Autopilot 2.0 is dangerous

[Tribaltech]

Hi everyone,

So I understand that AP 2.0 is still "beta" which is kind of frustrating to begin with because I purchased my AP 2.0 Model S in December and its almost May and I don't see any light at the end of the tunnel when a fully rolled out version will become available. Anyway now that I have gotten that out of the way I want to state that AP 2.0 with its TACC is downright dangerous in a ton of day-to-day situations. And here's how:

1) When I turn on AP 2.0 while driving on a highway the car swerves like a bloody drunk person sometimes veering so close to the vehicles in adjacent lanes that I have no option but to take over. Its just friggin dangerous?

2) At other instances the car chooses to slam on the brakes while going on curves at highway speeds. Like without any rhyme or reason it'll literally brake so hard for a second that it's enough to make an un-belted passenger fly out of the windscreen.

3) At certain occasions I have experienced that going under bridges and overpasses it momentarily again slams the brakes. Like what the heck Tesla? What the heck!?

4) Lane changing is another horror to tackle altogether. While driving in AP 2.0 with autosteer when I turn on the indicator to change lanes and when it has executed the lane change and I switch of the indicator, the bloody car just keeps on going till I have to take over, otherwise it would fly into the shoulder.

All of these dangerous situations considered I feel its a miracle my car hasn't slammed into another vehicle or that someone else hasn't rear-ended me terribly by now. I am so disappointed with this useless and irresponsible roll out of AP 2.0. If Tesla cannot ensure its technology to work safely then why roll it out at all? Why don't just test it at your own facilities and roll it out only when you're quadruply sure. Until then just give us regular old non-TA Cruise Control. But to roll out half baked "features" like this that can cause accidents is downright awful.

My question to everyone on the forum are the following

1) Have you experienced such scares with AP 2.0?
2) Is there a way to engage cruise control without any of the radar and ultrasonic bs and just activate regular plain-Jane cruise control? Asking because I am sick and tired of this extra "Intelligent" car that keeps scaring the crap out of me and my family at highway speeds.
3) Any idea as to when AP 2.0 will be rolled out non-beta without any glitches?

Thanks in advance for your responses

 

[Krypto Kat]

All your observations are valid.
I would propose a feature that would activate the lane keep assist, while leaving the gas and brake controls to me. Would take a lot of fatigue out of driving and still keep it safe. I speak with experience in my Acura which has a very smooth lane keep assist feature (you just have to keep your hands on the wheel but it does the steering for you). It doesn't ping pong between the lines and does a phenomenal job of keeping the car centered in the lane, and since I control the brake and gas pedals I end up with a much smoother drive than in my Tesla. Not to mention I stay attentive to the road and don't panic in situations that trip up every intelligent adaptive cruise control system out there.
Tesla could easily do that via OTA update.

 

[Patrick W]

When I read reports like this I find myself wondering if my HW2 S is somehow special.

Probably not but while I won't call it perfect and not yet up to the HW1 S I traded for the one I have now, I've yet to encounter (after 5 months of driving) any of the events listed above with the exception that before the latest update I did have a couple of instances where the car slowed suddenly when approaching an overhead sign. But even those instances lasted less than a second.

Just last evening I spent over an hour freeway driving and the car stayed in the lane, slowed when coming up on other vehicles and changed lanes as well as my HW1 S did.

I read a while back that some HW2 cars came off the line with misaligned sensors that needed to be adjusted by service. Perhaps that's the cause of some of the negative reports I've read.

And, as for beta, boy, when my HW1 S first got auto pilot it really was a "thrill" ride at first. But it learned. HW2 started off ok and is only getting better.

Personally I've no problem with being a beta tester, knowing that I'm helping to improve the line. With tongue partially in cheek I've said before that better for Tesla to put cars on the road that drivers know are in beta rather than companies like Microsoft that issue stuff they claim to be finished product only to find it is not.

 

[Burnt Toast]

I feel there is a lot of hyperbole in your post. I don't think the brakes alone can generate enough force to fling an unbelted passenger through the windscreen. I also don't believe it applies brakes when passing under signs/bridges, merely stops accelerating (and then begins a momentary regen).

 

[alfagiulia]

I don't understand the fascination with autopilot. I would love to hear why people are so excited about it. I think it is gimmicky and not particularly useful.

It's definitely an interesting software development problem. And there's a niche for it maybe in shipping, taxis, the elderly/blind etc. I also think auto safety is important, but there are many other ways to improve safety. It will also be a very long time before you can fully relax and trust that an AP car won't harm you - even if it's statistically safer than the average human driver. I like the sensation of driving and I am much more comfortable being in control of the vehicle. The BEST thing about Tesla is that it is fun to drive. The movement towards autonomy really dampens that experience.

