NoA is worthless, Autopilot is essentially unimproved from 2016-2017, and everybody has caught up.

[EinSV]

I wonder if part of the reason for the disparity in experiences is simply statistics. Let's say, just for argument's sake, that NOA is 99% reliable. That sounds high. And certainly, that might be good enough for Tesla to release as a L2 system that requires driver attention. But with a large fleet driving all over the US, there are going to be plenty of folks who say NOA is great but 99% will still leave plenty of folks who will have bad experiences sometimes too.

There may be some objective differences in how NOA and AP behave from car-to-car, region-to-region and just statistical blips, but I also suspect a significant portion of the different "experience" is subjective and depends more on the driver's expectations and reactions than differences in how NOA actually behaves.

For example, I use NOA much more than my wife does mostly because she gets impatient that it doesn't pass quickly/aggressively enough even in Mad Max mode. I'm usually a fairly "spirited" driver myself but with NOA on I am fine with a more relaxed approach and don't mind that it makes different decisions than I would.

On the other hand, when I'm a passenger and my wife is zipping in and out of traffic, I often wish she would turn on NOA.

 

[aadams1278]

Thanks.

Weird how people have such different experiences.

Have you never had NOA abort a lane change in the middle, or give you the red hand on wheel alert, or phantom brake, or oversteer into an exit or miss lane exit entirely?

Don't get me wrong. I've had good experiences with NOA too. But I have experienced some issues too.

I have a 2018 Model 3 with AP2.5. Commute 60 miles one way weekly, 90% highway.

I'll use the above questions to frame my experience. I have had lane changes abort, mostly manually commanded changes, and maybe once from NOA commanded change. All for identifiable reasons (missing lane markings or approaching cars). I've never gotten the red hand alert during normal driving on a highway "for no reason". I have gotten it on non-highway roads when something weird happens, usually lost lane markings during a curve or traffic blocking the view or something. Phantom braking used to be much worse. Now it's far better than it once was. Yes I see it but rarely, and more importantly, predictably. It happens at certain points, which I know where those points are and can plan accordingly. I won't call this an "oversteer" but it does aggressively steer into a certain exit lane I use during my commute. Never ever missed an exit.

My general outlook is probably more optimistic than most have stated here. I agree with other comments that have said overall AP has improved over time while some updates have regressions, which are usually temporary. I think we have to keep in mind that AP is still a work in progress - aka not perfect and not finished. The way I approach it is to learn its strengths and weaknesses and use it when it's practical. "Set it up for success" you could say. I know when I can trust it and I know when to be cautious because either I'm unsure if it will succeed, or I know it's likely not to succeed. As an example, we all know it isn't intended to automatically stop at traffic lights and make a left turn through a complex intersection, so I don't "expect" it to and don't get disappointed if I try to make it do that and it fails (although it HAS successfully made a left intersection turn before). Based on that usage pattern, it's not perfect, but I'm happy with it.

Many have said it starts to try to change lanes for an exit roughly 1 mile early. As far as I can tell, it still isn't designed to/doesn't have the capability to intelligently seek out a gap in traffic to perform a lane change if there isn't one directly next to you. In other words, it won't accelerate to get into a gap ahead of you because it knows there isn't a gap behind you. Since I know that, I don't expect it to succeed in situations when that occurs (yet). If that does occur, I manually accelerate or decelerate (using AP speed, not disengagement) to put it into a position that it CAN lane change, and then it does great. It's not FULL self driving yet, it's driver assistance. I like when it changes lanes for me, so I help it get into position and it helps me change lanes.

There are two places on my commute where it hiccups. It's consistent every time. At one exit, the road merges from two lanes to one and it wants to put me in the one that disappears. This is consistent and a very simple fix. I just cancel the lane change and all is fine. This is clearly a lack of map/lane knowledge of that particular exit and I expect is highly likely to improve over time. If I DON'T cancel the lane change, it still works, it just merges back over when the lane ends. While it's not exactly like a human would do it, technically it still "works". Another exit has a slight upward slope to it compared to the highway. It tends to phantom brake here but only for a brief instant and then continues on, only a few mph lost. This did not occur on a previous update, just like the aggressive lane changes into exit lanes. I chock this issue up to the AP code being more broad and capable of detecting more "stuff" than it used to, which is a step toward it being a more general solution than just staying in lanes and taking exits - this is a good thing. My hunch is it thinks it "sees" something that it wasn't previously trained to "see" at this upslope exit, but then it realizes it isn't actually there. While it's a bit buggy in certain cases like this, I believe, overall it is getting more capable on the back end, which we don't directly see, and will improve.

