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 Useless (2018.42.3)

This site may earn commission on affiliate links.
My impression is that wk057 is spot on in his description of current NoA. I tried to use it this week-end, but I found it decelerated without reason every time I moved back into right lane - almost like brake-checking..... Overtaking is currently implemented as an afterthought. I had to quit using it midway through the trip, as on occasion I felt it created confusion to the other drivers.

I hope this kind of buzz gets back to Tesla so they can work on it.
 
  • Informative
Reactions: cwerdna
1. V9 (39.7 and 40) has big improvements for lane changing compared with V8. For V8, I need to check and make sure it is safe and there is no car in blind spot before turning the signal. For V9, I have tested in all different cases and I have built a trust that I will just turn the signal and let the car change it.

This is remarkably dangerous behavior, and a clear example of why when manufacturers add safety measures there is in increase in fatal accidents.

2. For NoA (42), it is very useful to not have to worry about missing exit or change to a wrong freeway, this itself is very helpful to me.

The request put to the public is to monitor the system and make sure it works right. You're already way too trusting of a system that the creators themselves don't fully trust. This is exactly what I was talking about in my earlier post.

4. I usually change lanes myself for over taking (just turn the signal, no need to check blind spot or traffic:), the car will change lane. If the car suggest a lane change to overtake, I usually confirm it immediately (one tap is easier than enraging the turn signal).

No need to check blind spot. This is the most insane behavior I can imagine. And to think, you're on public roads endangering other peoples loved ones. The good news is your post is here forever, so it can be used against you in a wrongful death suit.
 
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


By sheer luck today I used NoA on a destination that was just after the interstate turned in to a county road while it was raining out. I'm honestly shocked with how well it handled the transition. Not only did it know when to get over (twice) before the stop light, it took the turn and merged me over before it ticked down the counter and had me take over control.

I think we've all had our issues with NoA being beta for sure, but in this one instance I was floored with how well it performed.

 
  • Helpful
  • Like
Reactions: kbM3 and scottf200
Are you sending control commands or are you just calling a changeLaneRight() method?

It would be interesting if you can control the steering and throttle and brake.

I'm more interesting in the throttle/brake since that's safer.

I'm currently working on open source autonomous driving layer that you can feed perception data and it will be able to perform advanced lane changes, planning, merging, cut-ins. based on a model based reinforcement learning using mobileye's rss.

So even if you are just calling a changLaneLeft() and a setSpeed() function. I could probably still get it to work sufficiently.

With this system you would be able do a cross country drive (highway only) without ever touching any controls.

I still have some kinks I'm working out. But will be interested in working with u on ap2.

My custom code sits as an intercept between the DAS module, the SCCM, and the rest of the car. So any data seen by the DAS or the SCCM is able to be manipulated, and anything seen from these modules equally malleable.

I've broken most of this out into simpler functions, though, like initiateLaneChange(direction) and setCruiseTarget(speed_in_mph) and such. There is no direct acceleration/deceleration control. The DAS tells the motor and braking system what it wants via "I want to go X speed and I was to get there Y quickly" where X is a speed and Y is a maximum acceleration/deceleration. The system limits both to reasonable values.

Also can you access and grab frames from model 3 driver facing cam?

Not as of yet. It's still not enabled at all as far as I can tell.


Now that I have disabled lane change confirmations the experience is at least somewhat positive and I look forward to improvements in the future.

Disabled it?
 
My impression is that wk057 is spot on in his description of current NoA. I tried to use it this week-end, but I found it decelerated without reason every time I moved back into right lane - almost like brake-checking..... Overtaking is currently implemented as an afterthought. I had to quit using it midway through the trip, as on occasion I felt it created confusion to the other drivers.

I hope this kind of buzz gets back to Tesla so they can work on it.

I had this happening to me until I turned off HOV lanes. It was slowing down to try and make it into the HOV lane. Wayyy better after I toggled that off. Your luck may vary.
 
  • Informative
Reactions: Cirrus MS100D
I realize I haven't been super active here on TMC lately, but I'd suggest to the people immediately jumping to conclusions that you check out some of my post history for the past ~5 years before.

I love Tesla's products. They're the best of what they are, currently. Hell, I've purchased quite a few Tesla's and currently have... five?... on my My Tesla account (haven't bothered to update my signature in a while, either).

So people start with the name calling and other hating, perhaps do minimal research.
I can tell how excited you are about NOA because it brought you out of hiding! I don't have NOA yet, but I'm pleased with how quickly they are fixing things in V9 like the rear facing camera and UI features. I'd rather have a product we can discuss and enable and disable than waiting for it to be ready for prime time. Again, as long as they keep at it... I'll keep taking the updates.
 
My custom code sits as an intercept between the DAS module, the SCCM, and the rest of the car. So any data seen by the DAS or the SCCM is able to be manipulated, and anything seen from these modules equally malleable.