Does anyone else worry about so many resources being poured into autonomous driving at the expense of development of the rest of the company, e.g., superchargers, service centers, gigafactories, manufacturing innovations, battery development, other mechanical innovations, new products etc...? The other big car makers - VW group, Ford, Mercedes, GM, the Koreans, and the Japanese will eventually catch up to Tesla in terms on the cars themselves if Tesla does not keep innovating. I don't want a robot car. I want a unique, driver's car. Tesla won't be unique if it doesn't stay ahead of the curve in terms of development of the actual car.

 

[alfagiulia]

Tesla is really banking on autonomous driving with model 3. But what if people don't really want that? Then they've wasted a ton of resources in a feature that doesn't inspire the masses... they've blown an opportunity to make the model 3 the best non-luxury car ever built. That would be a shame. I hope Tesla hasn't compromised the model 3 too much to develop the AP hardware and software.

 

[Krypto Kat]

I think it is gimmicky and not particularly useful

For now yes; but the dreamers among us hope that it won't be so, for long.

 

[nomeus]

Tesla is really banking on autonomous driving with model 3. But what if people don't really want that?

Why wouldn't anyone want it? It's like saying why would anyone want an automatic gear box? I for one look forward to autonomy. Now I waste so much time driving, why wouldn't I want to do something else (like work) while the car drives me to my destination? I would take autonomy over manual driving any day.

 

[ehabich]

When driving along the german Autobahn I get the occasional swerve. We have a November 2016 Tesla Model S 75. Yesterday it aimed straight for the side of a truck whilst driving at 130km/h. On a straight stretch of road! The problem was possibly bad lane markings, which are very common in Germany.

Since I'm aware of the dangers of the current system I was able to recover the faulty driving yesterday again.

I'm wondering if those instances should be recorded manually by me or us, and sent to Tesla.

On the other hand, Tesla ought to be able to track each instance when a user has to take control away from AP2,
and the video should be sent to Tesla headquarters automatically, for analyzing the problems...

Or would such an "automatic trouble reporting system" too advanced?

 

[alfagiulia]

For now yes; but the dreamers among us hope that it won't be so, for long.

But what is the dream exactly?
That your car will drive you around like a Taxi?

 

[alfagiulia]

Why wouldn't anyone want it? It's like saying why would anyone want an automatic gear box? I for one look forward to autonomy. Now I waste so much time driving, why wouldn't I want to do something else (like work) while the car drives me to my destination? I would take autonomy over manual driving any day.

That's a really good example. An automatic transmission is a little bit easier to drive and gives the driver a little bit less control. I lament the fact that it's getting increasingly hard to find cars with manual transmissions, even in Europe. Automatic transmissions are BORING.

If you don't really like driving, why would you buy a Tesla? There are plenty of more comfortable cars and cheaper cars, and there always will be. Or why don't you take a train, bus, or Taxi? I ENJOY driving. For me it's a source of PLEASURE. I get a thrill from manipulating the controls and feeling the acceleration.

 

[alfagiulia]

When driving along the german Autobahn I get the occasional swerve. We have a November 2016 Tesla Model S 75. Yesterday it aimed straight for the side of a truck whilst driving at 130km/h. On a straight stretch of road! The problem was possibly bad lane markings, which are very common in Germany.

Since I'm aware of the dangers of the current system I was able to recover the faulty driving yesterday again.

I'm wondering if those instances should be recorded manually by me or us, and sent to Tesla.

On the other hand, Tesla ought to be able to track each instance when a user has to take control away from AP2,
and the video should be sent to Tesla headquarters automatically, for analyzing the problems...

Or would such an "automatic trouble reporting system" too advanced?

I've never noticed bad lane markings on an Autobahn. Although the construction lane markings can be very confusing because they just put down orange markings over top of the original white markings. But many other roads completely lack lane markings of any kind... including 2-way 100kph limited roads.

 

[J1mbo]

I'm wondering if those instances should be recorded manually by me or us, and sent to Tesla.

On the other hand, Tesla ought to be able to track each instance when a user has to take control away from AP2,
and the video should be sent to Tesla headquarters automatically, for analyzing the problems...

Or would such an "automatic trouble reporting system" too advanced?

Pretty sure that Tesla would want to collect info on each "disconnect" as part of their "fleet learning". Make sure you have Data Sharing turned on (seems to be off by default).

For me, the biggest problem I have with TACC is where to rest my right foot?

Lane holding is also OK. I have 17.16.35 and find it to be more confident than the previous FW at motorway speeds.

The worse AP scenario that I have personally experienced (on an older FW) was when changing from Lane 3 to Lane 2 on a slight incline with a bridge perhaps a 1/4 mile ahead, and a truck slightly ahead of me in Lane 1. In that case, the car slowed down much more than necessary as it moved into Lane 2, which would have been unexpected for a driver behind me in Lane 3.

But honestly, the whole lane change thing is so slow and overly cautious that I don't bother with it, just change lanes manually. Takes a couple of seconds and then I can re-engage AP.