My only real issue with NOA is speed commanded lane changes. I'd say it's about 50/50. There are definitely times when it recommends a change as I approach another car and I'm thinking "why? This car isn't going slower than me." Then 10 seconds later it is decreasing speed because the car ahead of me is in fact going slower than me, and i'm totally impressed! The negative side is when both lanes are relatively slow and the other lane is slightly faster it will recommend a change. I personally don't like continuously changing lanes like this so I just cancel it, but it keeps suggesting it. Because it immediately puts on the blinker when it recommends the change and doesn't allow you any time to cancel before the blinker comes on, I look like an idiot who keeps signaling over and over and never changes lanes. For this reason alone, I disabled unconfirmed lane changes and I've been much happier with it since then. This last case is really not AP's fault. That's when it recommends a lane change, and changes lanes, just for the offending car to then accelerate and me not be able to pass (without accelerating), now blocking the lane. Tesla can't change how OTHER people drive, so good luck to them on fixing that problem.

Last, I'll say this. NOA progress has slowed/stopped for a while. I don't know a ton about code but I know some, and I think it's safe to say the jump from NOA to full self driving is a massive leap. They can't just turn on FSD when it's 80% ready and say "have at it". That's a tall order. Also I think it's worth considering that Tesla has recognized that leap can't happen on AP2.5 hardware, so now they're at this point where they have to split the code between AP2.5 and AP3. If it's MUCH harder to get beyond the current state of NOA with the complaints we have mentioned in this thread, it makes total sense to kindof pause trying to "make it work" in its current form on AP2.5 and focus on improving it on AP3. As of now, we have only seen "FSD Preview" rolled out to the public, which leads me to believe we are basically all still on AP2.5 code, even if you have AP3. Elon has also said that everything so far has been optimized for AP2.5. I expect when AP3 is fully "turned on" with code designed for AP3, that's when the significant improvements will start to happen.

 

[spatterso911]

Your experience mirrors mine. I am in an HW3 Raven S. I drive about 200 miles a week on AP. 95% is with little correction from me. I wonder if the other automakers have simply caught up to HW2.0/2.5 cars. I really don’t see the issues that the OP has.

 

[_jmk]

I wonder if part of the reason for the disparity in experiences is simply statistics. Let's say, just for argument's sake, that NOA is 99% reliable.

Sounds about right. It only tries to kill me once every 99 seconds and the rest of the time I just feel horrified

 

[diplomat33]

There may be some objective differences in how NOA and AP behave from car-to-car, region-to-region and just statistical blips, but I also suspect a significant portion of the different "experience" is subjective and depends more on the driver's expectations and reactions than differences in how NOA actually behaves.

For example, I use NOA much more than my wife does mostly because she gets impatient that it doesn't pass quickly/aggressively enough even in Mad Max mode. I'm usually a fairly "spirited" driver myself but with NOA on I am fine with a more relaxed approach and don't mind that it makes different decisions than I would.

On the other hand, when I'm a passenger and my wife is zipping in and out of traffic, I often wish she would turn on NOA.

I don't mind the subjective differences. People do have different driving styles. I do not consider it a failure if NOA does not drive the same way I do. For me, I judge NOA based on whether it is safe and does it what it is supposed to do. So if it moves over to the right lane 2 miles before taking an exit and takes that exit smoothly and safely, I am happy. So my main concerns are when NOA does something unsafe or fails to do what it is supposed to (like fails to take an exit or fails to do a lane change).

 

[MorrisonHiker]

I use regular Autopilot/TACC on my daily commute. I don't use NoA most of the time since it wants to change lanes based on the car right in front of it and can't see traffic way ahead. However, I love using NoA on road trips. When I'm driving across Colorado, Nebraska, Kansas, etc., it works perfectly, changing lanes to pass cars and taking exits. When doing these road trips, the car easily does 99% of the driving for me and I typically only disable NoA when I'm back in the Denver metro area.

