Navigate on Autopilot is Useless (2018.42.3)

[SongjinDK]

@verygreen

The manual doesn't say if we should wait until it is safe to change lane before confirming the lane suggestion, which I don't think so. I think we should confirm the correctness of the suggestion, not if it is safe now. I do agree we should watch it closely during the lane change and be ready to take over.

Do you think we should wait until it is safe to change lane before confirming it? In that case, we will never do it in heavy traffic since the car wont turn on blinker before confirmation and nobody knows we intend to change lane so no one will yield.

 

[verygreen]

@verygreen

The manual doesn't say if we should wait until it is safe to change lane before confirming the lane suggestion, which I don't think so. I think we should confirm the correctness of the suggestion, not if it is safe now. I do agree we should watch it closely during the lane change and be ready to take over.

Do you think we should wait until it is safe to change lane before confirming it? In that case, we will never do it in heavy traffic since the car wont turn on blinker before confirmation and nobody knows we intend to change lane so no one will yield.

the manual says if you don't confirm in 3 seconds, it will be unhappy. Also it's my reading that once confirmed the maneuver would be attempted right away safety-permitting (but the manual is ambiguous about it, so I suggest filing a bugreport with Tesla. Manuals should not be ambiguous.)

 

[Matias]

Similar to Boomer19’s question above, what are the reasons behind AP2 deficiencies and why hasn’t Tesla able to address them? I understand the breakup with Mobileye, but that was two years ago. Why hasn’t Tesla progressed any further? Do they have the wrong approach? Are they incompetent? Are they being conservative?

What is the consensus on what’s holding them back?

A couple of the contributors to this tread seemingly have the solution, or at least some good ideas. Will these ideas not scale to what Tesla needs? Most of us want Tesla to be successful. How can this community help get us where we want to be?

Well.. I think the most obious answer is, that this is a difficult problem. I mean, no-one has a production car which is any better. If there would be other production cars with better AP, I would say Tesla has dropped the ball. But there is not.

And don't bring any reseach models to this conversation.

 

[electronblue]

Allow dumb cruise control when radar sensor is blocked with snow/salt/sand.

I actually did this mod on my S a while back when I had a broken radar bracket that would eventually slide out of usefulness. The problem with it is that you get used to TACC and it limits your reaction time, even if you've been told by the car you're in dumb mode now. So, honestly, this is something that shouldn't be done. It actually is a problem resolved on the facelift cars, X, and 3... since the radar is protected.

This is incorrect if you refer to sensor blocking being solved. It is not. Moving radar behind bumper meant it can no longer melt any snow in front of it. Snow is known to block radar and disable Tesla Autopilot in some climates on facelift or later cars. Elon Musk tweeted about possibility of adding a new radar heater some days ago.

Elon Musk on Twitter

 

[avoigt]

Well, my AP1 car behaves differently in different days in the exact same spot. On my daily route road makes turn to the left. Line markins are poor. 9 times of ten car tries to drive straight, but for some reason 1/10 it follows the road. This is without lead car. So I don't find it strange, that different people experience different behaviour in the same kind of situations.

I believe it is time to stop talking about AP 1 and I mean this also with respect to all who use it and like it. Its a systems that is not sold any more and soon we have the 3rd generation of AP out. Its also not helpful as a benchmark as too much changed to be able to compare it really. To be clear I do not say that it may not be better in some situations still but that will decrease and vanish. Also the vast majority of people have only experience with AP2 and that number does increase versus the AP1 number not.

I understand the frustration of many people but that discussion what AP1 once did and do, does not add value. To go from AP1 to AP2 was/is certainly painful and a long way to go. AP1 is gone and will not come back.

Many people will not like to hear that but its just the reality we are in.

About the discussed issues around NoAP I did search yesterday and only found positive videos how it performs. Maybe just only positive experiences are posted but that is in my view unlikely.

Here is another one:

 

[Matias]

I believe it is time to stop talking about AP 1 and I mean this also with respect to all who use it and like it. Its a systems that is not sold any more and soon we have the 3rd generation of AP out. Its also not helpful as a benchmark as too much changed to be able to compare it really. To be clear I do not say that it may not be better in some situations still. Also the vast majority of people have only experience with AP2 and that number does increase versus the AP1 number not.

