Video: Here's why you have to keep your eyes on the road when using Autopilot

[sorka]

Engagement chime and disengagement chimes were noticeable in the intro section, not perfectly clear but easily identifiable. Engagement nearest time of indecent is audible but only one of the two tones stands out above the road noise (on my desktop speakers), the other half of the engagement tone was too quiet vs ambient.

I'm not going to do it again tonight but the other night I put on headphones and listened to the seconds around the audio gap you referred to by time stamp and all I got was an audio gap and a turn signal. If the autopilot disengage sound is quiet enough that road noise covers it up such that I can't find it even with boosting my volume on a PC then Tesla needs to modify the sound of that chime (repeat it twice or play it once with more volume?)

You seem certain its there but I'm not hearing it. I guess we need a third party to be the tiebreaker.

So tell me this, since the engagement chime is so easy to hear, then where is it when the OP "took control"? The only possible answer if it isn't where I said it is, which is, is that there's a bug that prevented the disengagement chime from player, or the OP didn't actually take control and AP corrected itself.

Honestly, everyone in my household can easily hear it without headphones once they know where it is. It's not hard and another poster in this thread also said they heard it at that time stamp as well.

 

[sorka]

what kind of dash cam are you using? this is a very clear video, and I'd consider getting your brand.

Its the dual channel BlackVue 650 that most Model S owners seem to have. I have the same one myself.

 

[dhanson865]

So tell me this, since the engagement chime is so easy to hear, then where is it when the OP "took control"? The only possible answer if it isn't where I said it is, which is, is that there's a bug that prevented the disengagement chime from player, or the OP didn't actually take control and AP corrected itself.

That's the kicker there.

Just writing the prior reply I was thinking the same thing. I sure didn't hear the disengagement when I saw the car correct back into the correct lane.

But going back to your disengagement video you had smoother pavement, less road noise, and your music set to a higher volume.

Does the disengagement sound match the music volume? If he had his music turned way down would the disengagement chime be less noticeable vs road noise?

 

[Nick Deng]

Yesterday my family and I were driving on a divided freeway when a seam in the freeway started on the left side of my lane and then crossed across my lane. The Autopilot computer incorrectly interpreted the seam as a lane marker (which is odd because there were lane markers which were much more obvious). The autopilot took me into the lane next to me which was occupied by another car. I immediately corrected manually and went back into my own lane, but I'm sure I looked like a jerk to anyone behind me. I didn't hear the sensors detect the car next to me, and while I'd like to think they would have detected the other car and avoided a collision, I didn't wait long enough to find out. My wife was very upset because she assumed it was my driving (she's very protective of the Tesla for obvious reasons) and my son immediately chimed in with comments echoing mommy. You may enjoy a laugh if it resembles your family. The video is 1 minute long, the lane seam appears 26 seconds in.

I was hoping you'd be able to hear autopilot "disengage" tone when I corrected, but I can't hear it on the video. I should probably email this to Tesla so they can help improve the software. I'm a huge autopilot fan, and I realize Tesla is being attacked due to idiots failure to use it correctly, so I'm tempted not to share this. But I want the product to improve and I don't want any of my fellow Tesla fans to get in an accident, so that is my motivation for sharing this. The video is unlisted, so only people on the forums should be able to find it. 1st time I used my BlackVue for something useful. Glad I have it.

You do need to be careful at driving anyway. Here I learned few AP may not good at and to share with all: 1, if you are on the road mixed new and old , sometime no detect on new concrete surface. 2. Sometime line color not seen by AP cam? 3. Is very important, when come to a "Y" lane like car pool lane exit etc AP Confused and can not make decision for you!!! 4. If some road too wide. 5. Only one side of road has mark/line. And more... Be safe, be happy with your tesla!

 

[sorka]

That's the kicker there.

Just writing the prior reply I was thinking the same thing. I sure didn't hear the disengagement when I saw the car correct back into the correct lane.

But going back to your disengagement video you had smoother pavement, less road noise, and your music set to a higher volume.

Does the disengagement sound match the music volume? If he had his music turned way down would the disengagement chime be less noticeable vs road noise?

No. It's louder than the music volume. It's also just the very tail end of it because most of it happens during the audio dropout at the video transition.

 

[Garlan Garner]

No. It's louder than the music volume. It's also just the very tail end of it because most of it happens during the audio dropout at the video transition.

LOL... That's a tremendous amount of road noise. The lady in the car wasn't screaming. She and the other person were talking normal. In my car the chimes would have been louder than those people talking.

