hard right and moved forward.

At that point I could see the the mirror was just swinging freely and the little glass panel in front of the window was shattered. When I got the car out of the garage I could see the door and the front quarter panel were misaligned but fortunately the only other body damage appears to be a scrape on the plastic front left wheel well plastic strip.

I took pictures, and the car is now at the local Tesla service center. I have no idea why this happened, as previously the car would simply refuse to move if it thought it didn't have adequate room. I've never seen any dramatic steering adjustments in a summon, but this time it had to turn quite a bit to end up where it did. Some sensor probably either malfunctioned or was misinterpreted.

I haven't seen any mention of another similar incident, but thought I'd let everyone know. As I posted earlier, I did notice some minor summon oddities after the last update. Maybe there's a bigger problem, or maybe this was just a freak occurrence. If I learn more I'll provide updates.
 
  • Informative
Reactions: oktane
IMG_3954.JPG
IMG_3955.JPG
IMG_3956.JPG
Post your pics if you will, they will say more than the description of the damage.
 
  • Informative
Reactions: oktane and lunitiks
I had a major summon malfunction today. I've used it for a year or so to back out of my garage. Most of the time it works great, but sometimes if it senses the car is too far left or right to provide adequate clearance, it aborts. That makes sense.

Today the car was pretty well centered in the garage, and I initiated summon with the key fob, standing on the left side of the garage to get in the drivers side when it came out. It did what it usually does, e.g. comes out a little way, pauses and then comes out the rest.

Because I was standing to the side I couldn't see that it had swerved left and the next thing I knew it caught the driver's mirror, snapped it back, and then wedged the left side of the car, just in front of the driver's door, against the side of the garage. Even though it was immobilized against the garage door frame, for a short while you could see it continuing to try to move backwards--wedging it even tighter.

The driver's door couldn't open so I tried to summon forward, but that didn't work. So I climbed in through the passenger side and steered it hard right and moved forward.

At that point I could see the the mirror was just swinging freely and the little glass panel in front of the window was shattered. When I got the car out of the garage I could see the door and the front quarter panel were misaligned but fortunately the only other body damage appears to be a scrape on the plastic front left wheel well plastic strip.

I took pictures, and the car is now at the local Tesla service center. I have no idea why this happened, as previously the car would simply refuse to move if it thought it didn't have adequate room. I've never seen any dramatic steering adjustments in a summon, but this time it had to turn quite a bit to end up where it did. Some sensor probably either malfunctioned or was misinterpreted.

I haven't seen any mention of another similar incident, but thought I'd let everyone know. As I posted earlier, I did notice some minor summon oddities after the last update. Maybe there's a bigger problem, or maybe this was just a freak occurrence. If I learn more I'll provide updates.
I don't trust summon anymore after I had a similar experience, but not on this software version. My car was halfway out of my garage and stopped, so I initiated summon again. At that point my car turned its wheels to the right and started backing up. Thank God I was able to stop it with the key fob. It made no sense why it did this maneuver. I was shocked and I took a few pictures of the position the car was in after I aborted the summon. That happened on July 1st.

IMG_2788.JPG
IMG_2790.JPG
 
I came on here to say my car has attempted to swerve off the freeway at two different off ramps in the last couple of days. Definitely pissed the driver behind me off when I swerved my car back onto the freeway. Of course he was following me to close. This is the only software version I've experienced this. It always does a little twitch towards the off ramps but has never tried to take them.
 
I came on here to say my car has attempted to swerve off the freeway at two different off ramps in the last couple of days. Definitely pissed the driver behind me off when I swerved my car back onto the freeway. Of course he was following me to close. This is the only software version I've experienced this. It always does a little twitch towards the off ramps but has never tried to take them.
Me too, and it's consistent (repeatable). It tends to happen when freeway is curving to left slightly, and an off-ramp has well marked (painted) lines. The car ignores the left lines and wants to follow the right, and exit.

old |v8.1(17.28 c528869)| AP2 |
new |v8.1(17.34 2448cfc)| AP2 |
 
The confusion, that if the car should take the lane to the right (which happens to be an offramp) or the one to the left (which happens to be the freeway's right most lane) should not be surprising to anyone.

How will the car know which one is the legitimate lane ? The car has to either choose the right line or left line as the primary line during these kind of decision making points. Quite rightly the car gives prominence to the line on the right, which I am sure is the right decision 9 out of ten times, but not an exit. Happens to be problem for both AP1 & 2

If you put on your thinking hat a bit, you won't be this frustrated.
 
The confusion, that if the car should take the lane to the right (which happens to be an offramp) or the one to the left (which happens to be the freeway's right most lane) should not be surprising to anyone.

How will the car know which one is the legitimate lane ? The car has to either choose the right line or left line as the primary line during these kind of decision making points. Quite rightly the car gives prominence to the line on the right, which I am sure is the right decision 9 out of ten times, but not an exit. Happens to be problem for both AP1 & 2

If you put on your thinking hat a bit, you won't be this frustrated.
Agree. And FSD will have to navigate it. But short term, us testers will like to see TM reduce the "false exit" risk. For example, if no signal indicator is detected, follow the left line, I donno.......
Maybe a new thread listing the places where it happens everytime will be in order.