Better quality video of the same video: https://www.youtube.com/watch?v=6QCF8tVqM3I

Same driver just posted a new video, this time during the day, in which AP behaves correctly: https://www.youtube.com/watch?v=bkDsozA1Pb8

That's the scary and also dangerous part: the behavior might be ok at some point, and might fail at another point. Tesla's (and NTSB) investigations will likely find out what went wrong in this particular accident and on this part of the road, and likely will improve the systems. But users are certainly not enough aware of that kind of risk: they might try the Autopilot, see that it works, and trust that if it worked, it will work again and again and again... On a daily commute for example, if it worked many days in a row, and then fails for some reason (technical, external, whatever), the user will likely fail to notice.

> On a daily commute for example, if it worked many days in a row, and then fails for some reason (technical, external, whatever), the user will likely fail to notice

Yup. This is why I won't get into a self-driving car until they're fully driverless.

Drivers become less attentive as the system gets better.

During the day there is a black gap clearly visible. During the night I imagine that black line just disappears.

Since nobody noticed it, it failed with a AP2 car in night time, and succeeded with a AP1 car in daytime

That is waaay too close for no one to be screaming in the car! I don't understand how the Tesla couldn't see the centre barrier but it can detect when a crash is about it happen, can someone explain that?

It can't detect when a crash is about to happen. The driver manually pushes the brakes there.

I have only worked tangentially with self-driving systems but in general they are very bad at dealing with stationary objects - this is why it can't see the center barrier.

This article describes the reason why radar and stationary objects don't mix: https://www.wired.com/story/tesla-autopilot-why-crash-radar/

How far ahead can LiDAR detect stationary objects? Will there be enough time on a highway to brake or steer away from suddenly appearing obstacles?

The latest Waymo video [1] says its car can see up to 3 football fields away. I would guess this would let it operate at higher speeds. I haven't seen any demos of it on the highway. I think they're focused on cities and suburbs.

[1] https://youtu.be/B8R148hFxPw?t=1m28s

Ok. Can it see potentially dangerous obstacles that stick out of the ground by, say, 10 inches?

It can see ducks..

https://www.youtube.com/watch?t=26m15s&v=Uj-rK8V-rik

> I don't understand how the Tesla couldn't see the centre barrier but it can detect when a crash is about it happen, can someone explain that?

AFAIK the AP focuses more on moving targets for collision avoidance, stationary objects are harder because there's so many of them you'll never hit since they're above or next to the road, but that's hard to tell until they're very close (can't see where the road leads to 100m away).

That's the very definition of "you're driving way too fast (and if that's your only detection method, you shouldn't be on the road)".