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

Is the current NoA helpful or harmful?

Do you agree that current NoA requires significant intervention and is a potential safety risk?

  • Yes

    Votes: 80 45.5%
  • No

    Votes: 96 54.5%

  • Total voters
    176
This site may earn commission on affiliate links.
Musk has stated on numerous occasions that he firmly believes the sensor hardware available in current models is all that is needed. Some have questioned this logic, for various reasons - including stereoscopic vision, as well as radar sensor redundancy, but I will state that there is an area of study which reproduces stereo vision with a single camera operating multiple frames over time and distance. There have been studies, videos, and even educational coursework on this subject.

Here is one example of some prior work: Driving Computer Vision with Deep Learning

And he promised numerous times that the CPU in HW2 / 2.5 was powerful enough. So why do we all need HW3 now?

I already addressed the single camera issue - it works fine looking forward and when objects are moving relative to you. What's in your link looks just like what Tesla showed at the autonomy day briefing - a scene in front of the car where everything is panning relative to the car. The principle is the same as SAR - synthetic aperture radar - which has been around since the 1950's, so it's not that novel. The problem is when looking at other cars moving with you, such as when merging onto a highway, there's not enough relative angular motion for the computer vision algorithm to do it's thing. And that's why the car (at least my car) on NoA isn't confident merging onto a highway and doesn't show any cars in that lane I'm trying to move into - it's not sure what's there. Maybe it sees them, but can't figure out how close they are.

And it's possibly why it's not good at fast approaching cars either - until it can get both the rear and rear-side cameras on a car coming up from behind to triangulate, or that car gets close enough to have some angular motion, there's no motion one camera can detect as that car is approaching at a constant angle until it gets close. It would be different if the camera could pick up a doppler effect, but I'm not aware of that technology existing in the visible spectrum, as least for the small speed changes we're talking about. Or it could determine the car is getting "bigger" as it approaches, but in my opinion the rear facing cameras need to be higher resolution to pick that up from far enough away to be useful.

I guess we'll know in about a year...
 
But what struck me as odd is they said NoA cut someone off... NoA is such a granny that doesn't seem possible and I personally would need to see a video of it happening because I do not believe that NoA can cut anyone off with it's current programming.

I had it cut someone off once. They were approaching quickly on the right, but there was space if NoA made a timely move so I let it do it’s thing in the spirit of experimentation. It changed lanes into the gap soooo slowly that by the time it finally got in the other lane, the other car was what felt like inches from my bumper, boxed in, and tailgating me until I exited. It was a pure a-hole driving move by my car - the most passive aggressive cut-off ever, and I am sure the other driver was cursing my name. As the human driver, I would have either gotten into the lane decisively allowing the car behind time to move out if he wanted, or let him pass first.
 
I had it cut someone off once. They were approaching quickly on the right, but there was space if NoA made a timely move so I let it do it’s thing in the spirit of experimentation. It changed lanes into the gap soooo slowly that by the time it finally got in the other lane, the other car was what felt like inches from my bumper, boxed in, and tailgating me until I exited. It was a pure a-hole driving move by my car - the most passive aggressive cut-off ever, and I am sure the other driver was cursing my name. As the human driver, I would have either gotten into the lane decisively allowing the car behind time to move out if he wanted, or let him pass first.

See for me, I get the opposite, if a car is coming quickly, NOA just aborts even if it has plenty of time to slowly move it's arse over and even it it's already halfway through the lane change. I usually end up having to take over to finish the lane change.
 
See for me, I get the opposite, if a car is coming quickly, NOA just aborts even if it has plenty of time to slowly move it's arse over and even it it's already halfway through the lane change. I usually end up having to take over to finish the lane change.


It is possible the behavior I saw has been changed in the more recent firmware releases as I have also been "bullied" out of a lane change by a fast moving car more recently. It is possible CR was testing with the firmware shortly after the no confirmation lane change was added, which is when I had the car cut someone off.

So, it might be accurate to when they tested, but out of date now. CR doesnt have the resources to test every firmware drop I would guess, so they are probably only doing the major ones that add big features. Unfortunately, that will mean they always will have bad things to say about the releases because the initial release of a feature is always a bit half baked. Kind of the downside of how Tesla rolls things out.
 
  • Like
Reactions: Magellan55
Agreed on most of these points. My wife loves NoA and uses it all the time. Lane changes in heavy traffic are messy so its easier to grab control then flip it back on NoA, but in light traffic lane changes are great. Especially knowing updates will keep coming, not regreting the extra cash it cost whatsoever. I can make up for the few silly decisions it wants to make, while it makes up for the many more stupid decisions I make while driving (looking down at my phone for a moment and almost rear ending someone, etc.)
 