I understand the frustration of many people but that discussion what AP1 once did and do, does not add value.

Apparently I wrote badly, because I did not want to compare AP1 and AP2 or AP 2.5 at all. I also did not want to blame AP. I just wanted to point out the fact, that AP can on different days behave differently in the exact same spot, so it is not strange, that different people experience different behavior in same kind of situations.

But you are right, discussing AP1 does not bring much (if any) value to this.

I'll get my coat

 

[avoigt]

Apparently I wrote badly, because I did not want to compare AP1 and AP2 or AP 2.5 at all. I also did not want to blame AP. I just wanted to point out the fact, that AP can on different days behave differently in the exact same spot, so it is not strange, that different people experience different behavior in same kind of situations.

But you are right, discussing AP1 does not bring much (if any) value to this.

I'll get my coat

Thats fine and I did not wanted to call you out on that. My post has been aimed at a larger group here.

I do agree that AP can act differently at the same spot because situations are usually always different and the system does take more and more external factors into account. Not likely to have exactly the same situation twice unless you mimic it in a test environment.

 

[lunitiks]

!AP14LIFE!

 

[diamond.g]

CR's review comes across (to me) as meh as well.
Tesla's Navigate on Autopilot Shows the Promise and Problems of Self-Driving Cars

 

[Az_Rael]

@verygreen

The manual doesn't say if we should wait until it is safe to change lane before confirming the lane suggestion, which I don't think so. I think we should confirm the correctness of the suggestion, not if it is safe now. I do agree we should watch it closely during the lane change and be ready to take over.

Do you think we should wait until it is safe to change lane before confirming it? In that case, we will never do it in heavy traffic since the car wont turn on blinker before confirmation and nobody knows we intend to change lane so no one will yield.

Did you read the agreement before turning on NoA? It explicitly says to check your surroundings and blind spots prior to confirming. I read this to mean if it is not safe to change right then, do NOT confirm the lane change.

 

[PhaseWhite]

Did you read the agreement before turning on NoA? It explicitly says to check your surroundings and blind spots prior to confirming. I read this to mean if it is not safe to change right then, do NOT confirm the lane change.

View attachment 350234

In my experience the lane change confirmation works the same way auto lane change works without NOA enabled. It will not change lanes if there is a car it sees next to you. And it does start asking you to confirm a lane change even when cars are next to you and the display clearly shows this.

Now to other's points, should you just blindly trust this will work? Hell no, you should be watching and holding the wheel to override if there is a car there when it starts changing lanes.

The manual is a good reference, but my observations in this case are that it is poorly worded. To @verygreen 's point, it should be updated.

 

[scottf200]

the manual says if you don't confirm in 3 seconds, it will be unhappy. Also it's my reading that once confirmed the maneuver would be attempted right away safety-permitting (but the manual is ambiguous about it, so I suggest filing a bugreport with Tesla. Manuals should not be ambiguous.)

Your 'unhappy' remark is personification. It simply says is a reminder (ie. you may be waiting for an opening watching your mirrors, etc, which is *perfectly* normal): "If you do not confirm the lane change within three seconds, a chime sounds to remind you that Navigate on Autopilot requires your confirmation to change lanes."

Did you read the agreement before turning on NoA? It explicitly says to check your surroundings and blind spots prior to confirming. I read this to mean if it is not safe to change right then, do NOT confirm the lane change.

That is confusing and unrealistic based on his use-case scenario. As he pointed out if there is a steady stream of cars (i.e. in your imm left lane) going faster than you and you want to go there then it is pretty rational for him to put on his blinker and hope someone gives him some extra feet to merge in ... at least he is communicating/warning (blinker) when he makes his 'mad max' manual move.

 

[scottf200]

About the discussed issues around NoAP I did search yesterday and only found positive videos how it performs. Maybe just only positive experiences are posted but that is in my view unlikely.
Here is another one:

That video demonstration showed several features working pretty well. Seems NoAP likes additional msgs (informational, warning, etc) too.
NOTE: Good to watch on a PC in youtube where you have the GEAR icon so you can set the speed to x1.5 (but keep in mind that some actions the car is taking then is x1.5 so it seems faster than it was).