Whats going to be nice about this is that I believe in the future we will be able to quickly get black box proof of events. Much like what Tesla is doing to filter out those trying to fabricate stories.
Tesla Autopilot crash in Montana: Drivers reveals new details and claims a ‘cover up’ by Tesla

 

[Plaid21]

Yesterday my family and I were driving on a divided freeway when a seam in the freeway started on the left side of my lane and then crossed across my lane. The Autopilot computer incorrectly interpreted the seam as a lane marker (which is odd because there were lane markers which were much more obvious). The autopilot took me into the lane next to me which was occupied by another car. I immediately corrected manually and went back into my own lane, but I'm sure I looked like a jerk to anyone behind me. I didn't hear the sensors detect the car next to me, and while I'd like to think they would have detected the other car and avoided a collision, I didn't wait long enough to find out. My wife was very upset because she assumed it was my driving (she's very protective of the Tesla for obvious reasons) and my son immediately chimed in with comments echoing mommy. You may enjoy a laugh if it resembles your family. The video is 1 minute long, the lane seam appears 26 seconds in.

I was hoping you'd be able to hear autopilot "disengage" tone when I corrected, but I can't hear it on the video. I should probably email this to Tesla so they can help improve the software. I'm a huge autopilot fan, and I realize Tesla is being attacked due to idiots failure to use it correctly, so I'm tempted not to share this. But I want the product to improve and I don't want any of my fellow Tesla fans to get in an accident, so that is my motivation for sharing this. The video is unlisted, so only people on the forums should be able to find it. 1st time I used my BlackVue for something useful. Glad I have it.

Funny how things like that happen right when there is car next to you on the right. I wonder if there was no car on the right and you let the autopilot do its thing if it would actually completely change the line or come back to the same line.

 

[scottm]

..repeatable test.

Get 2 cars and go out and try.

Report back.

 

[Garlan Garner]

..repeatable test.

Get 2 cars and go out and try.

Report back.

Not necessarialy true.

My AP does not repeat the same driving responses consistently. - Using the same car.

 

[scottm]

Don't confuse a repeatable test with a consistent outcome.

It's a very repeatable test.

Variable outcomes is very interesting!

 

[ShadowR55]

Darn thing keeps blinking!!

 

[DrManhattan]

Just did my first road trip - a 1500 mile trip 95% using autopilot. I only had a few issues - 3 times on the way there, car swerved several times back and forth. A couple times were near semi's. After software upgrade on the way back, it didn't happen. On the way home, there was a huge silver bumper or ladder in my lane. I saw it in time and swerved out of the way. If I didn't see it, it would have caused massive damage. You always have to pay attention. Autopilot is a life saver for road trips - but only if you are PAYING ATTENTION!

 

[Garlan Garner]

Autopilot is a life saver for road trips - but only if you are PAYING ATTENTION!

I absolutely love your last sentence. That should be the slogan of AP.

 

[cinergi]

Wow. OK, so the sound at 14:19:26 ... yeah, no that's not the tail-end of AP disengaging. That same sound happens at least twice more between 14:20:46 and :51 ... it's something else. Give it up.

 

[Garlan Garner]

Wow. OK, so the sound at 14:19:26 ... yeah, no that's not the tail-end of AP disengaging. That same sound happens at least twice more between 14:20:46 and :51 ... it's something else. Give it up.

Very nice investigation.

 

[jackbowers]

No chime. Appears Autopilot corrected itself without disengaging right after it sensed the vehicle it was swerving into.

 

[HyperMiler]

Obviously, als stated many times in other posts, the symbol and chime for disengagement are way too subtle. Something to be easily corrected by Tesla.

BtW, this mobbing of people who are honestly trying to share their experience is gross. TMC is different, remember?

AP is great but disengages stealthily sometimes and would run you into the Armco (e.g. bend over hilltop) within a second if the conditions are not right.

There needs to be a chime and a large advertisement popping up, and not some anemic little steering wheel symbol changing into a camouflaged color.

 

[sorka]

Wow. OK, so the sound at 14:19:26 ... yeah, no that's not the tail-end of AP disengaging. That same sound happens at least twice more between 14:20:46 and :51 ... it's something else. Give it up.

Can you please point out the exact time indices you think you hear the same sound? I've listened to that 5 second range a half dozen times with headphones and do not hear the an fraction of the same tone that I hear at 14:19:26 after the transition.

 

[cinergi]

Can you please point out the exact time indices you think you hear the same sound? I've listened to that 5 second range a half dozen times with headphones and do not hear the an fraction of the same tone that I hear at 14:19:26 after the transition.

IIRC the second one was right at the :49 - :50 transition. The first one I think was at :47.
It's tricky because it's not prefixed by the audio dropout but they absolutely match the tail-end of the one at 14:19:26 after the dropout clears.

The fact that there's this much debate about the audio should tell you something, too.

 

[Garlan Garner]

IIRC the second one was right at the :49 - :50 transition. The first one I think was at :47.
It's tricky because it's not prefixed by the audio dropout but they absolutely match the tail-end of the one at 14:19:26 after the dropout clears.

The fact that there's this much debate about the audio should tell you something, too.

This is insane.

I performed a frequency read on the entire file. That's the easiest thing to do. I sampled my MS sound and got the frequency range. I then ran his video through my waves frequency meter and that frequency range only appears at the leading start of the video at 00.00.00.0021.