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

Navigate On Autopilot: automatic lane change results

This site may earn commission on affiliate links.
FWIW, @ggreen4077 , I can't fully interpret your list, but before your appt, what have you got to lose, try the following to do a proper reset:

1- Dust all the cameras off with a dry microfiber towel, no point in getting the lenses wet unless they are caked. Wipe all the secondary sensors on the car body with a damp microfiber towel and dry.
2- Reset the display computer by holding both scroll wheels until a few seconds after the Tesla logo re-appears.
3- Reset the main computer by setting your phone to airplane mode, tap the Power Off in Security menu, sit in car 5 minutes without moving out of your seat. Car will gradually power down to dead. Wake phone out of airplane mode so the car can be unlocked and ready to drive. Then tap brake to awaken car.

4- Go for a drive and see if things are better.

No matter all the naysayers who think it's "unnecessary", after seeing that it clears many anomalies, I now also do this after every update. It only takes a few minutes, and I've had far FAR fewer issues than most.

Except the one time I forgot to wake the phone and I couldn't silence the alarm until I did :rolleyes:
.
Thanks, I'll give it a try.
 
FWIW, @ggreen4077 , I can't fully interpret your list, but before your appt, what have you got to lose, try the following to do a proper reset:

1- Dust all the cameras off with a dry microfiber towel, no point in getting the lenses wet unless they are caked. Wipe all the secondary sensors on the car body with a damp microfiber towel and dry.
2- Reset the display computer by holding both scroll wheels until a few seconds after the Tesla logo re-appears.
3- Reset the main computer by setting your phone to airplane mode, tap the Power Off in Security menu, sit in car 5 minutes without moving out of your seat. Car will gradually power down to dead. Wake phone out of airplane mode so the car can be unlocked and ready to drive. Then tap brake to awaken car.

4- Go for a drive and see if things are better.

No matter all the naysayers who think it's "unnecessary", after seeing that it clears many anomalies, I now also do this after every update. It only takes a few minutes, and I've had far FAR fewer issues than most.

Except the one time I forgot to wake the phone and I couldn't silence the alarm until I did :rolleyes:
.

Based on your advice, rebooted display and main computers, took it for a test drive. Appears about half of the issues were resolved.

Using NOA, car still would not enter I69 on ramp, but once on the interstate, it did move to the correct lane and exit. It also passed slower vehicles. Rear facing blind-spot cameras still do not pickup on oncoming traffic at a safe distance.

It's a good start. Thanks for the help.
 
Great. Good start.

But I'll tell ya, we all need some trial and error time. It took me many weeks to get good at using the system.

For one thing: expectations. Like are those "delayed" rear views of oncoming traffic really affecting anything, or are they just displays? Does the car change lanes into dangerously close cars? I don't know in your case, and I hope your service tech will ride with you and help sort that out.

Some situations you know it won't handle well, so you learn to disable AP/AS/NOA, sometimes for a few seconds even, and then turn it back on. I see the AI as a personality to ride, more like a horse than a toaster. And it's not really a contest or a stern evaluation, it's a partnership, it's what works for you and makes your driving happier. After getting to know my Nicki, I think we drive better together, with less effort, than apart.

Another detail is that the software is constantly changing. Sometimes they improve X at the cost of Y. Over time it's gotten better and better, but not without occasional setbacks. I average well > 90% in AS/AP/NOA and it's great, but I'd be a fool to trust ANY driver, human or AI, completely. Never fully take your hand off the wheel or your eyes off the road.
.
 
Last edited:
Using NOA, car still would not enter I69 on ramp, but once on the interstate, it did move to the correct lane and exit. It also passed slower vehicles. Rear facing blind-spot cameras still do not pickup on oncoming traffic at a safe distance.

How did you determine that? The distance as shown in the car display is somewhat different from the way you would perceive it looking in read-view and side mirrors.
 
This is the final video with Enginerd's lovely data...Or is it??? :cool:

I made this vid using that last drop of Enginerd's data, it's a bit basic so considering making a dedicated reel showing the progress of Enginerds data over the past 12 months, as a bit of a thank you and testimonial. What do y'all think?