 

[boonedocks]

Someone with a good experience of Nav on AP

Vide of nav on Autopilot (42.2) handling the end of a highway in the rain

WTF I don’t have TACC or AutoPilot OR rain sensing wipers OR auto high-beams when it rains. Every driver assistance feature magically disappears INCLUDING ABS warnings ¯\_(ツ)_/¯. Had an hour long on hold wait and a 15 minute conversation which lead to a long email with tons of pictures. Here’s two of the 20 BTW my wife's AP2.5 Model X does NOT exhibit this behavior at all.

 

[wk057]

I believe it is time to stop talking about AP 1 and I mean this also with respect to all who use it and like it. Its a systems that is not sold any more and soon we have the 3rd generation of AP out. Its also not helpful as a benchmark as too much changed to be able to compare it really. To be clear I do not say that it may not be better in some situations still but that will decrease and vanish. Also the vast majority of people have only experience with AP2 and that number does increase versus the AP1 number not.

I understand the frustration of many people but that discussion what AP1 once did and do, does not add value. To go from AP1 to AP2 was/is certainly painful and a long way to go. AP1 is gone and will not come back.

Many people will not like to hear that but its just the reality we are in.

Actually, it's you who needs to get over the fact that AP1 isn't going anywhere. AP1 is the hardware still in use today by a sizable portion of the Tesla fleet (over 125,000 vehicles last I knew, with the AP2 vehicle count only recently passing that in the past 6 months or so with the Model 3 ramp up), and thus in use by a huge number of owners Heck, AP1 was still being sold less than 2 years ago, and is still sold by Tesla on CPOs and such... meaning a huge chunk of those vehicles aren't even out of warranty yet and you're saying it's irrelevant. Just because Tesla has moved on to AP2, AP3, AP4everdelayed doesn't mean AP1 is irrelevant and thus comparisons with it are not helpful.

In fact, AP1 is, in mine and many others' options, still the best publicly available drive assist system out there. AP2 has yet to match its features and reliability fully. This is objectively true, despite claims to the contrary. It doesn't matter if AP2+ can do new-feature-X (poorly) that AP1 doesn't have when AP1 has features Y,Z,A,B,C,D,E that AP2+ hasn't even gotten close to implementing yet.

So yes, AP1 is still going to be the benchmark for Tesla's own development. They should eventually surpass it in every metric if they get their **** together, but honestly, I don't see it happening in the very near term. In the meantime, AP1 folks can definitely continue to enjoy their cars, despite being forsaken by Tesla entirely (more broken promises).

Here's a funny one that someone brought up to me privately: "With the same level of human interaction, what can stock AP2+ do that AP1 can not?" ... and I thought about this, and the answer is actually "almost nothing." The only thing AP2+ can do that AP1 can not at this point is engage a lane change towards a ramp with less human interaction. Everything else you do with AP2+ requires the same or more human interaction than the same features on AP2+. NoA doesn't change lanes without human interaction, so that's a wash since AP1 can change lanes with human interaction, too. AP1 and AP2 can do auto steer with hands-on nagging. Etc etc. There's plenty of cases where AP1 performs better, and plenty where AP2 performs better. But the answer to the original question is still the same.

 

[avoigt]

Actually, it's you who needs to get over the fact that AP1 isn't going anywhere. AP1 is the hardware still in use today by a sizable portion of the Tesla fleet (over 125,000 vehicles last I knew, with the AP2 vehicle count only recently passing that in the past 6 months or so with the Model 3 ramp up), and thus in use by a huge number of owners Heck, AP1 was still being sold less than 2 years ago, and is still sold by Tesla on CPOs and such... meaning a huge chunk of those vehicles aren't even out of warranty yet and you're saying it's irrelevant. Just because Tesla has moved on to AP2, AP3, AP4everdelayed doesn't mean AP1 is irrelevant and thus comparisons with it are not helpful.

In fact, AP1 is, in mine and many others' options, still the best publicly available drive assist system out there. AP2 has yet to match its features and reliability fully. This is objectively true, despite claims to the contrary. It doesn't matter if AP2+ can do new-feature-X (poorly) that AP1 doesn't have when AP1 has features Y,Z,A,B,C,D,E that AP2+ hasn't even gotten close to implementing yet.

So yes, AP1 is still going to be the benchmark for Tesla's own development. They should eventually surpass it in every metric if they get their **** together, but honestly, I don't see it happening in the very near term. In the meantime, AP1 folks can definitely continue to enjoy their cars, despite being forsaken by Tesla entirely (more broken promises).