Autopilot has gotten much better since 2016/2017. When we got our first AP2 cars, they had virtually no AP functionality so to say it is essentially unimproved is totally false. We did tests almost weekly in 2017 and AP couldn't go above certain speeds, handle curves, continue going straight instead of diving to a turn lane, make automatic lane changes, etc. We stopped doing AP tests because it eventually passed all of our tests.

While some other auto manufacturers have added lane keeping to their current vehicles, I don't think they've all "caught up" because none of them have added such functionality to all the cars they've manufactured since late 2016.

AP isn't perfect but it has improved greatly over the years and continues to improve in nearly every firmware update.

 

[Amp Chaser]

After 6 months with my M3 Dual / FSD and 12,000 miles, I found the secret to FSD / AUTOPILOT. Here is the secret HOW/EOR!

 

[bhzmark]

all of those companies now have rivaling technologies that have caught up or are getting really close.

I haven't seen an update to this good study, but this showed that even the original 2018 Model 3 was much better than everyone else, even the older Tesla AP1 (here represented by the Model S).

IIHS examines driver assistance features in road, track tests

 

[bhzmark]

I use NOA (no confirmation setting) on essentially every freeway drive. It keeps getting better, handling more challenging merges and exits and complex situations involving unexpected maneuvers from other cars.

I find it very relaxing to have the car do most of the work and rarely feel the need to take over, unless the traffic is very heavy and I'm in a big hurry for some reason.

Same

I am not aware of any current commercially available system that can come close to keeping up with the capabilities of a 2016 Model S/X or 2017/18 Model 3 with AP2, and none of the currently available cars have a chance of keeping up with HW3 equipped vehicles.

+1

Weird how people have such different experiences.

I suspect it probably has everything to do with expectations. Some people expect AP to do exactly what they would do at the time and in the manner that they would do it. Other people see AP as tool with a certain functionality that will be different from their driving functions, but will still get to the same place roughly the little slower, and roughly the same way. And they easily accept that trade off -- perhaps based on understanding the underlying concepts.

25% of the decisions are flat out wrong (getting on the wrong lane / trying to change lanes even when the current lane is fine) and about 70% of the right decisions have wrong timing.

See. A tool isn't "wrong." It just responds to its inputs in relatively predictable ways that may not optimize -- or appear to optimize -- in the same way that the driver might choose to. Getting over way early for an exit is probably more "right" than waiting until the last minute and then squeezing in. The latter was previously my typical driving and I was probably about 70% "wrong" on timing according to AP, and good safe general driving for all drivers.

 

[Goose66]

I also have similar experiences to the OP in my Model 3 with 2.5 hardware (August 2018). I would also add that, through the recent updates, Navigate on Autopilot has gotten markedly worse for me, not better. I drive 8 miles or so to work everyday - about 5 of that on the Interstate with one highway transition. The car works fine in stop-and-go, bumper-to-bumper traffic and also does fine in open roads at speed. But in bumper-to-bumper traffic at speed, the Tesla NoA is now a nightmare. It immediately works over to the far left lane when getting on the Interstate only to figure out (too late) that it can't get back over to the right to make the highway interchange. And then, getting over to exit at the end of the commute has gotten really squirley, with sometimes it telling me to take over because it can't make the maneuver, and other times I forcefully take over because I see it's going to miss the exit.

Something that popped up in the last month is equally disturbing. At the highway transition and/or at the exit ramp, when I intervene using the steering wheel and brake, it releases control. Then, anywhere from 10 to 15 seconds later, it sounds a loud alarm and pops up a message saying "Take over immediately!" Of course, at that point there is nothing to take over - I've been driving the car manually for 15 seconds. But that kind of latency is very disconcerting. Repeated logging of this issue with Tesla support has gone with no response.

Anyway, I used to use it every morning just to make my commute more relaxing, but these days I don't use NoA at all for morning commutes (just for open-highway and bumper-to-bumper stop-and-go traffic on occasion) and will probably not try it again until I get the HW3.0 upgrade (whenever that is).

 

[RandyS]

I doubt it. Elon is adamant about his vertical integration approach and doing everything in-house. It's why he dumped Mobileye and started AP2 in the first place.

Hmmm. I thought that Mobileye dropped Tesla as a client for safety reasons...