  • Like
Reactions: silenteski
And if you do not understand the risks of beta software turn it off and do not use until it is officially released. You are a danger to yourself and others without knowing it will bite you on the a$$ at the worst possible moment. As for track mode maybe Tesla should require you acknowledge all the risks and issues every time you turn AP on.

There are still many situations which it either does not know how to handle or handles poorly, like school zones or roundabouts or intermittent lines on the pavement. It is up to Tesla to figure out how to deal with these things, it is not up to these things to be Tesla OK.
 
  • Like
Reactions: Magellan55
While CR has its merit, but it is a beta software and I expect there are many scenarios that it cannot handle. I use NOA all the time and know when it excels and when it's time to take over. I have the one on-ramp that is simply waiting to crash if I let noa handles it, but I also have other succesfull on ramp and off ramp stories. I take CR report as a one off test report and be cautious about it but wouldn't completely dismiss NoA.
 
I just returned from a 1400 mile road trip on 2019.12.1, and NOA was indeed helpful. I did NOT let it change lanes on its own, but let it prompt me. I'd say about 40% of the prompts were stupid or useless. Lots of times I'd be driving along in the right lane, and i'd be prompted to move left to "change lanes out of the passing lane"

However, where it was most helpful was in reminding me to change lanes when slowly approaching a slower car. Eg, I'd be doing 70, and the car I was approaching was doing 68. With NOA off, I would sometimes not change lanes until I'd lost speed and slowed down to 68. However, NOA was fairly good at prompting me to change lanes before I would have to slow down.

I did the drive in 2 13-14 hour stretches, and the NOA provided enough entertainment / distraction to keep me more focused. Especially when the car kept binging as it cut in and out in heavy rain. All those bings were quite valuable in terms of keeping me alert.
 
CR reviews things like it's still 1965 and software updates don't exist. I take their old-school approach with a giant grain of salt. It's good for some things and downright misleading for others, like complicated driving software that improves over time. CR just doesn't get that.
 
I am excited for future updates of EAP, but I currently regret purchasing EAP.

Problems I have:
Phantom breaking - aggravating
Sometimes hugs lines on curves
Not so great in lanes that merge or spilt
Detects vehicles changing lanes too late and brakes to hard to compensate

I am constantly disengaging AP. It gets better with each update, but I dare not user it when my family is in the car.

Waiting for more innovation. Fingers crossed.
 
  • Like
Reactions: Magellan55
They are a bunch of lying fraudsters.....if you had THAT many cameras, why can't you post JUST ONE VIDEO of NoA cutting off a car, like you claim?

Tesla should sue them for fraud...I'll represent them for free.

50 bucks says all of those losers shorted Tesla stock 2 weeks ago.
 
NoA is a tool and a work in progress. Most new tools are feared, and their designers shunned (or worse). AP and NoA are better than they were, and continuing to get better. Use them wisely and carefully, and enjoy the rewards. Abdicate your responsibilities for using them to the tools themselves, and you will be surprised. Not all surprises are good ones.

Most of our automotive infrastructure was not designed for high speed travel. Frequent low speed crossing, dangerous potholes, and construction zones with no buffer zones for disabled vehicles create dangerous situations for all of us, AP and NoA included.

Where are the statistics for automotive accidents "by vehicle type, make, and model"? Why aren't these referenced as much as Tesla's infrequent accidents? Because they're "not news worthy"? We are a strange people, aren't we? /s
 
The article is about a software update. They own the car.

Yes, now they need to start testing these updates as updates, not the final state of the car forever and ever like they still are. They're comparing apples to oranges if they're testing Autopilot the same way they test Cadillac's Super Cruise. One is updated in perpetuity, and one is what it is for the most part.
 
Yes, now they need to start testing these updates as updates, not the final state of the car forever and ever like they still are. They're comparing apples to oranges if they're testing Autopilot the same way they test Cadillac's Super Cruise. One is updated in perpetuity, and one is what it is for the most part.
The article was a review of an update! They bought the car a while back, received an update, and then reviewed the update. I'm sure if Super Cruise receives an update then they'll probably review that too.
 
This YouTube video presents an interesting critique of the Consumer Reports recent article on Navigate On Autopilot limitations. It was well produced. I agree with some of the producers comments but he is definitely a Tesla fan who only presents a one sided point of view.

 
Driving in and out on I-66 has been problematic due to construction. I've missed two intersections where the published Google map has not been updated to the new interchange configuration (one temporary and the other permanent). It has also shut down after giving up on figuring out the new lane changes - might be a function of lane markers painted out black (sort of) and new lane shifts marked in.

Yes I watch the FSD carefully but the analogy of watching a newby driver is accurate.