Autopilot False Positive Braking

[Blissedout]

Because Tesla hasn't officially confirmed what EAP features are actually working - we are all guessing.

It's possible EAP is only operating with the front camera(s) and radar today - and that, at best, it's only looking at the moving vehicle(s) immediately in front and basing decisions on the changing spacing between vehicles.

Human drivers use more information to make driving decisions - such as emergency vehicle sirens and brake lights - inputs that are probably not being used by AP2 software. A human driver will see traffic slowing down ahead - with brake lights from many vehicles, and be prepared to slow down and stop. If AP2 is just looking at the vehicles immediately ahead and ignoring the brake lights, the software is likely to react later than a human driver.

From a safety perspective, it would help considerably if Tesla could tell us what is and isn't working in EAP - so the driver is prepared to handle situations not currently supported in the software.

Autopilot 2.0 appears to be using the three front facing cameras. See

 

[Lectrik]

After a braking event I previously described (slowing of a car in the exit lane) I've developed a keener eye for this instance and can hold the wheel left in anticipation of the brakes coming on. Slight acceleration helps too.

 

[Buster1]

UPDATE:

After a third braking/deceleration event, Tesla Engineering dug into my events through my local SC manager. I even sent in detailed reports with times and last/longs etc.

Got a final reply from my SC manager last night...
“We have a response to the braking issue. The sensors are not the issue. The radar is. It is located behind the bumper and it can be affected by the extra thickness added to the bumper by the clear bra and wraps. Autopilot function is not guaranteed when aftermarket wrap or clear bra is installed.”

Guess it’s my clearbra.

 

[The Duke]

I am happy to hear they did not charge you for a problem you caused.

My friend has aftermarket wheels on his AMG and MB is refusing to deal with a tire wear problem unless he pays them to find it. I am pretty sure he nicked a curb and the car is out of alignment but he is convinced MB should be taking care of it for free. So his car sits while they fight.

 

[rsg123]

My 2014 S had random unexpected braking with autopilot enabled in the past week, although it was pretty moderate both times. I chalked it up to having a bunch of dead bugs on the windshield/front end, but my car also has the clear paint protector in the front.

 

[Doug_G]

I experienced a false braking situation today, which could have been dangerous had the car following me been closer.

Driving on a four lane divided highway on autopilot. There was a vehicle stopped well off the highway to the right, with its brake lights on.

I always give a wide berth. I used the turn blinker to instruct the autopilot to move to the left lane, which it did. Then, as I was about to pass the stopped car, which was now far off to the right, the autopilot slammed the brakes.

I immediately overrode that with the accelerator, but that took a moment. I'm really glad the car behind me was not close, because they might have slammed into my back.

Rather inexplicable. I have noticed that the cruise control will brake when it spots a flashing red light off to the side (e.g. I've had it hit the brakes when one of those speed warning signs stared flashing - quite repeatable); but there was no flashing light here and the stopped vehicle was way off to the right.

 

[GatorGuy]

UPDATE:

After a third braking/deceleration event, Tesla Engineering dug into my events through my local SC manager. I even sent in detailed reports with times and last/longs etc.

Got a final reply from my SC manager last night...
“We have a response to the braking issue. The sensors are not the issue. The radar is. It is located behind the bumper and it can be affected by the extra thickness added to the bumper by the clear bra and wraps. Autopilot function is not guaranteed when aftermarket wrap or clear bra is installed.”

Guess it’s my clearbra.

Find that verrrry hard to believe. Radar has no problem going through plastic. Now if he had said it was the ultrasound sensors I would have believed that.

 

[bandido]

This is what drives me crazy. I love Tesla, but they continue to ignore REAL issues. I was driving in a new 2018 loaner, latest-esque' software-(2018.16.1 cbe3dea), new MCU, etc.. At 75mph on large open highway, clear sunny day, and SLAMMS brakes for about a 1/2 second. Absolutely would have been rear ended if anyone was behind me. My 2015 AP1 has been solid with no weird autopilot behavior. I was just telling several friends that I think AP2 has made parity with AP1 and then this happens.

Can someone tell Tesla to get a mediator and grovel back into bed with MobilEye. There is just no way they are ever going to get self driving without some serious help...

 

[chillaban]

Find that verrrry hard to believe. Radar has no problem going through plastic. Now if he had said it was the ultrasound sensors I would have believed that.

Note that shaped plastics can act as lenses and warp the directionality of a radar beam, which is very important for things like radar braking which whitelists/blacklists objects based off their signal strength and coarse location.

 

[GatorGuy]

Just to add my experience, my whole car is wrapped including all ultrasound sensors for two weeks. I have had zero problems parking or with radar braking.

 

[pedriscoll]

My June 2015 85D with AP1 never had false positive braking events for the first 2 years of AP use. About 6-9 months ago I began to experience these false positive braking events. Overall I feel that AP1 is less reliable than it once was. I now know that I need to remain ever vigilant for this type of event. AP is still better than any other driver assistance system on the market, but the driver must always be ready to take over instantly.