Mobileye spills the beans: Tesla was dropped because of safety concerns

 

[diplomat33]

Hmmm. I thought that Mobileye dropped Tesla as a client for safety reasons...

Mobileye spills the beans: Tesla was dropped because of safety concerns

Yes that too. But Elon also believes strongly in vertical integration. So I think Elon was probably ok with Mobileye dropping Tesla since he wanted to do FSD his way and not be constricted by a supplier with a different vision.

 

[EinSV]

I don't mind the subjective differences. People do have different driving styles. I do not consider it a failure if NOA does not drive the same way I do. For me, I judge NOA based on whether it is safe and does it what it is supposed to do. So if it moves over to the right lane 2 miles before taking an exit and takes that exit smoothly and safely, I am happy. So my main concerns are when NOA does something unsafe or fails to do what it is supposed to (like fails to take an exit or fails to do a lane change).

Since human drivers routinely cut me off or get distracted and drift across lane lines, personally I would feel much safer if everyone I had to share the road with was using Navigate on Autopilot.

 

[linux-works]

Yes that too. But Elon also believes strongly in vertical integration. So I think Elon was probably ok with Mobileye dropping Tesla since he wanted to do FSD his way and not be constricted by a supplier with a different vision.

I'm seeing a few approaches in trying to get the full suite of 'car self driving' software to work.

some are doing an entire DIY approach, using primarily (if not entirely) in-house written and designed software. that's the long, slow, expensive route - and you better be right since going back is going to lose a lot of time (we see that with the hw3 native sw port concept).

some are in the 'buy' mode and are buying premade components (like autosar from the various vendors, electrobit, vector, etc). there's a lot that can be said about going with autosar, but if you are not already there, converting from proprietary to standards based is going to be HARD and SLOW. and EXPENSIVE (if you ever saw what vector charges, oh boy!)

v2x software, walk-up unlock beacon software, visual recognition (nvidia and TI are big, here) - lots of that can be bought. the 'buy' method means that once the market has settled on a few successful vendors, interop (ie, driving on shared roads) will be better since the algorithms will be more inline and less about fighting each other.

there is no clear choice, though.

tesla has a lot of software investment in their own code, so I don't see them going standards-based any time soon. the other vendors, they'll prefer to buy instead of write - at least the hard parts.

 

[mikegre]

For what its worth, I use NoA every day from on ramp to off ramp on my way to work, the maps are good and it does what you would expect and significantly makes my commute more enjoyable. I haven't had a phantom brake for a long time and aborted lane changes are typically for identifiable/justifiable reasons. I've learned to live with the NoA style and just let it do its thing. Not always how I would do things, but generally safe and effective.

I've also experienced some routes that show NoAs weaknesses, in which case I just use AP, which is now pretty robust actually. Being able to use the blinker for lane changes (and having it actually know if it is safe or not to a pretty high degree of accuracy) is well and above most other systems on option, including AP1. I've been in AP1 cars and I would say they are nearly as good in 'normal' situations but not remotely comparable in difficult situations (like hard braking, hard turning, weird lanes etc). So to say progress has not been made is kind of a trolling comment really.

I think what you mean is, it doesn't do exactly what you want, where you want it to, all the time.

That's a very good point about letting it do its thing. At first I was unhappy with the choices NoA was making but I've learned to relax and let it do its own thing.

 

[clydeiii]

Given that I use it every day on every commute I make (25 miles each way mostly on I-95 between DC and Baltimore), it certainly isn't "worthless." It is worth a ton to me. I hardly ever have to intervene even.

 

[S4WRXTTCS]

Here is my analysis of it, but keep in mind that 90% of my experience with TACC/AP/NoA is on 200+ mile road trips as my commute to work doesn't involve any highways.

TACC is a pile of crap compared to AP1, and compared to offerings from other manufactures. Quite simply Tesla tries to do too much with TACC on AP2, and it's not nearly as smooth as it could be. Thankfully in the latest version of the SW it seems like false braking has been decreased. But, even with decreased false braking it still does silly things like tracking the semi next to me and slowing down for it. I haven't had a 200 mile drive without it at least making a few minor mistakes, and at least one moderate one. I haven't had any major speed change one in quite awhile. So I don't think it's dangerous as much as it's simply not good enough.

