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

Riding too far to the left

This site may earn commission on affiliate links.
But, it did get me to wonder why a car with side sensors would have trouble following a concrete barrier such as what they have on Hwy 17. I would think it would be a rather easy surface to follow at a safe distance, i.e. in the lane. Can the ultrasonic sensors not see the barrier? Or, are they ignored (and if so, why)?
Or because of that the algorithm was written assuming very confident of the distance, so they give a larger buffer on the right that they are less sure about? *shrug*
 
First off, I agree with the posters who would never even think about using AP on Highway 17 through the mountain. I think twice to even drive it myself. And to think that once upon a time, that barrier was no taller than the yellow stripe down the middle of the road.

But, it did get me to wonder why a car with side sensors would have trouble following a concrete barrier such as what they have on Hwy 17. I would think it would be a rather easy surface to follow at a safe distance, i.e. in the lane. Can the ultrasonic sensors not see the barrier? Or, are they ignored (and if so, why)?

That has to be balanced with all of the potential side effects of falsely sensing something via side sensors and making an unwanted steering decision based off that.
 
Thanks everyone for your input. For the record, I am on 10.5

I am sorry but I have only one week of experience driving Teslas, so I cannot be out of my mind, maybe out of my depth.

That is why I brought the question here. It seems the consensus is that no one trusts the car to stay away from the barrier, the interesting part is that after the first few times when I grabbed the wheel, wondering what it was thinking, I started to trust it, and that's when it got as close as, you can't really tell, in the video. It's close. I am OVER the yellow line and riding in the gutter going over the grates for the drains.

But the car DIDN'T hit the wall, and the sensors detected it, and is showing red, red, red on the display.

What got me to agree with the sentiment here is the car did the same thing on a two-lane windy road and in that situation. The car crossed into oncoming traffic. I was watching and there were no cars, and it came back into the correct lane itself. I didn't have to disengage, just wonder if it would ever actually hit the wall. If none of us are trusting it to drive for us, then we may never know.

The video released earlier showing someone reproducing the accident from this week makes the way the Model X hit the divider make sense. I have certainly not heard of anyone having their Tesla run into a wall it's driving parallel to. So I don't know if I am actually expecting trouble, but I do expect to travel this road a LOT. I'll report any issues here

-Randy
 
  • Like
Reactions: ℬête Noire
Hmm...my main reason for going with the 3 is to do a 30m commute in LA traffic. Mostly in the left carpool lane with concrete barrier a few feet away.

Is this really not a good idea based on people’s experience with AP in the 3? This would change things considerably.

I mean I suppose a lot of it is low speed and I had a woman behind me the other day doing her makeup with a 3 on AP. We were doing maybe 2-10mph and she was full on looking in the visor doing full makeup just casually glancing at the road once in a while for a good 20min.

I guess it’s still helpful when the traffic is very heavy but the limitations implied earlier in this thread are kind of discouraging ...
 
  • Funny
Reactions: Robert831
Thanks everyone for your input. For the record, I am on 10.5

I am sorry but I have only one week of experience driving Teslas, so I cannot be out of my mind, maybe out of my depth.

That is why I brought the question here. It seems the consensus is that no one trusts the car to stay away from the barrier, the interesting part is that after the first few times when I grabbed the wheel, wondering what it was thinking, I started to trust it, and that's when it got as close as, you can't really tell, in the video. It's close. I am OVER the yellow line and riding in the gutter going over the grates for the drains.

But the car DIDN'T hit the wall, and the sensors detected it, and is showing red, red, red on the display.

What got me to agree with the sentiment here is the car did the same thing on a two-lane windy road and in that situation. The car crossed into oncoming traffic. I was watching and there were no cars, and it came back into the correct lane itself. I didn't have to disengage, just wonder if it would ever actually hit the wall. If none of us are trusting it to drive for us, then we may never know.

The video released earlier showing someone reproducing the accident from this week makes the way the Model X hit the divider make sense. I have certainly not heard of anyone having their Tesla run into a wall it's driving parallel to. So I don't know if I am actually expecting trouble, but I do expect to travel this road a LOT. I'll report any issues here

-Randy

The Model 3 is NOT a self-driving car. AP is a driving aid. Tesla says, and the owner's manual repeats over and over, under big red caution symbols, that when using AP you must remain alert and ready to take control of the car at an instant's notice. An alert driver with AP is (maybe?) safer than a driver without AP. But a driver who treats AP as though it were FSD is begging for a Darwin award.

Do not use AP or trust it to avoid hitting something under conditions when it appears to be having trouble staying in its lane. Randy: the behavior of the car you are describing, is the car telling you, "I can't really deal with this road. Keep using AP if you want to die. Otherwise turn off AP and drive the car yourself." AP might save a life when there's a moment of inattention. That does not mean we can ignore the road.

I do not know the road in question, but it sounds to me like one that AP is not yet ready for. Maybe in a future software version.

And I'd try to get the license plate of the woman mentioned in the above post who was ignoring the road while she applied makeup and report her to somebody. It probably won't do any good. But when she does kill herself in a car crash, there will be a record people can point to, to show that she habitually drives recklessly, and not blame the car.
 
I am not putting on makeup at 65 on an insanely windy road, I am driving it with AP on and my hands on the wheel. My query was about "Do I trust it when it gets closer to the barrier than I would". People here said no, and I hope a software update will get it back in the center of the lane.

I am experimenting with AP as it seems all Tesla owners are, trying to find where the line is currently. Saying Just don't use it, isn't helpful, that doesn't move the product forward.

I have used AP like the makeup lady, in 2-10 MPH stop and go traffic I am comfortable tuning the radio or something, as no one is gonna die if I tap their bumper at those speeds and that is probably her calculation. I think many Tesla owners are comfortable in that situation and it's what relieves stress driving at Tesla to work.

I mostly use AP as a second set of eyes, technically a third, the wife always complains about my driving, now she can complain about the car's driving instead. I am still there if the car heads into a 45 MPH curve at 72 (and of curse the wife complained about that too)

-Randy
 
It sounds like you're doing it right. I was not saying don't use it. I was saying when it gets closer to the barrier than you would, or crosses the line, then take over from it. From the comments of others, it sounds to me as though AP may not be able to handle that particular road.
 
Hmm...my main reason for going with the 3 is to do a 30m commute in LA traffic. Mostly in the left carpool lane with concrete barrier a few feet away.

Is this really not a good idea based on people’s experience with AP in the 3? This would change things considerably.

I mean I suppose a lot of it is low speed and I had a woman behind me the other day doing her makeup with a 3 on AP. We were doing maybe 2-10mph and she was full on looking in the visor doing full makeup just casually glancing at the road once in a while for a good 20min.

I guess it’s still helpful when the traffic is very heavy but the limitations implied earlier in this thread are kind of discouraging ...

If you are in LA traffic I would say it is no problem, plenty of time to react if the car starts to veer. In general I'm a bit wary about autopilot at full freeway speeds and careful when I use it, but heavy slower traffic is when I always use it. Even highway 17 I would use it if the traffic is stop and go.
 
Thanks everyone for your input. For the record, I am on 10.5

That is why I brought the question here. It seems the consensus is that no one trusts the car to stay away from the barrier, the interesting part is that after the first few times when I grabbed the wheel, wondering what it was thinking, I started to trust it, and that's when it got as close as, you can't really tell, in the video. It's close. I am OVER the yellow line and riding in the gutter going over the grates for the drains.

But the car DIDN'T hit the wall, and the sensors detected it, and is showing red, red, red on the display.
I find Autopilot a great aid on the road. Particularly construction zones. Also when carrying grandchildren.

Early last November a friend and I drove my wife’s birthday present - 2017 S100D AP2 - over 750 miles from Delivery in Chicago home to Philadelpia, in one long day. There were many construction zones on the Pennsylvania Turnpike, which was between dinner and midnight.

We were each on our second shift behind the wheel by then, familiar with Autopilot and using it as much as possible.

Autopilot did well through construction zones. It kept us centered in the lane and properly separated from the vehicle ahead. Much less stress for me, and I had done many miles driving a full-size box truck through narrow cattle chutes.

When we arrived home at midnight, we were far less tired from 15 hours of travel than one would expect.

On the other hand, last weekend, with 2018 10.4 8bbdc66, I found Autopilot getting very close to the underpass wall while in the left lane of 676 crossing under Philadelpia. When I heard the rumble strip, I took over and moved to the center lane.

I also regard Autopilot as an additional safeguard when carrying grandchildren. The stakes are higher, so is the distraction factor. Autopilot is an unblinking, if incomplete tool that will help keep the car in lane and safely separated if I’m momentarily distracted.
 
Hopefully, when they are tuning where the car rides in the next autopilot they will also tune the crazy fast-breaking, I would appreciate it if it started slowing earlier. Maybe add it to a setting like following distance, since there are places where if you let them, people will cut you off. I really want to uncheck the Herky-Jerky box, have it drive more like a person.

-Randy
 
Hopefully, when they are tuning where the car rides in the next autopilot they will also tune the crazy fast-breaking, I would appreciate it if it started slowing earlier. Maybe add it to a setting like following distance, since there are places where if you let them, people will cut you off. I really want to uncheck the Herky-Jerky box, have it drive more like a person.

-Randy

This is part of the reason I don't use AP in town. Where there are cars stopped at a light, it seems to maintain set speed until it has reached the follow distance and then stops very quickly. It needs to be able to see that there are stopped cars ahead and slow down more gradually. Then when the traffic starts to move, it waits until the car ahead has reached the follow distance before it begins to accelerate, rather than starting up smoothly with the flow of traffic.

But as I said, I just figure these BETA features are not ready for city driving yet. (The owners manual says that both TACC and Autosteer are BETA features.)
 
  • Like
Reactions: Robert831
1) Ouch. Total I expect?
2) That's not "running parallel". That barrier is angled across the lane, check the lines. It was a end of lane, looks like a temporary construction detour.
3) Still a good example of how AP isn't a driverless system, requires a level of alertness by the driver. EDIT: Assuming the AP was actually turned on? Was there confirmation about it?

EDIT: What appears to be the similar (same?) section of road using AP, in comments further down the comments:

EDIT2: Wait, no. That's someone "grabbing the wheel". Seems to be confirmation that AP isn't handling it.

 
Last edited:
  • Like
Reactions: T-Will
Mine stays more centered. However centered does not always equal to "good".

When I pass a semi for instance that uses nearly it's enfire lane, id rather be more on the left of the lane as opposed to the dead center.

What would be awesome would be traffic awareness of adjacent lanes with AP making dynamic adjustments to the side of the lanes to leave more room. That and of course using that information to never stay in someone's blind spot.

That would really take safety to the next level.
 
1) Ouch. Total I expect?
2) That's not "running parallel". That barrier is angled across the lane, check the lines. It was a end of lane, looks like a temporary construction detour.
3) Still a good example of how AP isn't a driverless system, requires a level of alertness by the driver.
I don't know, that vid was on youtube for a long time. There's even a chance they fixed something in the AP for this scenario(although I doubt, b/c AP is bad for stationary objects, like that fire truck that AP recently ran into). But then we just had a fatal X crash for another scenario... I assume for a while there will be some scenarios that are not accounted for and not handled properly. So, there will be few more deaths. Hopefully, few years down the road we can trust AP more.
 
Hmm...my main reason for going with the 3 is to do a 30m commute in LA traffic. Mostly in the left carpool lane with concrete barrier a few feet away.

Is this really not a good idea based on people’s experience with AP in the 3? This would change things considerably.

I mean I suppose a lot of it is low speed and I had a woman behind me the other day doing her makeup with a 3 on AP. We were doing maybe 2-10mph and she was full on looking in the visor doing full makeup just casually glancing at the road once in a while for a good 20min.

I guess it’s still helpful when the traffic is very heavy but the limitations implied earlier in this thread are kind of discouraging ...
My limited experience with it is it can take you by surprise and when it does i would rather not have 1/ center divider next to me and 2/ going fast (which the fast lane is intended for).
 
I don't know, that vid was on youtube for a long time. There's even a chance they fixed something in the AP for this scenario(although I doubt, b/c AP is bad for stationary objects, like that fire truck that AP recently ran into). But then we just had a fatal X crash for another scenario... I assume for a while there will be some scenarios that are not accounted for and not handled properly. So, there will be few more deaths. Hopefully, few years down the road we can trust AP more.

You're two edits behind! ;)

I added a link to a video that someone claimed suggested AP worked in that sort of situation (which I'd find surprising for the reason you give). But the sounds and I think dash icons suggest they grabbed the wheel because they didn't think AP was handling it?
 
But the sounds and I think dash icons suggest they grabbed the wheel because they didn't think AP was handling it?
Yeah, I believe those are "take control" signs and the driver had very little time to comply. AP is not FSD, so it is not supposed to know how to handle complex scenarios, it's driver's responsibility.
Note, however, that in the crash scenario the car in the front was blocking the view, so potentially due to that the AP was not able to give an advanced warning. This is still very real concern for anybody, who would like to trust AP beyond reasonable limits.