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

Navigate on Autopilot is Going to Cause Accidents

This site may earn commission on affiliate links.
Status
Not open for further replies.
In the few times I've used it I've had to take over a number of times, maybe sometimes it would have been successful, but it just makes me uncomfortable in traffic. EAP in general is great for really light traffic and really heavy traffic (stop and go). For heavy traffic traveling near the speed limit it still seems sketchy at best unless you just camp out in one lane.
 
Nav on AP is NOT going to cause accidents. Drivers relying on a Driver "Assist" system too heavily will cause accidents.

I think it's great that Tesla released this feature as it needs real world data to be able to improve. As with AP, it's NOT a driver "replacement". It simply is a tool that's there to assist the driver. Sometimes the assist may be better than no assist at all, but that's just part of the learning curve. The other option is for Tesla to not release it until it's 100% ready. Which means they have to do all the testing themselves. To rack up that many test miles on their own would take years and years and years.

If you follow the instructions and do not abuse the system, hoping it will be more than what it really is, then you'll be fine. Unfortunately, it takes responsible discipline. It does do some things very well, which naturally lures drivers into a false sense of security. They then get lazy and start to trust it too much. Need to always remember, it has significant weaknesses and will need the driver to take over on MANY occasions.

Those that use it responsibly, will get some enjoyment out of it. Those that choose to ignore instructions and place their lives in the hands of a system that they expect to be more than what it is, will very possibly end up in an accident or worse. Remember, AP or Nav on AP is a CHOICE. You don't have to use it. But if you do CHOOSE to use it, then use it responsibly. Doing so, will assist Tesla in developing the system to be better and better.

The release and capability of Nav on AP with its clear weaknesses reflects how hard it truly is to develop fully autonomous vehicles. Similiar to going to Space, IT'S NOT EASY. Tesla releasing this stuff to us in BETA mode is cool as it allows us to see first hand where we're at in the progress towards FSD. Clearly we're still many, many, many years away from being able to hop in the back seat and be just a passenger. In the meantime, am enjoying what they do give us, even if it scares the crap out of me sometimes!!
 
Speaking of which, what situations would you want to have a follow distance of 7?!?! I started with 4, and it felt like an eternity away. 3 was better, but I think 2 is just enough.

Ah, my favorite topic. 7 is too close for me. I would probably select 10 if 10 were available. According to Pennsylvania law, I am supposed to leave 4 seconds worth of distance between my car and the car up front. Which is almost 500 ft at 80 mph. At 7 I am nowhere near that distance behind the car up front. I have heard many times from people in other states that they have problems with people cutting them off if they leave more than a few car lengths to the car in front of them. Where I drive I find that it is perfectly reasonable to drive with AP set at 7. Almost no one cuts in front of me. Either people here are more polite, or the roads are not as congested, or both.
 
Tried using NoA for three consecutive days commuting to work and unlike Autopilot which I love to use, NoA is scary and unpredictable. To me, it is too stressful to monitor. Be hyper vigilant. I’m giving up until I read consistent reports on TMC that it is working great and a true stress-reducing driver's aid.
 
  • Like
Reactions: insaneoctane
I will continue to use it when I can to help Tesla learn. It is just very limited. It is not useful in heavy traffic. Of course I must say even EAP can be out of it's league in extremely heavy traffic merges. EAP can't maintain the requisite infinitesimal gap necessarily to prevent cars from cutting in non-stop during heavy merges. When I transition from 101S to 405S during rush hour, it is an extremely busy transition with 2 of the 101S lanes transitioning to the right onto the 405S. If I used EAP while I sat in either of the right 2 lanes I would hardly move an inch as EAP (even on a distance setting of 1) isn't willing to prevent drivers from cutting in front. So I know it's limits and use it accordingly. NoA also has limits, particularly in heavy traffic. No way would I move to the far right lane for a transition that is 2.5 miles of parking lot traffic even though NoA thinks that I should. I am more aggressive than that and will merge somewhere closer to 1/2 to 1 mile and literally save 20 minutes over what NoA wants. I realize that NoA has limitations and it is currently of limited value on my busy commute. This sure creates an interesting problem for Tesla to try and solve!
 
