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

Autonomous Car Progress

This site may earn commission on affiliate links.
Nice quick overview of Mobileye products.


He says SuperVision, now deployed on the Zeekr001 in China, provides "hands free, point to point, navigation-on-autopilot, on all road types, up to 130 kph".

At the end of the video, he mentions that L4 on personal cars is currently being implemented on the Zeekr consumer vehicle in China, which will provide "eyes off navigation-on-pilot on highway, rural and arterial roads."
 
  • Like
Reactions: Dewg

Wow, pretty slow reaction time, probably about 1 second slower than an adequately attentive human. Note how the vehicle accelerates from 17-18mph even though it's clear the pedestrians are going to enter the crosswalk.
It's not an active pedestrian crosswalk when the light is red for them. There's no expectation that they would start walking across the road, nor should anyone be prepared for them to do so. There is no right to cross against a red light, that's jaywalking.

Sure, if you can clearly see someone intentionally or absent-mindedly walking into a green-light lane then it would be excellent to anticipate your stopping. That's not what we see here, this is two people waiting by a red-light and jumping into the street on purpose at the last second. They do hunch over like they might enter, but that's just a feint, aka JackAssery. This does not showcase anything other than their stupidity, and the fact that the car did manage to stop when they actually carried through with their feint and entered the roadway.
 
Last edited:
  • Like
Reactions: diplomat33
It's not an active pedestrian crosswalk when the light is red for them. There's no expectation that they would start walking across the road, nor should anyone be prepared for them to do so. There is no right to cross against a red light, that's jaywalking.

Sure, if you can clearly see someone intentionally or absent-mindedly walking into a green-light lane then it would be excellent to anticipate your stopping. That's not what we see here, this is two people waiting by a red-light and jumping into the street on purpose at the last second. They do hunch over like they might enter, but that's just a feint, aka JackAssery. This does not showcase anything other than their stupidity, and the fact that the car did manage to stop when they actually carried through with their feint and entered the roadway.
Again, as detailed above, given the lack of any traffic behind, there were clearly other acceleration and speed profiles that could have been used to minimize the risk presented by the (predictable) jack-a**ery.

And to be honest, since this sort of behavior is so predictable, I’d prefer to hear Cruise say: “we expected this, and in this scenario with obvious misbehavior and intent of pedestrians (which is predicted accurately by our planner) we have programmed the car to be on high alert, while not modifying the profile in response, in an attempt to dissuade people from this behavior (while not increasing risk of collision).”

That’s what I’d prefer to hear, rather than a spurious explanation that they don’t slow down due to possibility of rear-ending (which was zero risk in this case, and can always be calculated as an input to the planner).

Cruise should always be modifying planner behavior based on tailing vehicles - just like a human. The lead driver is usually the one who controls whether a rear end collision occurs. They have a similar amount of control over it as the trailing driver, through their behaviors and by creating a buffer (or removing the tailing car entirely!). This is an example of defensive driving!

As mentioned, Cruise planning seems to have a few weak points, and this explanation was not reassuring. I am sure the reality is that they do account for nearly everything in their planning, but still would like to hear it said.
 
Again, as detailed above, given the lack of any traffic behind, there were clearly other acceleration and speed profiles that could have been used to minimize the risk presented by the (predictable) jack-a**ery.

And to be honest, since this sort of behavior is so predictable, I’d prefer to hear Cruise say: “we expected this, and in this scenario with obvious misbehavior and intent of pedestrians (which is predicted accurately by our planner) we have programmed the car to be on high alert, while not modifying the profile in response, in an attempt to dissuade people from this behavior (while not increasing risk of collision).”

That’s what I’d prefer to hear, rather than a spurious explanation that they don’t slow down due to possibility of rear-ending (which was zero risk in this case, and can always be calculated as an input to the planner).

Cruise should always be modifying planner behavior based on tailing vehicles - just like a human. The lead driver is usually the one who controls whether a rear end collision occurs. They have a similar amount of control over it as the trailing driver, through their behaviors and by creating a buffer (or removing the tailing car entirely!). This is an example of defensive driving!

As mentioned, Cruise planning seems to have a few weak points, and this explanation was not reassuring. I am sure the reality is that they do account for nearly everything in their planning, but still would like to hear it said.

Disagree. This is not predictable behavior at all, at least in cities I've driven in. This is like an unwatched stroller suddenly getting blown by the wind into the street, almost never happens and it's a very back-of-the-mind possibility. Does every pedestrian commonly lunge into the street, and we have to expect that? It's freakishly rare behavior, or otherwise we'd never drive the speed limit through green lights. This is not an active pedestrian crosswalk when they are facing a red light.

Maybe I'm not understanding your point. Is there some new Cruise/Tiktok Dare Challenge out there? Because I would never drive like I expect every person to jump in front of my car. Why the h*ll would they do that? Why do you think people jumping into the street in front of a fast moving car is predictable behavior?
 
This is not predictable behavior at all, at least in cities I've driven in.
Watch the video again and tell me you could not tell what was going to happen before it happened!

I could, and without the benefit of hindsight.
Is there some new Cruise/Tiktok Dare Challenge out there?
Yes, clearly. I’m not talking about a scenario that would occur with a human driver (those would be different).

That’s what I’m talking about being able to predict and account for. This is definitely something that is going to happen to autonomous vehicles so it’s something they have to be very careful about dealing with.

I’m not convinced the strategy of “continue sailing along blissfully at 18mph towards a cross walk where people are obviously going to jump in front of the car, when no one is following the Cruise vehicle” is the optimal one for minimizing collision risk.

When someone does this and gets hit by the autonomous vehicle…it’s going to be the autonomous vehicle’s fault! Not saying the pedestrian shouldn’t have expected to be hit, but short of setting up a situation where it is physically impossible for the vehicle to stop in time (easy, just jump out from behind a truck on a tight street with perfect timing!), it’s the autonomous vehicle’s fault.
 
Last edited:
Ridiculous and atypical if true, but if that's where we are...
This is what college and high school kids do literally all the time. It’s fun, and their incomplete brain formation keeps them in perfect bliss.

Anyway Cruise and Waymo are already dealing with this, the only question is whether they have fully committed to taking responsibility for any collisions that occur (which they should, for their own selfish reasons!).
 
This is a great video by Engineering Explained that looks at how and when autonomous vehicles should break the rules to handle certain driving cases. The video is sponsored by Mobileye so it looks at how Mobileye's driving policy handles this question. I found the video to be super informative and easy to understand. And it shows lots of interesting and difficult driving scenarios and shows how Mobileye's autonomous driving handles them.

 
Last edited:
I would be happy with a vehicle that can drive itself through Atlanta on I 75 so I can just zone out. Then just wake me up when we get to Dalton going North or Macon when going South
Greyhound. $16 from Atlanta to Dalton. $18 to Macon. Certainly, not Tesla for the foreseeable future. Waymo/Cruise maybe someday. Maybe. Someday.

Amtrak would be nice, but not available for the route you want.