Here's a funny one that someone brought up to me privately: "With the same level of human interaction, what can stock AP2+ do that AP1 can not?" ... and I thought about this, and the answer is actually "almost nothing." The only thing AP2+ can do that AP1 can not at this point is engage a lane change towards a ramp with less human interaction. Everything else you do with AP2+ requires the same or more human interaction than the same features on AP2+. NoA doesn't change lanes without human interaction, so that's a wash since AP1 can change lanes with human interaction, too. AP1 and AP2 can do auto steer with hands-on nagging. Etc etc. There's plenty of cases where AP1 performs better, and plenty where AP2 performs better. But the answer to the original question is still the same.

It does not matter if AP 1 was sold to 125k users and still in CPOs. Its not sold any more and that is all what matters. 125k cars is nothing compared to AP2 deliveries of today.

Understand your frustration but it does not help anybody to discuss AP1 that was sold 2 years ago. Thats an outdated technology and will not be reactivated. AP2 is what we should look at and discuss because that is what people have to decide if they purchase or not today.

 

[AWDtsla]

It does not matter if AP 1 was sold to 125k users and still in CPOs. Its not sold any more and that is all what matters. 125k cars is nothing compared to AP2 deliveries of today.

Understand your frustration but it does not help anybody to discuss AP1 that was sold 2 years ago. Thats an outdated technology and will not be reactivated. AP2 is what we should look at and discuss because that is what people have to decide if they purchase or not today.

Hope you'll have an equal outlook on AP2 when it's obsoleted in 2019. No the computer upgrade won't save you.

PS we're on AP 2.5 now.

 

[wk057]

It does not matter if AP 1 was sold to 125k users and still in CPOs. Its not sold any more and that is all what matters. 125k cars is nothing compared to AP2 deliveries of today.

Understand your frustration but it does not help anybody to discuss AP1 that was sold 2 years ago. Thats an outdated technology and will not be reactivated. AP2 is what we should look at and discuss because that is what people have to decide if they purchase or not today.

Respectfully disagree. This is not frustration, just facts.

Fact: A double-digit percentage of owners are on AP1.
Fact: AP1 is a benchmark for Tesla's progress on future versions.
Fact: AP1, to-date, has nearly every usable functionality that AP2+ has, plus others AP2+ does not. (AP1+: Autosteer, Auto lane change, blind spot detection, rain sensing, auto high beams, AEB, SCA)

Just because Tesla bumps the version on things doesn't make the previous version irrelevant.

Perhaps if AP2+ could withstand comparison to AP1 with any actual differentiation/improvement, this discussion wouldn't be nearly as relevant.

 

[Az_Rael]

That is confusing and unrealistic based on his use-case scenario. As he pointed out if there is a steady stream of cars (i.e. in your imm left lane) going faster than you and you want to go there then it is pretty rational for him to put on his blinker and hope someone gives him some extra feet to merge in ... at least he is communicating/warning (blinker) when he makes his 'mad max' manual move.

Right, I agree about manually making that type of lane change (assuming you are in a part of the country where people are nice if you turn your blinker on and let you in).

But we are talking about autopilot lane changes.
I am guessing if you got an official response from Tesla, the answer would be to not confirm the change until it was safe to actually change.

I think the system is not currently capable of handling those tight situations and is more for light traffic in its current iteration.

 

[scottf200]

...
Everything else you do with AP2+ requires the same or more human interaction than the same features on AP2+. NoA doesn't change lanes without human interaction, so that's a wash since AP1 can change lanes with human interaction, too. AP1 and AP2 can do auto steer with hands-on nagging. Etc etc. There's plenty of cases where AP1 performs better, and plenty where AP2 performs better. But the answer to the original question is still the same.

Again total respect for your work and contributions.

You are talking about physical human interaction as that is the only measurement. It definitely is not for many people. NoAP provides *guidance* for many things to get you to your destination. For example getting you in the right lane because the road you are on is going to split and you need to be in the, for example, right 2 lanes to keep on your intended route. This is great for people traveling to new places *OR* to places they only occasionally go to. This is not FSD or nothing. It is steps in a measured and safe way to get to higher goals.