I am relatively new to the site.
I am a bit confused. Your later post comments on your modified commute and then the video is from England?
For some reason, my experience we the new update wasn't comforting as it didn't appear to slow down as it approached the traffic light. Even though the light was green, going through the intersection at the higher speed was discomforting. Thank you.
 
upload_2020-5-15_15-37-3.png

I picked up an interesting tidbit today during a visit from Tesla mobile service (hood latch recall). I didn't have many complaints about the car, so I decided to poke him about the HW 2.5 failure to exit passing lane issue. He said it's a known issue, and that as soon as Tesla's firmware developments start to take advantage of new hardware features, Tesla basically doesn't look back. This bug was introduced as they began to utilize HW3. He explained that everyone will/can eventually get the HW update, which will eliminate the bug for those who get the upgrade. My understanding was that non-FSD HW 2.5 folks (with or without Enhanced Autopilot) would not get the hardware upgrade. He said that eventually, everyone will be eligible for the HW3 upgrade, regardless of their EAP or FSD status. The implication is that this was basically an IOU for people who bought the car before the HW was fully cooked. He said that Tesla has several priority batches, with the top tier being the early adopters with top trim levels. Some received the HW3 upgrade when it was offered/available to us. But he insisted that even the lowest priority batches (e.g. no autopilot) will eventually be eligible for the free hardware upgrade. Those folks will not get any functionality that they haven't paid for. But they would get full functionality on the features they are eligible for. No timeline commitment, of course.

I know it's hard to be certain when a single individual from the sales/service side makes a broad statement about Tesla's intent. But this long-time Tesla employee was giving me the scoop as he knows it, and I'm just passing it along, FYI.
 
[QUOTE = "drtimhill, post: 4684151, member: 101950"] Rond-point. Vous n'avez pas à vous arrêter du tout lorsque le trafic est faible, ce qui augmente l'efficacité (CO2, etc.) tout en économisant du temps et de l'usure. Ils s'adaptent également eux-mêmes aux changements des modèles de flux de trafic. L'inconvénient est qu'ils sont gros et plus coûteux à déployer. [/ QUOTE]
Oui bien plus couteux.
 
I'm adding one more recent trip to the data set. We ended up driving back to North Carolina one last time (?) for my son's drive-thru graduation ceremony. The ceremony itself was kind of a let-down, but he enjoyed the visit. So I ended up logging another drive in each direction with firmware 2020.20.12 and HW3. Results were solid, and I identified a couple of small improvements.
  • Sometimes NOA only moves out of passing lane when traffic is observed approaching from the rear. I think this might be a new behavior, or perhaps it was masked previously by the HW2.5 passing lane hogging problem. This seems to be a fair balance in terms of driving policy, although it might only be apparent in light traffic situations. In my experience, it always detected the car approaching from the rear, in either the slow lane or fast lane, with sufficient time to get out of the passing lane. For balance, @DirtyT3sla also has examples of this that didn't seem to work as well.
  • Drive 63: Highway exit improvement: NOA exited I-95N @ the I-20E Florence supercharger exit. Previously NOA would signal the intent to exit without executing. Perhaps this can be attributed to HW3, or the GPS stability & tracking improvement first introduced in 2020.20.5. The other isolated issue from post 25 is still present, where NOA maneuvers to exit without apparent reason to do so (I-74N at NC exit 77 Spero Rd).
  • Drive 64: Driver abort was when NOA tried to cut off a heavy truck with insufficient clearance. Not a collision risk, but it surely would have elicited an air-horn.
  • Drive 64: Driver decline was when NOA signaled the intent to merge into the side of a heavy truck. It seemed that aside from confusing everyone including myself, it was likely to cut off the truck as soon as the lane was clear (similar to the cut off and driver abort noted above).
  • Drive 63: Driver decline was when NOA tried to merge into a row of several faster vehicles in the passing lane. At the time, it seemed to promote peace and happiness to simply wait for a few additional cars to pass first.
Thanks to everyone for their interest and comments on this experiment. Here are some concluding statistics:
  • 16,832 NOA miles (est)
  • 4,526 NOA auto lane change events
All together, if I exclude the drives from 2019.40.2.1 until 2020.12 (with the fail to exit passing lane problem):
  • 93.2% success or uncomfortable success
  • 1.0% AP aborted lane changes
  • 1.6% driver aborted lane changes (generally for safety)
While I only have a small data set with 2020.20.12, it's interesting to note that 2020.20.12 had a success rate of 97.4%, showing a considerable improvement with respect to the 93.2% success rate of the overall experiment.