 

[Don85D]

Our S85D will likely be our last car and I intend to drive it myself for as long as possible. Fatigue during highway travel (10 hours of driving each day) have never bothered me and as we get older long trips will probably diminish.
I'm OK with AP being the future but I don't need it. The future for AP is probably a ways off as I tend to watch the behaviour of other drivers as much as I watch the roads. I'm not sure how AP will know that the driver beside you is tailgating and itching to lane change in an effort to get somewhere faster. An intuitive AP makes sense but tracking lines on the road is too basic and of little value.

 

[Candleflame]

im pretty sure AP1 also had really dodgy behaviour at the start and it took ~1 month until the software learned to routes properly?? if i recall correctly?

 

[Sir Guacamolaf]

When AP1 was released, we had the same issues. Never having seen or heard of AP before, I thought "this is it?". Did I just pay 100K for a gimmick?

Since then AP1 has improved to the point that at least on a freeway I feel confident letting the car take over, and it is a huge value add to have AP. I still do not feel comfortable in it's "self-parking" abilities, and probably never will.

I'm sure AP2 will get there, but another 6-12 months. Meanwhile, since Tesla's focus has shifted to AP2 100%, looks like AP1 users are left out to dry. Heck I'd appreciate if they'd at least take exits based on nav, and summon private property - as they had promised. But it's all about AP2 now.

My guess is, when Model 3 rolls out, you'll see a significant improvement to AP2. I think they are really pushing the AP2 team to synch with the Model 3 release. After which perhaps, maybe AP1 will get some love. Or maybe not.

 

[brkaus]

I don't understand the fascination with autopilot. I would love to hear why people are so excited about it. I think it is gimmicky and not particularly useful.

It's definitely an interesting software development problem. And there's a niche for it maybe in shipping, taxis, the elderly/blind etc. I also think auto safety is important, but there are many other ways to improve safety. It will also be a very long time before you can fully relax and trust that an AP car won't harm you - even if it's statistically safer than the average human driver. I like the sensation of driving and I am much more comfortable being in control of the vehicle. The BEST thing about Tesla is that it is fun to drive. The movement towards autonomy really dampens that experience.

Does anyone else worry about so many resources being poured into autonomous driving at the expense of development of the rest of the company, e.g., superchargers, service centers, gigafactories, manufacturing innovations, battery development, other mechanical innovations, new products etc...? The other big car makers - VW group, Ford, Mercedes, GM, the Koreans, and the Japanese will eventually catch up to Tesla in terms on the cars themselves if Tesla does not keep innovating. I don't want a robot car. I want a unique, driver's car. Tesla won't be unique if it doesn't stay ahead of the curve in terms of development of the actual car.

You're right that the car is fun to drive. But there are many situations that are not fun. Like the 50 mile drive from San Antonio to Austin when the interstate has very heavy traffic and the speed is varying between 25 and 75mph the entire way.

 

[xyeahtony]

These "scares" are simply unfamiliarity with the vehicle. You're supposed to keep your hands on the wheel at all times, so when i think the car is going to get too close to a car in the adjacent lane, i keep my hand locked on the wheel and prevent it from itching further. If AS wants to keep going one way, it'll eventually disengage...but it never does. Maybe it gets too close for your comfort, but i feel like even on distance setting 1, i'm still too far away from the car in front of me where im constantly getting cut off.

For the naysayers that are saying Autopilot is a gimmick, i drive over 100+ miles a day on autopilot without issues. I'm less fatigued and i feel great and enjoy driving again. The multiple Tesla youtube personalities can attest to that as well. If you don't like autopilot, then don't pay the $5000 unlock fee and just drive a regular car. Maybe Autopilot is not as useful in Germany, but it's definitely useful for a lot of Americans who have long commutes and/or roadtrips. Maybe the autopilot data is better adjusted for American roads as well who knows. Elon Musk's vision is for autonomous vehicles, that's why the Model 3 doesn't have a driver dash console. That's his vision for the future, because human beings can't be trusted to drive without getting into accidents.

And to people saying AP could cause an accident, well still to date no self-driving vehicle has ever gotten into a car accident due to the fault of the system, it's always been another human-driven vehicle causing the accident. Google has had their self driving cars drive millions of miles with no incidents, and their self-driving technology is older and a lot clunkier. AP is a drivers assistance feature, not a let me turn it on and not pay attention feature.

And to the OP, no there is no way to disable traffic aware cruise control and have regular cruise control. personally i find regular cruise control useless but that's just me i guess. "unbelted passenger out of the windscreen" is a bit much, i've had the random occasional ghost brake incidents and nothing has ever gone flying in my car (i leave a lot of junk on the passenger seat).

 

[McRat]

I must admit it's interesting to see the results. More than 1/3 think it's great just as it is. Might as well stop development.

 

[boaterva]

I must admit it's interesting to see the results. More than 1/3 think it's great just as it is. Might as well stop development.

Where do you get that? 'Very Happy' or 'Happy' in NO way says there isn't room for improvement.