I've broken most of this out into simpler functions, though, like initiateLaneChange(direction) and setCruiseTarget(speed_in_mph) and such. There is no direct acceleration/deceleration control. The DAS tells the motor and braking system what it wants via "I want to go X speed and I was to get there Y quickly" where X is a speed and Y is a maximum acceleration/deceleration. The system limits both to reasonable values.



Not as of yet. It's still not enabled at all as far as I can tell.




Disabled it?

Sorry I wasn't clear just referring to the setting with disable/mild/average/mad max options. This at least reduces the number of error prone lane change confirmations that I have to deal with.
 
No need to check blind spot. This is the most insane behavior I can imagine. And to think, you're on public roads endangering other peoples loved ones. The good news is your post is here forever, so it can be used against you in a wrongful death suit.
CuriousG is yet another person who seems to not bother checking his blind spots before changing lanes given his derision of SMOG (Signal, Mirror, Over the Shoulder, Go) and describing changing lanes as "barely nudg(ing) the wheel". Maybe he doesn't check his mirrors either? See below.

Two days with Model 3 impressions/review- This is not a mini model S
Two days with Model 3 impressions/review- This is not a mini model S
Turn signal “discussion”
Turn signal “discussion”

Let's just hope he doesn't use NoA in the same way as SongjinDK.
 
Last edited:
  • Like
Reactions: DrDabbles
Mine is also on the verge of useless, requiring a number of rather abrupt manual takeovers as it second-guesses itself. I'd imagine there were a few adjacent drivers who assumed I was drunk or distracted over the last week. It has never tried to steer me into any other cars or objects though.

I've found that turning off the speed-based lane change suggestions improves NoA performance. The system never accepts that you're ignoring a suggested speed-based lane change so it slows the car while leaving a huge gap with traffic ahead, assuming that at some point you will accept its suggestion. Of course other cars pile in to the gap and the cycle continues.

For me the only time NoA is useful is when I'm in a lane that is splitting between two freeways (it successfully signals and takes the correct path each time) or if I'm already in the correct exit lane and then it will signal, take the exit, and slow. Otherwise NoA is just too "polite" for LA traffic and always second guesses its own lane changes.

As @mtndrew1 describes above, I've noticed two situations where phantom braking has occurred since the NoAP update (I have NoAP with lane change suggestions enabled with "Mad Max" on an AP2.0 MX with v2018.42.2):
  1. A car "changes lanes" into your lane, but so far ahead of you that hard braking should not be necessary. I've had this happen once with plain AP after a car turned left onto a two-lane city street I was driving on, and the car was probably ~2 city blocks ahead of me. The other time I think I saw this was after I disengaged NoAP, a car pulled out (~10-12 car lengths ahead of me) from a bumper-to-bumper lane on my left into the merge lane I was in, and I got the "red vehicle" warning (yes, about one second after disengaging NoAP). The plain AP case feels like a big regression as I've noticed AP reacting to other cars turning into my lane (or turning in front of me), and I have not had it overreact this much.

  2. At the end of an interchange (off-ramp to on-ramp between highways) where traffic to the left was moving markedly slower than I was going in the merge lane at the end of the interchange. I believe NoAP knows it needs to merge left (eventually) in these cases, so it starts trying to monitor the traffic to the left. However, if you're going 15-20 MPH faster than the traffic in the left lane (even with clear road in front of you that has no merge lines in sight), NoAP apparently brakes HARD so you can merge into the left-lane traffic instead of forging ahead (and slowing gently). This phantom braking is what makes NoAP so dangerous to use right now IMO. I'll have to try disabling lane change suggestions to see if that "fixes" (prevents) this issue.

NoAP isn't all bad. It's amazing when it changes into a new lane for an exit, or takes an exit properly. (Disclaimer: I live in California.)

However, I've also had other wacky behavior such as:
  • NoAP activated on a private road after I enabled Autopilot. There was no on-ramp in sight, but I got a single blue line—perhaps because the lane markings looked like some kind of an on-ramp?! Surprised me, but the vehicle performed the same as on AP, and eventually figured out it wasn't at an on-ramp and went back to plain AP. (There was an on-ramp later in the trip; just not on the private road.)
  • I've gotten red hands to take over while using NoAP. (I was expecting this behavior, so didn't really commit details to memory.)
  • There is one on-ramp (with a metering light on at certain times of the day) that has a speed limit of 35 MPH all the way up to where it merges with a 65 MPH highway. I always override that speed with the accelerator pedal, so I don't know what it actually does on its own.
  • There is an off-ramp (leading to an interchange) where the posted speed on the instrument cluster stays at 65 MPH through a cloverleaf with a posted 25 MPH speed limit, although the vehicle does slow down on its own.
  • NoAp completely missed an exit after it changed into a new lane properly. (I made the exit but turned late out of NoAP.)

