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

How Satisfied Are You With Autopilot?

How Satisfied Are You With Autopilot?

  • Very Unsatisfied

    Votes: 18 7.2%
  • Unsatisfied

    Votes: 8 3.2%
  • Neutral

    Votes: 16 6.4%
  • Satisfied

    Votes: 80 32.1%
  • Very Satisfied

    Votes: 127 51.0%

  • Total voters
    249
This site may earn commission on affiliate links.
I thought I read 85mph somewhere. But any faster and the sensors just can't keep up. Happens with BlindSpot sometimes, too, at high speeds

If I'm not mistaken TACC is 90mph. The blind spot detection turns off at either 80mph or maybe it was 85mph (someone here says 88mph).

- - - Updated - - -

Has anyone found a way to keep your hands on the wheel while in AP that 1) didn't cause AP to disengage and 2)didn't make your arms tired supporting the weight of your arms mid air since you can't place much weight on the wheel without causing condition 1)?

I keep my elbow on the drivers door arm rest. THat's also the way I normally drive.
 
All I see here is the car making what looks to be a slight but sudden centering correction to the right while entering a gradual left turn, causing the meat servo to violently overcorrect to the left. Gentle inputs, people... gentle inputs.

What I see is that the lane markings suddenly end, and the Autosteer can't find the road. Lane markings and stripes are informational to our brains. We can still see where the road is and adapt. To the AS, lane markings are a vital, necessary controlling factor in being able to work!
 
Found the blindspot official numbers "Blind Spot Warning is active when Model S is travelling between 20 mph (30 km/h) and approximately 85 mph (140 km/h)." (Page 68 owners manual)

Can't find the official TACC numbers
 
What I see is that the lane markings suddenly end, and the Autosteer can't find the road. Lane markings and stripes are informational to our brains. We can still see where the road is and adapt. To the AS, lane markings are a vital, necessary controlling factor in being able to work!
The lines to the left transition into detectable Armco barrier and the lines to the right remain until after the human overcorrection is made. A curb appears to the distant right of the lane markings (eventually bending in to replace the markings, but that point is after the correction). It's possible that a centering correction was being made by Autopilot to more evenly space the car between the Armco barrier and the distant curb that appeared (both hypothetically detectable by the radar/ultrasonic systems), as I did not notice any alert from Autopilot indicating that the driver should intervene. The system also had a lead car in view the entire time.

All of that is beside my point, though... regardless of the cause, the Autopilot system made a very small corrective input to steering angle (in error or otherwise) and the response from the human was sudden overcorrection. My concern here is with people overreacting to the system doing anything unexpected, regardless of whether the system is operating nominally or not. As the fleet is learning and improving, it's possible that people reacting to the system will cause more accidents than the system itself.
 
You are using highly inflammatory and inaccurate language in your post. The Autopilot software had no intent to harm you. You are attributing human malevolence to software code.
On a two lane road with a lot of curves and no physical divider between your lane and oncoming traffic you had Autopilot engaged even though it is clear based on statements from Tesla that is not the intended use case. I am glad you had your hands close to the wheel and were able to take control but you should not be using Autopilot on roads like that.


Ouch! @ecarfan Its not my video or my car, I am just the messenger. This was posted on reddit. I agree with your points though.
 
Has anyone found a way to keep your hands on the wheel while in AP that 1) didn't cause AP to disengage and 2)didn't make your arms tired supporting the weight of your arms mid air since you can't place much weight on the wheel without causing condition 1)?

I tried putting my hands on my knees with one thumb gentry resting on the bottom of the steering wheel. That seems to usually be enough to keep the Nagging Nanny at bay. And my intuition is that if I had to take over quickly I would be OK.

The key to auto steering at this point is to be very vigilant and aware of when the auto steering may have trouble. We will all get better at this with practice. (Then Tesla will release an upgrade and we'll have to learn all over again :)

The danger is that it works well enough that one can get away with not paying attention for long periods. Sooner or later someone will be killed for that reason. The question is whether that will happen more or less often than it would have without autopilot.
 
I wonder what % of the "very satisfied" live in California, where the majority of development and testing was done. In my area, the car has serious problems identifying some types of highway lane markings, which is why I voted neutral. I'm sure it will improve.

I also wonder what % of lane recognition is developed in-house by Tesla vs. built in to Mobileye.
 
Folks, you cant fix stupid...so dont try to make Tesla take away awesome new features if some people are too stupid to use them in a reasonable way...this couls easily parallel the gun control argument. If you think auto pilot tried to kill you you need to turn the feature off and go away...stop pissing on the parade.

Tesla Autopilot is not perfect I think it's easy to agree on that. Disabling Autpilot for corner cases, such as secondary highways, would be the best way to improve it until better technology and logic has been developed. Doing so would be better for Tesla investors, Tesla drivers, insurance companies - not to mention everyone else on the road.
 
Tesla Autopilot is not perfect I think it's easy to agree on that. Disabling Autpilot for corner cases, such as secondary highways, would be the best way to improve it until better technology and logic has been developed. Doing so would be better for Tesla investors, Tesla drivers, insurance companies - not to mention everyone else on the road.

I agree with the first sentence, I disagree with everything else you wrote.

- - - Updated - - -

The speed limit is supposed to be the maximum safe speed under ideal conditions....
...where "ideal" means a blind quadriplegic driving an overloaded 1957 school bus in a snow storm.

You are absolutely right, I was thinking road designs. Road designs are designed with buffer for inclement weather at the given speed limit plus speeding plus some other nonsense, IIRC.
 
I found a position that is comfortable with my hands "on" the wheel but not so heavy handed (as I've been driving my adult life) that causes the AP to disengage.
i try and see some of the limits of the system, as others appear to do also, so I was trying to find something similar to the "tesla tried to kill me" video. What I seemed to notice is that roads with the double yellow center lane, especially in bright sun and shady sections, the AP tended to lose the yellow easier than I would have thought visually. It then would see the white across the opposite lane, then tend to steer to between the two white lines (I.e. Crossing over the double yellow). Then it would again pick up the double yellow and correct back into my lane. Obviously, this isn't desired, but the system may not have as much color deciphering as yellow lanes are not typical on divided highways / interstates (the intended realm of the AP beta).

anyone else see anything like that's?
 
I found a position that is comfortable with my hands "on" the wheel but not so heavy handed (as I've been driving my adult life) that causes the AP to disengage.
i try and see some of the limits of the system, as others appear to do also, so I was trying to find something similar to the "tesla tried to kill me" video. What I seemed to notice is that roads with the double yellow center lane, especially in bright sun and shady sections, the AP tended to lose the yellow easier than I would have thought visually. It then would see the white across the opposite lane, then tend to steer to between the two white lines (I.e. Crossing over the double yellow). Then it would again pick up the double yellow and correct back into my lane. Obviously, this isn't desired, but the system may not have as much color deciphering as yellow lanes are not typical on divided highways / interstates (the intended realm of the AP beta).

anyone else see anything like that's?

That makes sense, I had something happen to me where it tried to go against traffic, through a double yellow line, if it was trying to center, that would explain it
 
The speed limit is supposed to be the maximum safe speed under ideal conditions....
...where "ideal" means a blind quadriplegic driving an overloaded 1957 school bus in a snow storm.

yup. Speed and safety haven't necessarily been related since that era either. They removed speed limits in Montana for a time and traffic accident rates DECLINED. but then again, so did enforcement revenue.

Im treading dangerously close to moving off the thread topic, but let me bring it back - speed doesn't kill stupid drivers do - same with this AP. The first accident will be caused by some idiot driving through a residential area and blowing through a stop sign and creating an accident. It will be his/her fault and the tools will be blamed by the driver.
 
A better question for me: is it worth the $2500/3000? I'm configuring and trying to decide if it is worth it. I use simple cruise control sometimes. Haven't felt a serious need for all the above, although it looks like fun to play with.
 
Depends on what kind of person you are. I am a techie and I wouldn't consider a tesla without an autopilot - at all. I have always liked the cutting edge (sometimes the bleeding edge) of technology, and I am aware of the pitfalls. I use the autopilot, and use it cautiously. I take the word "Beta" seriously and don't treat a beta release as if it is production grade fully functioning, will-never-make-a-mistake system. I even expect a production version software to make a mistake. As a human, I am good at judging how far to let the autopilot go, and when to take over.

But then there are my flight students who "try to kill me" - and themselves - every time we go flying. Then they learn how to fly, take their children and family up (along with other 300 or so people) and make good decisions and become safe pilots. I am treating the autopilot the same way - it will learn and become better.

My advice - go for it!
 
What I seemed to notice is that roads with the double yellow center lane, especially in bright sun and shady sections, the AP tended to lose the yellow easier than I would have thought visually. It then would see the white across the opposite lane, then tend to steer to between the two white lines (I.e. Crossing over the double yellow). Then it would again pick up the double yellow and correct back into my lane. Obviously, this isn't desired, but the system may not have as much color deciphering as yellow lanes are not typical on divided highways / interstates (the intended realm of the AP beta).
Don't know whether the software has a problem picking up yellow lines, but it's a little strange if true. Don't the interstates have a solid yellow line on the median side?
 
Depends on what kind of person you are. I am a techie and I wouldn't consider a tesla without an autopilot - at all. I have always liked the cutting edge (sometimes the bleeding edge) of technology, and I am aware of the pitfalls. I use the autopilot, and use it cautiously. I take the word "Beta" seriously and don't treat a beta release as if it is production grade fully functioning, will-never-make-a-mistake system. I even expect a production version software to make a mistake. As a human, I am good at judging how far to let the autopilot go, and when to take over.

But then there are my flight students who "try to kill me" - and themselves - every time we go flying. Then they learn how to fly, take their children and family up (along with other 300 or so people) and make good decisions and become safe pilots. I am treating the autopilot the same way - it will learn and become better.

My advice - go for it!
i agree with every word! I recall doing a LRJET type rating for a well educated articulate MD who tried to kill me on his check ride V1 cut. I fear that people like him will assume "beta" is meaningless and foment disaster. There are people on this forum who equate "autopilot" with "autonomous". If it were not beta it would still be an autopilot rather than autonomous driving.

I'm thrilled with my 500 miles do far with AP. I'm happier as I'm learning it 's limitations. This is complex technology. It deserves respect.
 
First chance to try it today. 100 mile drive on I-70 heading east into Denver, half before dawn and half just after sunrise driving into the sun. I disengaged it through the construction zones. Several times it tried to take an exit to the right and I had to intervene. There was a couple of times where it made a sudden jerky swerve out of the current lane with no warning and I had to intervene. I'm not sure if it was the sun glare or what caused this. It seemed to work best when we were going slower in the heavy traffic through Denver. I hope it improves quickly with crowdsourcing learning, but am hesitant to really use it at this point.