-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Phantom braking is the biggest issue with AutoPilot.
- Thread starter WilliamG
- Start date
A couple days ago I used TACC pretty much all the way to Canada without much issue.
I can't say I remember having any significant false braking events on the way up, and my only thought was I wish it would do a better job smoothing out the slowing. There was someone in front of me who kept hitting their brakes instead of backing off a bit. So that was a bit annoying until the person got out of the left lane. There was a quite a bit of traffic so I didn't bother using NoA.
Then on the way home I used NoA from the the US border all the way down to my exit near Everett, and had quite a few annoyances.
There wasn't much traffic to speak of so it was the perfect opportunity to test it in non-challenging conditions.
Here are the areas it completely failed
1.) It would hunt for the middle during the merge points in the right lane. There were times where it would get confused, and would actually brake. On a few of these it was bad enough that autopilot canceled either due to its own confusion or me going "wtf" and holding the wheel more to prevent it from going so far over to clip the exit lines.
2.) Excessive braking when the speed differential between me, and the car next to me was higher than it wanted. This was in situations where I was in the left lane, and people were merging into the lane to the right of me. I could understand, and accept a little slowing but it was way higher than expected.
I didn't have any issue with overall smoothness as there was hardly anyone in front of me.
I didn't have any false braking due to overpasses.
I didn't have any issues with canceled lane changes. It typically got into the passing lane before any significant slowing, and got back over after passing. in only one case did it slow considerably before getting over, and I have no explanation as to why. There wasn't anyone in the left lane at the time.
All in all it was both pretty awesome, but also a complete failure.
Once I got to the point where there were 3 lanes then it was smooth sailing in the middle lane.
On my exit it didn't handle it exactly like I would have as a semi-truck was in the right most lane, and the car initially wanted to pass it. Then decided it was best to get behind it. That was fine, and that part is a pass despite not being what I would have done.
Part of me wants to document the test along with the expectation for a passing grade to see when/if any firmware version is capable of passing the test. To this date no version of NoA has passed any of my testing.
I can't say I remember having any significant false braking events on the way up, and my only thought was I wish it would do a better job smoothing out the slowing. There was someone in front of me who kept hitting their brakes instead of backing off a bit. So that was a bit annoying until the person got out of the left lane. There was a quite a bit of traffic so I didn't bother using NoA.
Then on the way home I used NoA from the the US border all the way down to my exit near Everett, and had quite a few annoyances.
There wasn't much traffic to speak of so it was the perfect opportunity to test it in non-challenging conditions.
Here are the areas it completely failed
1.) It would hunt for the middle during the merge points in the right lane. There were times where it would get confused, and would actually brake. On a few of these it was bad enough that autopilot canceled either due to its own confusion or me going "wtf" and holding the wheel more to prevent it from going so far over to clip the exit lines.
2.) Excessive braking when the speed differential between me, and the car next to me was higher than it wanted. This was in situations where I was in the left lane, and people were merging into the lane to the right of me. I could understand, and accept a little slowing but it was way higher than expected.
I didn't have any issue with overall smoothness as there was hardly anyone in front of me.
I didn't have any false braking due to overpasses.
I didn't have any issues with canceled lane changes. It typically got into the passing lane before any significant slowing, and got back over after passing. in only one case did it slow considerably before getting over, and I have no explanation as to why. There wasn't anyone in the left lane at the time.
All in all it was both pretty awesome, but also a complete failure.
Once I got to the point where there were 3 lanes then it was smooth sailing in the middle lane.
On my exit it didn't handle it exactly like I would have as a semi-truck was in the right most lane, and the car initially wanted to pass it. Then decided it was best to get behind it. That was fine, and that part is a pass despite not being what I would have done.
Part of me wants to document the test along with the expectation for a passing grade to see when/if any firmware version is capable of passing the test. To this date no version of NoA has passed any of my testing.
I had a mobile service appointment today that took me to Kamloops. It's a route I don't take often and normally wouldn't use AP on, but I used to use traditional cruise in my other vehicles. The route has fairly low traffic, with an oddly high number of government vehicles for whatever reason and a few semis. For science, I tried to use AP as much as possible and recorded events until my USB drive stopped working.
I only primarily used AP on the way there (I got sick of the anxiety, read on). On that ~120km route, I had 5 "phantom" braking events. 1 was minor enough that I didn't care. 3 were around leftward corners with a semi in the opposing traffic (left side). 1 was a very hard 15 km/h drop with seemingly no explanation. Numerous other braking events were expected and warranted, if a bit over-reacting (e.g. sharper corners coming up). I eventually stopped using AP because I couldn't tell if the vehicle behind me was a cop or not, and I didn't want to explain the reason for my erratic driving was evaluating the limits of a driver assist system.
It should be noted especially for the semi cases that AP in my car (and apparently many others?) hugs the left, which may have made it sensitive to the conditions.