 

[Buster1]

I’m starting to think it’s software. Seems to be a lot of false braking events for all of us over the past many months. As of the last 3-4 weeks and new software, I’ve had no AP braking events. Knock on wood.

 

[bob_p]

While there have been improvements with AP2, I still consider it as reliable as a student driver - it's going to make mistakes and you have to be ready to take over control when it does.

On some surface streets, there are several areas where TACC will automatically slow down briefly before resuming the set speed. Last year this happened periodically when going under large highway signs or bridges, now it seems to be doing it if there is a quick coloration change in the road - going from light to dark. Evidently the software is misinterpreting this as a potential object.

The software is also to conservative when vehicles are crossing the road in front of your, or when a car in front turns into the adjacent lane to turn at an intersection, causing the car to slow unnecessarily.

These are inconvenient, but not a huge concern.

What is a HUGE problem is when the software gets confused at highway speeds (65-70 MPH) and decides to slow down, either because AP thinks there may be an obstacle ahead OR because the speed limit database believes the speed limit is considerably below what is posted.

Two nights ago, driving home late at night, in heavy traffic - all moving at 65+ MPH, hit a spot where the speed limit database thought the highway speed dropped from 60 to 20 MPH - and started quickly slowing down - in the middle of highway traffic. As expected, this resulted in the car behind me closing the gap quickly - and honking their horn at the unexpected AP response.

Treating TACC/AP as a student driver, because I monitor the above situations - as long as I'm paying attention, I can correct this by applying pressure to the accelerator and getting back to the desired speed before TACC/AP is ready to drive at the set speed.

The software is still learning - so we should expect this type of behavior. Though the numerous errors in the speed limit database really need to be fixed. Prior to the change over to the new speed limit provider last summer, the speed limits were relatively correct. Now, there are many more areas missing any speed limit (which prevents AP from running) or with incorrect speed limits (which causes TACC to slow down or improperly set the default speed too high) - and this needs to be fixed...

 

[ai4px]

I understand it not being able to take a turn right or not dedecting a stationary object on time. That is why we are still responsible on Autopilot. However slamming the brakes hard in stop&go traffic, reducing my reaction time to correct Autopilot to virtually zero is something I haven't read anyone report before.

My wife drives a 2016 honda civic with this emergency brake issue. Once we had a car turn left in front of us and that caused the brakes to activate. There was no real issue, but it surprised our little honda and it locked up the brakes. The bigger issue was that while in e-brake mode, it was dead pedal. As I recall, it came to a complete stop on a 45mph road and no amount of pressing on the gas would make it go. We nearly got rear ended by the poor guy following us. The car that had turned across in front of us was long gone by the time we were able to move again. To me, the dead gas pedal is the worst part! Occasionally on a right hand curve in the road, the radar will "see" an oncoming car and false trigger the brakes as well.

I was hoping the tesla system was better!

 

[geirrosset]

I just need to chime in here. I have experienced false positive braking on all my longer trips (>100km) since getting the Model X at the end of March. About 10 times. I have been unable to spot what caused these events, but at least three of them have been upon entering or exiting tunnels, which leads me to suspect altered lighting conditions as the culprit. If that's the case it is rather disconcerting as being rear ended is but a large bush by the side of the highway away... I believe the car had 2018.14.2 when I took delivery, then upgraded to 2018.16 and now I'm on 2018.18.X. I have not seen any improvement on this issue yet. The problem here isn't just that I trust AP even less, it's the fact that so far 100% of the times it has done this it has been without reason. What happens when I reach emergency braking incident nr. 100 and the previous 99 have been false positives? I will have developed an automatic response where I step on the accelerator, only this time there might actually be something there...

 

[SSonnentag]

Don't we have the option of turning off AEB? It might be a feasible option when in heavy traffic if you're getting a lot of false detections.

 

[SSonnentag]

When I was at the shop getting my S XPELed, I asked about the sensors and they said always cutout holes for them. Apparently it's a known issue.

 

[AWDtsla]

Don't we have the option of turning off AEB? It might be a feasible option when in heavy traffic if you're getting a lot of false detections.

If you are driving yourself, throttle inputs override AEB. If you are not driving yourself, it does not matter.

 

[SSonnentag]

If you are driving yourself, throttle inputs override AEB. If you are not driving yourself, it does not matter.

I know, but with AEB off you can continue to use TACC and AP.

 

[Enginerd]

When I was at the shop getting my S XPELed, I asked about the sensors and they said always cutout holes for them. Apparently it's a known issue.

I think they're talking about cutting out holes for the ultrasonic sensors (easily identified little circles on the bumper). However, AEB is probably more driven by the radar. This is my concern with getting the front end wrapped. Has anyone avoided wrapping the front radar? Or regretted wrapping over it? Tesla has a generic universal disclaimer something like: if anything goes wrong, it's because you wrapped over the sensors.