I've also been good about filing bug reports after nearly every incident—sometimes more than once to get the bug report text close enough to make it understandable/searchable. Hopefully that won't fill up the disk space too quickly on my MX. :)
 
This is remarkably dangerous behavior, and a clear example of why when manufacturers add safety measures there is in increase in fatal accidents.



The request put to the public is to monitor the system and make sure it works right. You're already way too trusting of a system that the creators themselves don't fully trust. This is exactly what I was talking about in my earlier post.



No need to check blind spot. This is the most insane behavior I can imagine. And to think, you're on public roads endangering other peoples loved ones. The good news is your post is here forever, so it can be used against you in a wrongful death suit.

Some people may have misunderstanding of the confirmation of the current NoA. I think the confirmation is to confirm if the lane suggestion is correct. The car simply says "I need to change to that lane, but I don't need to change it right now if it is not safe, so if you think my suggestion is correct, please confirm, then I will change to that lane when safe." That's why it often suggests a lane change even if there are cars in that lane and the line is read. Don't worry it will hit those car if you confirm. As long as the suggestion is correct, we should confirm it even if there is a car in the blind spot (and the line turns red). What we should do is to watch closely when the car actually starts to change lanes, if it is not safe, we should take over immediately. I think this is to prepare the real NoA where the confirmation is not needed. Correct me if my understanding is wrong.

This is especially true in heavy traffic, it will suggest lane change when there are cars back to back in the target lane. If we wait until there is no cars in blind spot, we will never be able to change to that lane. Instead, we should confirm immediately and let the car turn on the blinker and trying to negotiate, eventually some car will kindly yield and let me in.
 
Last edited:
Some people may have misunderstanding of the confirmation of the current NoA. I think the confirmation is to confirm if the lane suggestion is correct. The car simply says "I need to change to that lane, but I don't need to change it right now if it is not safe, so if you think my suggestion is correct, please confirm, then I will change to that lane when safe." That's why it often suggests a lane change even if there are cars in that lane and the line is read. Don't worry it will hit those car if you confirm. As long as the suggestion is correct, we should confirm it even if there is a car in the blind spot (and the line turns red). What we should do is to watch closely when the car actually starts to change lanes, if it is not safe, we should take over immediately. I think this is to prepare the real NoA where the confirmation is not needed. Correct me if my understanding is wrong.

There have been numerous reports here and in other forums of the blind spot detection system not working. Blind spot detection isn't a replacement for using your eyes and neck to do the bare minimum amount of work to not endanger another person's life. It's meant to augment your normal behavior. People quite literally die this way, and you can't be bothered to turn your head to look. That doesn't seem dangerous at all to you?
 
  • Like
Reactions: cwerdna
Some people may have misunderstanding of the confirmation of the current NoA. I think the confirmation is to confirm if the lane suggestion is correct. The car simply says "I need to change to that lane, but I don't need to change it right now if it is not safe, so if you think my suggestion is correct, please confirm, then I will change to that lane when safe." That's why it often suggests a lane change even if there are cars in that lane and the line is read. Don't worry it will hit those car if you confirm. As long as the suggestion is correct, we should confirm it even if there is a car in the blind spot (and the line turns red). What we should do is to watch closely when the car actually starts to change lanes, if it is not safe, we should take over immediately. I think this is to prepare the real NoA where the confirmation is not needed. Correct me if my understanding is wrong.

This is especially true in heavy traffic, it will suggest lane change when there are cars back to back in the target lane. If we wait until there is no cars in blind spot, we will never be able to change to that lane. Instead, we should confirm immediately and let the car turn on the blinker and trying to negotiate, eventually some car will kindly yield and let me in.

When id doubt, read the manual ;)
<span class="warningtitle">Warning:</span> Navigate on Autopilot
may not recognize or detect oncoming vehicles, stationary objects, and special-use lanes
such as those used exclusively for bikes, carpools, emergency vehicles, etc. Remain
alert at all times and be prepared to take immediate action. Failure to do so can cause
damage, injury or death.</div>
<span class="warningtitle">Warning:</span> Never depend on
Navigate on Autopilot to determine an appropriate lane at an off-ramp. Stay alert and
perform visual checks to ensure that the driving lane is safe and appropriate.</div>

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.</p>

<span class="notetitle">Note:</span> If you ignore a route-based lane
change suggestion (for example, you are driving in the left lane while approaching an
off-ramp on the right side of the highway), Navigate on Autopilot will be unable to
maneuver onto the off-ramp. In this case, the navigation system re-routes you to your
destination.</div>
 
@tomc603
The question is not about blind spot detection, it is whether we need to wait until it can safely change lane before confirming the lane suggestion, which I don't think so. I do think we should watch closely when the car actually changes lanes and be ready to take immediate control.
 
Last edited:
  • Like
Reactions: dhanson865
Also I believe all the experiences shared here are fair and accurate but what is very interesting to me is that it seems the personal experiences are very much different
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.
 
  • Informative
Reactions: Cirrus MS100D