I will continue to use it when I can to help Tesla learn. It is just very limited. It is not useful in heavy traffic. Of course I must say even EAP can be out of it's league in extremely heavy traffic merges. EAP can't maintain the requisite infinitesimal gap necessarily to prevent cars from cutting in non-stop during heavy merges. When I transition from 101S to 405S during rush hour, it is an extremely busy transition with 2 of the 101S lanes transitioning to the right onto the 405S. If I used EAP while I sat in either of the right 2 lanes I would hardly move an inch as EAP (even on a distance setting of 1) isn't willing to prevent drivers from cutting in front. So I know it's limits and use it accordingly. NoA also has limits, particularly in heavy traffic. No way would I move to the far right lane for a transition that is 2.5 miles of parking lot traffic even though NoA thinks that I should. I am more aggressive than that and will merge somewhere closer to 1/2 to 1 mile and literally save 20 minutes over what NoA wants. I realize that NoA has limitations and it is currently of limited value on my busy commute. This sure creates an interesting problem for Tesla to try and solve!

LOL, so while everyone else patiently waits in line for the transition to the next freeway, you're one of the ones that speeds past it all and then cuts in yourself?? LOL. And then you say you want the car to be an ass and not let anybody in? LOL. Basically, you're saying you don't want the car to let drivers like you cut in!! LOL

Not criticizing, but just had to laugh at the irony of your message!
 
So...I had the opportunity to try "Navigate on Autopilot" and I survived the experience. However, I had to take control numerous times when the car did some pretty dodgy maneuvers.

The weakness of these driver assist technologies: cruise control, adaptive cruise control, autosteer, and now "navigate on autopilot" is that to use them safely, I think the operator has to be engaged with the road and traffic environment and *anticipate* when the system is going to fail - and then take over in plenty of time. This avoids putting the system in a situation it can't handle.

For example, it is pretty intuitive that if you are using plain old cruise control that if you are coming up to stopped traffic you need to engage the brakes and stop. Less obvious, but still pretty basic is that if you are behind a slow moving vehicle on ACC, if that vehicle moves out of your lane - your ACC is going to make you accelerate. At each step, however, it takes more and more experience to figure out when the autosteer (or now "navigate on autopilot") is going to fail.

So where did "Navigate on Autopilot" get dodgy? First, lane changes to faster lanes.The system now suggests a lane change when maybe it is still too complex a traffic situation that I wouldn't have tried on plain old autosteer. A car will zoom up or change into a lane ahead from a far lane. I'm not sure autopilot is really aware of cars in the *next* lane over. Then it starts hesitating. And then I have to take over when if is waffling between lanes. Not good.

Then, there is exiting the freeway. The system now leaves you in the left fast lane for a while. Of course, you have to get in the right lane ahead of your exit. The dodgy situation is that the exit you want is 2 or 3 down from a series of entrances letting in a lot of traffic into the right lane. Navigate on Autopilot didn't get me into the right lane ahead of all that incoming traffic. Rather, asked me to merge in. I don't think it sees the cars on the on ramp until they are one lane over. So once again, hesitation and aborted lane change.

Lastly, there is auto exit onto an off ramp. You better be ready to take over and hit the brakes. Under plain auto steer, the inattentive driver misses his exit. Now he's heading to the traffic stop.

So yes, I get it- you are supposed to be watching. And I think I will be able to figure out when and how I could trust this "navigate on autopilot" and *liberally* take over. The question is: Will the average Joe think this a whole lot more capable than it is - and then get caught in some dangerous situation prior to learning the should have anticipated a failure a lot earlier?

So how has hat it worked for other people? What level of trust do you have?
Frankly this is the reason I did not buy EAP. For me it takes more mental energy assessing the road, assessing if EAP is correctly assessing the road, and continuously deciding if corrective action must be taken than just driving myself.

Driving myself takes ~10% of my brain power because the entire decision making system has been habitualized from 35 years of driving. Often I get to a place and don't even have a memory of how I got there because I operate on such a high level of biological automation. Babysitting an infant driver (i.e. EAP) takes a lot of effort and brain power -- at least three times as much for me.

My opinion is that yes EAP is not only not worth it but yes it can make drivers more dangerous in its current form. They should be paying us to use it since we are collecting data for them and taking a big risk. 20 years from now when the baby neural network grows up and can finally drive as good as a human or better and I can sleep or work while the car drives itself then it may be worth the purchase.
 