Detailed results:
upload_2020-6-20_15-8-47.png


Weighted score:
upload_2020-6-20_15-9-7.png


Category breakdown with 2020.20.12:
upload_2020-6-20_15-9-35.png


Wishing y'all many miles of safe, stress-free driving.
 
For my part, I think when in the passing lane, the car is extremely sensitive to cars to the right still moving slowly. So to me it's like "what the heck get out of the passing lane, there's a huge gap!" but my car is thinking "I'm going 78, and that guy to the right is going 77, I should pass before exiting passing lane."

Just my guess.
 
Based on what I've seen in your videos and in my car, it does seem to be weighting the speed of vehicles ahead of you rather strongly. If I'm set at 74, and I'm in the passing lane behind (and to the left of) a car going 73, my car will cheerfully wait there despite someone doing 85 coming up from behind me. I'd like to see stronger weighting on approaching vehicles from behind triggering a deferent lane-change.
 
I like the recent replies and I think but do not really know that if the human just nudges the throttle to create enough space for the Tesla to safely change lanes and it does in fact do it after removing the foot from the throttle then it may be possible that you have 'taught' the NOA how to properly execute a passing maneuver. Of course Tesla cannot express this action that breaks the speed limit so they will never admit that this driver temporary throttle intervention may be preferable to them.
 
I'm adding one more recent trip to the data set. We ended up driving back to North Carolina one last time (?) for my son's drive-thru graduation ceremony. The ceremony itself was kind of a let-down, but he enjoyed the visit. So I ended up logging another drive in each direction with firmware 2020.20.12 and HW3. Results were solid, and I identified a couple of small improvements.
  • Sometimes NOA only moves out of passing lane when traffic is observed approaching from the rear. I think this might be a new behavior, or perhaps it was masked previously by the HW2.5 passing lane hogging problem. This seems to be a fair balance in terms of driving policy, although it might only be apparent in light traffic situations. In my experience, it always detected the car approaching from the rear, in either the slow lane or fast lane, with sufficient time to get out of the passing lane. For balance, @DirtyT3sla also has examples of this that didn't seem to work as well.
  • Drive 63: Highway exit improvement: NOA exited I-95N @ the I-20E Florence supercharger exit. Previously NOA would signal the intent to exit without executing. Perhaps this can be attributed to HW3, or the GPS stability & tracking improvement first introduced in 2020.20.5. The other isolated issue from post 25 is still present, where NOA maneuvers to exit without apparent reason to do so (I-74N at NC exit 77 Spero Rd).
  • Drive 64: Driver abort was when NOA tried to cut off a heavy truck with insufficient clearance. Not a collision risk, but it surely would have elicited an air-horn.
  • Drive 64: Driver decline was when NOA signaled the intent to merge into the side of a heavy truck. It seemed that aside from confusing everyone including myself, it was likely to cut off the truck as soon as the lane was clear (similar to the cut off and driver abort noted above).
  • Drive 63: Driver decline was when NOA tried to merge into a row of several faster vehicles in the passing lane. At the time, it seemed to promote peace and happiness to simply wait for a few additional cars to pass first.
Thanks to everyone for their interest and comments on this experiment. Here are some concluding statistics:
  • 16,832 NOA miles (est)
  • 4,526 NOA auto lane change events
All together, if I exclude the drives from 2019.40.2.1 until 2020.12 (with the fail to exit passing lane problem):
  • 93.2% success or uncomfortable success
  • 1.0% AP aborted lane changes
  • 1.6% driver aborted lane changes (generally for safety)
While I only have a small data set with 2020.20.12, it's interesting to note that 2020.20.12 had a success rate of 97.4%, showing a considerable improvement with respect to the 93.2% success rate of the overall experiment.

Detailed results:
View attachment 554040

Weighted score:
View attachment 554045

Category breakdown with 2020.20.12:
View attachment 554047

Wishing y'all many miles of safe, stress-free driving.
Thanks for the one last time test buddy.
 
  • Like
Reactions: Enginerd