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.
We have learned through a trusted source that FSD v12 has started rolling out internally with Tesla update 2023.38.10.

FSD v12 is the update that is expected to remove "beta" from the title. The initial rollout to employees appears more limited in scale than previous updates. Considering the magnitude of the changes in this version, it makes sense to start slow.


NioVfkM.jpg
 
Last edited:
  • Informative
Reactions: JB47394
Gotta wonder if that was one-trick-pony SW.
Right. Impressive, especially backing, but it feels set up. It passed a couple of open parking spaces before choosing what it did.

So many questions - was the garage pre-mapped or did it read signs? Why did it leave in a different direction that it came from? Does it understand handicapped/reserved parking signs? What does it do if it cannot find a spot?
 
  • Like
Reactions: rbt123
Right. Impressive, especially backing, but it feels set up. It passed a couple of open parking spaces before choosing what it did.

So many questions - was the garage pre-mapped or did it read signs? Why did it leave in a different direction that it came from? Does it understand handicapped/reserved parking signs? What does it do if it cannot find a spot?
I can't find the video now, but I remember watching another Chinese system years ago with advanced auto park and the car is driven in the parking lot to map it out. I don't remember however if it is manually driven during this step or if the car does it automatically. It is not pre-mapped in the sense of the car company going there with a different car and mapping the lot.

Tesla did a similar demo in the past for AI Day back in 2021 for mapping out a parking lot's spaces:
1*KmJrVmYaEjR9xJH32r0nOA.png
 
  • Helpful
Reactions: scottf200
I can't find the video now, but I remember watching another Chinese system years ago with advanced auto park and the car is driven in the parking lot to map it out. I don't remember however if it is manually driven during this step or if the car does it automatically. It is not pre-mapped in the sense of the car company going there with a different car and mapping the lot.
NIO has similar feature; you drive once for it to map and learn the parking lot then it can park by itself in that same parking spot every time when you pull into the entrance, but you can't summon it to come back to your location, the Huawei feature is more advanced and seems to be a true valet parking feature that parks itself and comes back when you need it to.
Tesla did a similar demo in the past for AI Day back in 2021 for mapping out a parking lot's spaces:
1*KmJrVmYaEjR9xJH32r0nOA.png
That's not the same thing. They are discussing about Monte carlo search tree algorithm for path finding. The simulated environment is a parking lot.
 
  • Helpful
Reactions: scottf200
As we can see from the map, Mobileye has mapped the majority of US roads in just 10 months.

Slight correction: they've said they've mapped "the majority of the US" not "the majority of US roads." Big distinction, especially on such a small map with relatively thick lines representing the roads.

I'm interpreting "majority of the US" to mean they have mapped roads in at least 51% of the areas of the US. The CIA World Factbook estimates there are over 6.5 million km of roads in the US. I don't think Mobileye is claiming to have mapped over 3.2 million km.
 
Slight correction: they've said they've mapped "the majority of the US" not "the majority of US roads." Big distinction, especially on such a small map with relatively thick lines representing the roads.

I'm interpreting "majority of the US" to mean they have mapped roads in at least 51% of the areas of the US. The CIA World Factbook estimates there are over 6.5 million km of roads in the US. I don't think Mobileye is claiming to have mapped over 3.2 million km.

Point well taken. Thanks. From the map, it looked like most roads are mapped, especially on the east coast. But you are right that the lines are thick and might not be showing every road. And, a lot of the north west part of the US is mostly empty so there are not a lot of roads there anyway.

I am still impressed that they have mapped as much as they have in only 10 months.
 
Video segment of Europe and the USA with those map from 1 day to 10 months here:

16:50 minutes in

Video from this page: The Road to the Future of Mobility is Being Mapped by REM™ | Mobileye Blog

Thanks. I do think it shows the power of crowdsourcing. I still think Tesla should adopt this approach. Tesla has a big enough fleet and the cars are all equipped with cameras. Tesla could build their own map in a few months that is shared with the fleet and could help FSD beta a lot IMO.
 
  • Like
Reactions: KArnold
Point well taken. Thanks. From the map, it looked like most roads are mapped, especially on the east coast. But you are right that the lines are thick and might not be showing every road. And, a lot of the north west part of the US is mostly empty so there are not a lot of roads there anyway.

I am still impressed that they have mapped as much as they have in only 10 months.

I think the best way to present this kind of data would be the contiguous areas of addresses reachable by the areas mapped. Kind of like a geofence map for a hypothetical robotaxi running on the data. For an individual, it doesn't matter much if Mobileye has mapped all of the highways through their state; they need to have mapped the exits and local roads in order for a useful robotaxi to actually take them from any one address in their area to another.
 
For an individual, it doesn't matter much if Mobileye has mapped all of the highways through their state; they need to have mapped the exits and local roads in order for a useful robotaxi to actually take them from any one address in their area to another.

Sure. But the maps are generated from cars driving around. People don't drive on highways and them stop and not take the exit. So I think we can assume that if the highways are covered, the exits and local roads are covered as well. But the bottom line is that Mobileye will probably work to cover all those roads before deploying their system since they know that consumers will need to be able to drive on those roads. I doubt Mobileye would deploy a system that has too many "holes" in the map where the system would not be useful to the consumer.

And the map is not meant to suggest thst mobileye is done with mapping, but to show what their mapping approach can do in 10 months.
 
Last edited: