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

Firmware 8.0

This site may earn commission on affiliate links.
@TrafficEng and @zambono -- we are now seeing some symptoms that appear to indicate why the car suddenly slowing down on a divided, limited access road is taking place. One theory is that the AP is now reacting to what it perceives as a stopped vehicle or object in the adjacent lane, slowing the car to around 50mph. It may well be that the AP is treating the toll-booth and the bridge support leg as a stopped vehicle in the adjacent lane thus triggering the slowdown until it is past the object and then considering things "all clear."

As we have no release notes or explanation from Tesla, none of us know for sure, but I would encourage you to report these incidents as possible bugs. Assuming the "stopped vehicle in an adjacent lane" trigger HAS been deliberately added to AP software, they clearly need to distinguish that from a bridge support or other fixed object where there is no expectation a passing vehicle should slow down.

I think it is far more likely that what you are seeing is NOT a bug in the software but rather the software is doing what it is programmed to do, but the software isn't compensating/disregarding the bridge support. In systems terms, we have a valid detection (the bridge support for example) but the system is failing to identify it as such (or rather identifying it as a stopped vehicle) and thus triggering the "stopped vehicle in adjacent lane" logic. That makes more sense to me than that the AP is randomly doing this....

it would be interesting to see if you consistently are getting this when passing objects of this type.

I like your theory. It may be detecting the bridge as an adjacent stopped vehicle, or maybe an artifact on the map. Or also could be part of the 'white listing of overhead radar detected objects' which Tesla said they are doing. We'll see if this improves over time, or whether it is an algorithm problem which will require an update
 
I still fail to see how this exact issue is a danger. On the contrary - One is told not to use auto steer above 90mph (wether it is documented somewhere or just a message, a warning, or maybe you aren't told somewhere, it really doesn't mater). One shall not use auto steer above 90mph. Then one ignores that, the car dislikes it and punishes bad the behavior - so you don't do it again. That is improving safety, not making it more dangerous.

If one fails to see that, then there are bigger issues in ones life that needs to be resolved than inability to have auto steer enabled at +90mph. The car will either way do more than 90mph, you just have to be actively driving it. Im plenty of other scenarios, the car will not enable auto steer, i.e. bad or lacking road markings, bad weather, in the middle of a too steep turn - you have to wait for conditions to be good enough for the system to activate. One condition obviously is speed below 90.

Do you deem inability to enable auto steer in those other conditions as a danger, or just the inconvenience with the inability above 90mph?
Will you sou Tesla over a crash at 90+ with auto steer on, because you then suddenly might agree that it wasn't safe after all (if you live to tell about it)?

Have you read this entire thread? Have you at least read the posts, including mine, from people who have explained, many times at this point, that no one here has been complaining about the lack of autosteer over 90 MPH. Let me say that again, in the hope that you and the others who just aren't getting it will read it:

No one in this thread has been complaining about the lack of autosteer over 90 MPH.


The issue that we are complaining about is that when the car has reduced speed back to a speed below 90 MPH, autosteer--the use of which DOES improve safety--is not available. That is our concern.

 
Triggering on overhead signs is interesting since Elon made specific mention of overhead signs being mapped and whitelisted with the neural network learning to support AP2.0. It may well be that the AP1.0 hardware patched software is now not disregarding things that are "up higher" than before, again supporting a theory of a software patch to have allowed detection of (and prevention of) the Florida crash.... If radar is now detecting things that high, it would have detected the white trailer despite the lack of contrast against the sky....radar doesn't care about colors and contrast.

I really do hope Tesla puts out some guidance on the new software changes so we have some facts and not just theories based on observed system behavior.
 
Have you read this entire thread? Have you at least read the posts, including mine, from people who have explained, many times at this point, that no one here has been complaining about the lack of autosteer over 90 MPH. Let me say that again, in the hope that you and the others who just aren't getting it will read it:

No one in this thread has been complaining about the lack of autosteer over 90 MPH.


The issue that we are complaining about is that when the car has reduced speed back to a speed below 90 MPH, autosteer--the use of which DOES improve safety--is not available. That is our concern.

+1 @Andyw2100

Please read the posts before replying with your own.
 
  • Like
Reactions: Richg and Andyw2100
Used the auto steer for the first time since the last update. I must say that they are getting close to an acceptable lane change. In the past the lane change was so violent that I was afraid to use it. Last night I used it 5 times and only once on a curved section of road did it under perform.
 
Used the auto steer for the first time since the last update. I must say that they are getting close to an acceptable lane change. In the past the lane change was so violent that I was afraid to use it. Last night I used it 5 times and only once on a curved section of road did it under perform.

My P85D never had a "violent" AP lane change since initial software release in Oct. 2015 and up to current release.
 
Anyone experienced this? Was driving on a 6-lane divided highway (3 lanes going in opposite directions). Was driving in the lane closest to the emergency lane with Autopilot set at the speed limit (70 mph). As I approached a parked car in the emergency lane, my car started to brake and then reduced the set speed to 50 mph. After passing the car, I tried to increase the set speed back to 70 but she wouldn't have it. I had to disengage and then reengage. Happened twice today. Maybe another safety feature? Running on 2.50.114.
 
Anyone experienced this? Was driving on a 6-lane divided highway (3 lanes going in opposite directions). Was driving in the lane closest to the emergency lane with Autopilot set at the speed limit (70 mph). As I approached a parked car in the emergency lane, my car started to brake and then reduced the set speed to 50 mph. After passing the car, I tried to increase the set speed back to 70 but she wouldn't have it. I had to disengage and then reengage. Happened twice today. Maybe another safety feature? Running on 2.50.114.

Discussed extensively in the last few pages of this thread and another couple threads. It'll actually let you increase the limit again after 15-20 seconds of not seeing another stopped car, without breaking out and restarting.
 
Anyone experienced this? Was driving on a 6-lane divided highway (3 lanes going in opposite directions). Was driving in the lane closest to the emergency lane with Autopilot set at the speed limit (70 mph). As I approached a parked car in the emergency lane, my car started to brake and then reduced the set speed to 50 mph. After passing the car, I tried to increase the set speed back to 70 but she wouldn't have it. I had to disengage and then reengage. Happened twice today. Maybe another safety feature? Running on 2.50.114.

This has been discussed in this thread (in the last 100 posts or so--not thousands of posts back) a fair amount.

The theory is that the software is not yet dealing well with all circumstances surrounding the new feature implementation of trying to slow down when one lane of traffic is moving much faster than another. It makes sense for the car to slow if one lane is travelling 70 MPH, and the immediately adjacent lane is stopped or travelling at 10 or 15 MPH. It does not make sense for the software to slow the car if it is travelling 70 MPH passing a car stopped on the shoulder of the road or in the break-down lane.

The expectation is that Tesla will improve the software to deal better with these situations.
 
This has been discussed in this thread (in the last 100 posts or so--not thousands of posts back) a fair amount.

The theory is that the software is not yet dealing well with all circumstances surrounding the new feature implementation of trying to slow down when one lane of traffic is moving much faster than another. It makes sense for the car to slow if one lane is travelling 70 MPH, and the immediately adjacent lane is stopped or travelling at 10 or 15 MPH. It does not make sense for the software to slow the car if it is travelling 70 MPH passing a car stopped on the shoulder of the road or in the break-down lane.

The expectation is that Tesla will improve the software to deal better with these situations.

It goes to show the need for PREDICTABILITY in autonomous driving system. We are all experienced AP drivers, and this (and other blog posts) show we are all just trying to understand the behavior, so we are not surprised that the car suddenly does something weird. So we would all be helped by an email/release note which said:

' In this release we have incorporated a safety feature to slow Autopilot when a stopped/slow vehicle is detected in an adjacent lane. This may result in an inadvertent slow down when passing overhead signs and bridge structures which can be seen by the sensors as obstructions in the adjacent lane. Behavior will improve in time as these fixed obstructions are white listed by the fleet'

But no such luck.... we are left to guess the intended behavior. Come on, @tesla, let us know how it is supposed to work?
 
It goes to show the need for PREDICTABILITY in autonomous driving system. We are all experienced AP drivers, and this (and other blog posts) show we are all just trying to understand the behavior, so we are not surprised that the car suddenly does something weird. So we would all be helped by an email/release note which said:

' In this release we have incorporated a safety feature to slow Autopilot when a stopped/slow vehicle is detected in an adjacent lane. This may result in an inadvertent slow down when passing overhead signs and bridge structures which can be seen by the sensors as obstructions in the adjacent lane. Behavior will improve in time as these fixed obstructions are white listed by the fleet'

But no such luck.... we are left to guess the intended behavior. Come on, @tesla, let us know how it is supposed to work?

Agreed and similar to a point I made in another thread - Im really confounded by Tesla's lack of communication with their software releases.
The newest .185 release is apparently not working for anyone, they all get a message that the camera needs to be calibrated, but no indication of how or what needs to happen for it to be calibrated. You now have drivers aimlessly driving for tens, if not hundreds of miles hoping that enough time on the road will calibrate the camera and turn the functions on.
 
Have you read this entire thread? Have you at least read the posts, including mine, from people who have explained, many times at this point, that no one here has been complaining about the lack of autosteer over 90 MPH. Let me say that again, in the hope that you and the others who just aren't getting it will read it:

No one in this thread has been complaining about the lack of autosteer over 90 MPH.


The issue that we are complaining about is that when the car has reduced speed back to a speed below 90 MPH, autosteer--the use of which DOES improve safety--is not available. That is our concern.
Can we please quit discussing this issue. I think maybe it needs it's own thread. I would like to follow a thread about firmware 8.0 issues, but I think this particular topic has been taking up way too much space on here.
 
Last edited:
  • Like
Reactions: scottf200 and jgs
Can we please quit discussing this issue. I think maybe it needs it's own thread. I would like to follow a thread about firmware 8.0 issues, but I think this particular topic has been taking up way too much space on here.

Why did you choose to quote me, instead of one of the people who keep trying to completely derail the discussion by arguing against things that no one in this thread has been arguing for? My explanation of that--the post you chose to quote--was intended to reduce the number of truly off-topic posts.

Also, although there have been a lot of posts about this issue, it --IS-- a Firmware 8 issue, so the discussion probably does belong in this thread.
 
  • Like
Reactions: Denarius
It goes to show the need for PREDICTABILITY in autonomous driving system. We are all experienced AP drivers, and this (and other blog posts) show we are all just trying to understand the behavior, so we are not surprised that the car suddenly does something weird. So we would all be helped by an email/release note which said:

' In this release we have incorporated a safety feature to slow Autopilot when a stopped/slow vehicle is detected in an adjacent lane. This may result in an inadvertent slow down when passing overhead signs and bridge structures which can be seen by the sensors as obstructions in the adjacent lane. Behavior will improve in time as these fixed obstructions are white listed by the fleet'

But no such luck.... we are left to guess the intended behavior. Come on, @tesla, let us know how it is supposed to work?

Is that a valid handle, @tesla? Because if it is, we should start including all known Tesla handles in these posts. That may garner some needed attention as well.