AP2 is improved from what it used to be. Like it used to ALWAYS recenter itself on merge points. It still does, but it's way less severe. I'd say AP2 still leads the vast majority of competitors especially since you can command lane changes. The commanded lane changes are beautiful. I'm a pretty harsh reviewer, but even I have to admit auto lane changes is pretty awesome. I'm also someone who always gets out of the passing lane. So I'm constantly going from the middle lane to the left lane to pass, and then back to the middle lane (or right lane depending on how many lanes and traffic). In my book consistency is what really matters, and driver initiated AP lane changes with HW2.5 (what I have) are very consistent. My only really dislike with AP2 is the torque sensor in my Model 3 isn't very sensitive so I have to hold the steering wheel in a less than ideal manner or it will nag me.

NoA - I think it's kinda pointless since it's completely inconsistent. I've had good experiences with it, and I've bad experiences with it. I would like it to be more consistent. One of the things that impressed me about it is it could determine better than I could what the vehicles speed was ahead of me. Like once it canceled a lane change because the person sped up. I thought that was pretty cool. I would have changed lanes to the left lane only to realize later that the car ahead had sped up, and then I would have had to get back over. I also had to turn off speed based lane changes because occasionally (regardless of aggressiveness level) it would to change lanes to pass on the right, and I can't stand for that. I also find NoA worthless since they broke so much about it in recent firmware builds with HW2.5. For example getting out the passing lane is broken, and unconfirmed lane changes are too slow. Supposedly HW3 doesn't have those issues, but I won't be able to try HW3 for another few weeks.

 

[Maitri982]

I’m pleasantly surprised that as of my writing this, the OP does not have any disagrees. The first thing I did when I saw the title was scroll down and check that. I echo everything mentioned as well. I’ve yet to find a scenario where NoAP is anything but a detriment, and I make sure to give it a try on all the updates.

I would also caution people who are holding out hope for the big development that’s just around the corner. First it was “now that we’ve gotten rid of MobilEye we can really move forward”. Then it was “Chris Lattner has rewritten the stack so we can really move forward”. Then it was “Andrej Karpathy has rewritten the stack so we can really move forward”. Now it’s “we’re almost done rewriting the stack so we can really move forward”. After this it’ll be “we’re bringing dojo online so that we can really move forward”.

Maybe one of these times they’ll get something to stick after throwing it against the wall, but history has not validated any of their previous claims. Meanwhile all the “dinosaur OEMs” actually understand the point of “the tortoise and the hare”.

I am huge Tesla/Msuk fan, but he has been full of it on FSD and he knows it. He said like 5 years ago that FSD was "basically a solved problem", which was ridiculous. He promised summon in a few weeks then it came a year later. When it actually was about to come he said it was "mind blowing"...it is OK and certainly no one else is doing anything like that, but I would not call it mind blowing.

Some guy asked about avoiding potholes, and he said it was "coming". That's BS and he knows it. Although he did not specify time frame but the implication was it was coming in near future, which given that it can't handle other basics of driving is also silly and not true.

I hope that FSD is solvable with the current sensor suite and with DL networks, but no one has solved it yet so it remains a theoretical conjecture at this point.

For me i would be happy if it just let me not pay attention on geofenced highways only because it would be much better than humans...city diving is fine but highways would be great.

So we wait and see...I had guessed 2021-2022 for anything really interesting.

 

[aadams1278]

“A delayed game is eventually good. A bad game is bad forever.” Shigeru Miyamoto - game producer at Nintendo. This quote references the importance of allowing games the necessary time to complete development sufficiently rather than rushing the process in order to meet an overly ambitious release date.

I think I’ve also seen this misquoted or paraphrased as “a late game is only late until it’s released, but a bad game is bad forever.” Which is probably more fitting in this case.

Ever since I first saw this quote years ago, it has always shaped my thoughts on software. While it’s not 100% accurate anymore since software updates weren’t a thing back then, I think it is still relevant as we are sitting here complaining about released features that aren’t perfect... and the ones that haven’t been released yet.

 

[malcolm]

Since the AP team reports directly to Elon and we know that he always has the latest build downloaded to his own car to test and provide feedback, maybe NOA just drives as badly as Elon but he thinks it's brilliant.