I only primarily used AP on the way there (I got sick of the anxiety, read on). On that ~120km route, I had 5 "phantom" braking events. 1 was minor enough that I didn't care. 3 were around leftward corners with a semi in the opposing traffic (left side). 1 was a very hard 15 km/h drop with seemingly no explanation. Numerous other braking events were expected and warranted, if a bit over-reacting (e.g. sharper corners coming up). I eventually stopped using AP because I couldn't tell if the vehicle behind me was a cop or not, and I didn't want to explain the reason for my erratic driving was evaluating the limits of a driver assist system.
It should be noted especially for the semi cases that AP in my car (and apparently many others?) hugs the left, which may have made it sensitive to the conditions.
SalisburySam
Active Member
I’ve submitted a bug report each and every time since this has first happened...literally dozens if not hundreds of times.
My own experience and others is that these are fixed. What is exact locations of the most common ones?
ljbad4life
Member
One thing to add is make sure your navigation data is up to date that is separate from the software updates. My car navigation data got stuck on an old build for a while and it made the car phantom brake a lot on a very specific spot on the LIE. After a navigation update, it went away (there was construction in that one spot for a long time and it was finally finished).
T3slaOwner
Member Extraordinaire
ljbad4life
Member
I went in to have Tesla rotate my tires, they updated everything while it was there. There does appear that navigation updates appear in the same place now that the regular software updates do in the menu. Perhaps just clicking on it will force it.
T3slaOwner
Member Extraordinaire
T3slaOwner
Member Extraordinaire
I will take a look. Thanks.
I have seen construction changes impact navigation for months after the work was done, in some cases the speed limit needed to be fixed from 25 (actually on a temporary side road) to 55 or higher on the new road. It is very disconcerting when the car brakes hard on a highway. It is potentially also dangerous. If someone were following too close behind me and I intentionally hit the brake that way, I'm pretty sure that would be illegal and could open me to responsibility for any resulting accident and damage. I don't know how I can be responsible for that when there is no way for me to prevent the car doing it.
ljbad4life
Member
I think it was $75 for tire rotation and balancing, but my tires were in really good condition that they said I didn't need to rotate them.
T3slaOwner
Member Extraordinaire
I had a vibration complaint and they wanted well over $100 to rotate the tires. Maybe they don't like me. I have an appointment this morning. Not sure I'm going in. The condition I called it in for has stopped.
ljbad4life
Member
Perhaps because you mentioned a vibration issue that they quoted an higher price.
On topic: I think you are right construction lingers in the navigation software for a long time. There was an S curve in that specific spot that required a significant slow down before. Now that it's gone, the highway is straight No need to slow down.
T3slaOwner
Member Extraordinaire
They also suck at inserting temporary changes. There was a rock slide in the mountains of North Carolina I was stuck in. Had to spend the night on the wrong side of the mountain. Next day the navigator wanted to route me through the rock slide still. Took them nearly a month to get the road clear. I want to say it was a week getting the blockage to show up in the navigator. I believe they use Google map data, but don't always have the same info at the same time. Maybe Tesla has their own copy of the data and their own servers so they can be a little behind.
A similar issue was a friend's house that showed up on the wrong place because of the road having been cut in two by a highway some 30 years ago. I guess Google never quite got it right, and I would need to use a different number to get routed to the right location in the Tesla. Even UPS had trouble finding their house. A recent package came with a print out of a Google map and highlighter showing the location for the driver. Not long after that the Google map started showing their address at their house and now so does Tesla.
Burrito
Member
i agree phantom braking is the worst part of AP i've experienced. im in NJ and people follow entirely too close and i'm paranoid that while on AP i will get rear ended
Try decreasing your Speed Limit Offset 1 or 2 MPH. I have mine set at 3 mph and have found this to decrease the # of false front collision alerts (based on my driving behavior). It might also decrease the likelihood of phantom braking since it's not been a problem for me.
Of course, it will require you to be more attentive since you've decreased your buffer, so it's a trade-off.
Of course, it will require you to be more attentive since you've decreased your buffer, so it's a trade-off.
T3slaOwner
Member Extraordinaire
I dropped off my X with the service center today and had a couple/three false breakings both on the back roads and the highway both in my car with up to date software and the S85 loaner on the way home. The S is much more the sports car. I would have loved to have bought one of those, but I'm 6 foot tall and my head and shoulders hit the door frame getting in and out.
DopeGhoti
Active Member
It's interesting reading that, because the 3 is smaller, I am of like height, and I haven't any issues with ingress or egress.
T3slaOwner
Member Extraordinaire
I had the same problem getting into the 3 only worse. I have to enter rear first while in my X I can step in with my foot and slide across the seat. I don't have any trouble with most other cars (I haven't tried a sports car lately). It is clearly due to the steep rake of the windscreen from keeping the wind resistance coefficient low. That same steep rake makes the post very, very wide and the angle makes them even more an obstruction to visibility. I've turned at city intersections where I knew there was a pedestrian at the curb, but could not see them while turning.
Similar threads