"Beta" is just Tesla trying to limit their liability; which I don't think is sound since it's enabled regardless. They really should require drivers take an online course (aka 'watch it like a hawk') prior to activation of AP or any newly introduced features.

Respectfully disagree; NoA feels like a classic beta product in the software development sense of the word. Generally feature complete but a fair number of bugs to be worked out. Autopilot really is an open beta. First, because of the novelty of having betaware in an automotive product and second, because autopilot is generally feature rich, people seem to forget that in classical software development a beta product has bugs by definition.
 
Ok, v42.3, spent the weekend with 4 drives around Orange Co (55, 57, 22, 5, 91).
It's going to take some getting used to before knowing how to goose it just right. Let's share ideas how to dance with it. It's definitely more stressful but can become reliably predictable probably after I'm over the learning curve.
Some initial thoughts:
1. If you know exactly were you're going adaptive cruise , auto steering is enough. NAP. I Can change lanes when I want . Then HOV is ok. I know when to start getting over.
2.. If I'm not as sure about the freeway transitions or exits and on NAP, I think HOV to often puts you too far away from the next exit/transition reliably. So I turn HOV off.
3. The predictable over braking with auto lane change seems to call for a foot on the accelerator to compliment it as soon as the lane change confirmation is done. Keeps working when overriding the over aggressive slowing. Agree, if too tightly spaced still won't change and sometimes interprets cars 2 lanes over as a reason to abort. (And generally having a foot near the accelerator for those few dramatic slowing for no reason events.)
4. Generally, the biggest complaint I have while on NAP is I'm spending too much time having to glance at the screen and squint to see the messages or gray lines.
a. Would love BIGGER FONTS on warnings or option for bigger font!
b. Destintive chimes for must take lane changes, and maybe nothing for suggestive ones (does it do this already; someone mentioned it, maybe I just didn't notice yet?)
 
  • Love
Reactions: MelaniainLA
Lane change control needs a lot of work, I think the main problem is every lane change maneuver is performed with the same amount of urgency. It should modulate aggressiveness (i.e. willingness to change speed, lateral movement rate, blind spot gap allowance) based on how urgent the change is needed, like if it's just intended for passing or miles before taking an exit, you can just maintain your speed until an opportunity passively presents itself. With the way it works now, it drives obnoxiously and abruptly slows so it can cut into gaps. It would be better at playing Tetris than driving.
I think at present, when confirming a lane change, I goose it a little on the throttle to make it smooth. Feels more natural. But, if looking at the screen and out to the right and applying gas, watch for the sudden unexpected slow down of car in front of you that the car would have slowed down appropriately for and is now over ridden. (My sore arm from my wife hitting it attests to that risk ;) )
 
I've driven in many places known for bad traffic and aggressive driving , and what's interesting is that while there are some things that they have in common, there's a unique flavor to the aggressiveness. This is particularly noticeable once you go international.

In every locale, there's a collective understanding on what is acceptable assholish behavior. If you deviate from that, like applying another locales assholishness to the current one, you are much more likely to cause a disruption (honking, excessive road rage, accidents, etc).

And so it is with AP. Don't have the expectation that it can handle heavy/aggressive driving, because that's the holy grail of autonomy in a human-dominated arena. It's going to be a long while before the AI is smart enough to emulate a particular locale's assholishness and fit right in.
 
I think at present, when confirming a lane change, I goose it a little on the throttle to make it smooth. Feels more natural. But, if looking at the screen and out to the right and applying gas, watch for the sudden unexpected slow down of car in front of you that the car would have slowed down appropriately for and is now over ridden. (My sore arm from my wife hitting it attests to that risk ;) )

I actually turn the thumb well and set the speed up 5 mph when doing a lane change. Seems much more natural and tends to create the move ahead and merge instead of slam on the brakes and merge that seems more natural.
 
LOL, so while everyone else patiently waits in line for the transition to the next freeway, you're one of the ones that speeds past it all and then cuts in yourself?? LOL. And then you say you want the car to be an ass and not let anybody in? LOL. Basically, you're saying you don't want the car to let drivers like you cut in!! LOL

Not criticizing, but just had to laugh at the irony of your message!
I don't think it is ironic. I am not in a hurry to wait in line. I'm also not the jerk who cuts in at the last minute. I am somewhere in the middle and want to make sure that there are no more breaks before I get into the line. Statistically there is a variety of flavors. My point is that NoA is the extreme early type and I am not.
 
Last edited:
I think whenever you guys are talking about a complaint with this feature, you should always mention the version software you are on.

I am on 2018.42.3 and I've had it go into an exit lane for the exit prior to the one I needed. After being in that lane for a few seconds it had me switch out, but it was too late and I had to take that exit. Got to take the long way to work. Also coming down a long exit ramp, it was going a lot faster than I would have gone, because there was an upcoming turn.

I use regular autopilot every day and love it. But with Nav on Autopilot the way it is right now, I won't be using that much until it gets smarter. Hopefully it can at least learn from my driving habits, as I know which lanes are fast and which lanes are slow already on my routes to and from work.
 
I think they should just program it to change lanes like most other California drivers. No blinker, no look, just swerve into the next lane whenever the hell they want to. If there's a car there, screw them. What the hell were they thinking being on your road and in your way?! :)

OMG, had a guy do this to me a few weeks ago. Entered the freeway and merged inches in front of my bumper with me braking for him, then without pause or blinkers, merged INTO the guy in the fast lane causing him to have to take the shoulder and slam on the brakes. The car that caused all that just happily sped along in the fast lane apparently oblivious to the chaos he had caused behind him.
 
OMG, had a guy do this to me a few weeks ago. Entered the freeway and merged inches in front of my bumper with me braking for him, then without pause or blinkers, merged INTO the guy in the fast lane causing him to have to take the shoulder and slam on the brakes. The car that caused all that just happily sped along in the fast lane apparently oblivious to the chaos he had caused behind him.
In CA? No way! :)

JK, everywhere nowadays.
 
First try of Navigate on AP. Worked fine for most of my trip.

Then I was getting close to my offramp.

A 1/2 mile from my offramp, it wanted me to change lanes left, which would be kind of dumb since I was about 30 seconds from making a right off the freeway.

I don't think it ever got to a point that it wanted to take the offramp, I couldn't really tell if it did or not, or it was going to stop and hit the median.
 
Nav on AP is NOT going to cause accidents. Drivers relying on a Driver "Assist" system too heavily will cause accidents.

I think it's great that Tesla released this feature as it needs real world data to be able to improve. As with AP, it's NOT a driver "replacement". It simply is a tool that's there to assist the driver. Sometimes the assist may be better than no assist at all, but that's just part of the learning curve. The other option is for Tesla to not release it until it's 100% ready. Which means they have to do all the testing themselves. To rack up that many test miles on their own would take years and years and years.

If you follow the instructions and do not abuse the system, hoping it will be more than what it really is, then you'll be fine. Unfortunately, it takes responsible discipline. It does do some things very well, which naturally lures drivers into a false sense of security. They then get lazy and start to trust it too much. Need to always remember, it has significant weaknesses and will need the driver to take over on MANY occasions.

Those that use it responsibly, will get some enjoyment out of it. Those that choose to ignore instructions and place their lives in the hands of a system that they expect to be more than what it is, will very possibly end up in an accident or worse. Remember, AP or Nav on AP is a CHOICE. You don't have to use it. But if you do CHOOSE to use it, then use it responsibly. Doing so, will assist Tesla in developing the system to be better and better.

The release and capability of Nav on AP with its clear weaknesses reflects how hard it truly is to develop fully autonomous vehicles. Similiar to going to Space, IT'S NOT EASY. Tesla releasing this stuff to us in BETA mode is cool as it allows us to see first hand where we're at in the progress towards FSD. Clearly we're still many, many, many years away from being able to hop in the back seat and be just a passenger. In the meantime, am enjoying what they do give us, even if it scares the crap out of me sometimes!!
This

You are still "pilot in command" no matter what. Until such time as they remove a steering wheel and pedals, you are responsible for the actions of the car. If using the system makes you uncomfortable or feel unsafe then don't use it. It's that simple.

Dan
 
Status
Not open for further replies.