Autonomous Car Progress

[diplomat33]

.

It just makes sense for them to leverage Google Maps and Waymo HD Map technology to provide HD Maps for 3rd party ADAS and higher vehicles. HD Mapping is a scalable business.

Definitely. It makes perfect business sense for Google.

 

[diplomat33]

From Mobileye CES talk here is diagram of Mobileye's ADAS and AV products with ODD, sensors and compute for each.

 

[diplomat33]

I like Mobileye's "levels of autonomy" that are focused on what the consumer needs to know.

And defined useful ODD:

 

[diplomat33]

Mobileye CES talk:

IMO, it was super interesting.

Things I loved:

Mobileye defined consumer based "levels of autonomy" and ODD.
They showed us the exact sensors, ODD, compute etc of all their AV products.
They have a path to scale from consumer to robotaxi. They have a path to scale ODD.
They explained how they validate safety and what their safety goal is.
They showed some good examples of SuperVision handling difficult cases.
They have clear business model and timeline for both consumer cars and robotaxis.
Redundancy in sensors

 

[gmta10101]

Mobileye CES talk:

IMO, it was super interesting.

Things I loved:

Mobileye defined consumer based "levels of autonomy" and ODD.
They showed us the exact sensors, ODD, compute etc of all their AV products.
They have a path to scale from consumer to robotaxi. They have a path to scale ODD.
They explained how they validate safety and what their safety goal is.
They showed some good examples of SuperVision handling difficult cases.
They have clear business model and timeline for both consumer cars and robotaxis.
Redundancy in sensors

That was an interesting watch.
Their confidence makes me hesitate because it's impossible to be that confident this early on. At this point, both tesla and Mobile eye, both don't know what they don't know. They are technically just predicting thry will solve all these issues eventually.


The other thing that has me scratching my head is that they are saying it's bad to take steps from highway to onramp offramp, then to city and urban as moonshots and instead just "step 5 it and draw the rest of the owl, first lol)

It's a little humorous that this is a novel idea, just start at the end and complete everything first, then everything else is easy. It's a monumental task regardless if you start at highways and develop the OOD, vs starting at the end and doing everything, you kind of have to do the same engineering work, just faster.

 

[Mardak]

From Mobileye CES talk here is diagram of Mobileye's ADAS and AV products with ODD, sensors and compute for each.

This large variety of Mobileye's product offerings sounds like automakers are still hesitant about adding more advanced driving assistance/automation features and are looking for the cheapest options not to fall behind the competition, so they're "just" matching the market demand. It's smart for Mobileye to be able to support all sorts of customer requirements while making the most reuse of what they have and want to build towards, and ideally this brings safety on the roads sooner with wider deployment than OEMs doing it on their own. I suppose also good business for them to sell their products to those who are competing with each other as they'll naturally want to come back to Mobileye to beat / match the capabilities of what's coming on the market.

It'll be interesting to see the portfolio demand distribution over time especially if a serious competitor achieves wider deployment towards the "no driver" side of things.

 

[diplomat33]

The other thing that has me scratching my head is that they are saying it's bad to take steps from highway to onramp offramp, then to city and urban as moonshots and instead just "step 5 it and draw the rest of the owl, first lol)

It's a little humorous that this is a novel idea, just start at the end and complete everything first, then everything else is easy. It's a monumental task regardless if you start at highways and develop the OOD, vs starting at the end and doing everything, you kind of have to do the same engineering work, just faster.

Shashua's "moonshot" slide felt like a strawman to me. I can't really think of who he was talking about. I am not really aware of anyone starting from scratch each time they want to jump a SAE level. Maybe he was talking about legacy automakers that deploy a year model with certain hardware for say L2 highway and then design another year model with different hardware for L3. The slide was designed to make Mobileye's incremental approach look better because obviously incremental will be better than having to "moonshot" every time you want to expand the ODD.

This large variety of Mobileye's product offerings sounds like automakers are still hesitant about adding more advanced driving assistance/automation features and are looking for the cheapest options not to fall behind the competition, so they're "just" matching the market demand. It's smart for Mobileye to be able to support all sorts of customer requirements while making the most reuse of what they have and want to build towards, and ideally this brings safety on the roads sooner with wider deployment than OEMs doing it on their own. I suppose also good business for them to sell their products to those who are competing with each other as they'll naturally want to come back to Mobileye to beat / match the capabilities of what's coming on the market.

It'll be interesting to see the portfolio demand distribution over time especially if a serious competitor achieves wider deployment towards the "no driver" side of things.

I agree. I think Mobileye knows that different automakers have different requirements. Some might just want a L2 hands-free system for highways while others might want to be more bold and sell a L3 or L4 system. So Mobileye wants to cover all their bases.

 

[scottf200]

Mobileye defined consumer based "levels of autonomy" and ODD.
They showed us the exact sensors, ODD, compute etc of all their AV products.
They have a path to scale from consumer to robotaxi. They have a path to scale ODD.
They explained how they validate safety and what their safety goal is.
They showed some good examples of SuperVision handling difficult cases.
They have clear business model and timeline for both consumer cars and robotaxis.
Redundancy in sensors

I am testing FSDbeta fairly recently and have watched a huge number of POV videos over the past few years ... just to give background to the below comments.

Mobileye seems incredibly more professional and thorough with their planning, breakdown, use cases, etc. They exude believability. By comparison, it seems like Tesla is seat-of-the-pants ... here is a good guess on what sensors we need and then it was PR'd as "FSD". There have certainly been many improvements along the way but so many limitations and problems show up in the real world.

At some point, I think Tesla's current hardware, defined as "FSD" will fall short for city driving. It is a wait-n-see for HW4, but all the buzz about the previous (current) cars being sold can do FSD does not seem realistic in a congested environment in the march of 9s sense. Sure it can handle 9X% of the cases but I'm not sure the masses and industry will truly defined that as "Full Self Driving / FSD". I think the stats will show they are 'safer' than humans but no way will it be considered genuinely 'Full Self Driving.'

 

[Supcom]

Mobileye seems incredibly more professional and thorough with their planning, breakdown, use cases, etc. They exude believability. By comparison, it seems like Tesla is seat-of-the-pants ... here is a good guess on what sensors we need and then it was PR'd as "FSD". There have certainly been many improvements along the way but so many limitations and problems show up in the real world.

Mobileye has to be more open about their products because they want to sell to OEMs for integration into their vehicles. OEMs are not going to commit to an outside vendor's solution without some assurance that the vendor has a viable product. Tesla integrates into their own vehicles, so they have no need to publicly disclose more than they feel is necessary to sell FSD to consumers and to attract engineering talent.

So, Mobileye needs to market their product differently from Tesla. The fact that Tesla does not share a great deal of information outside the company does not mean that they are just winging it.

 

[scottf200]

From Mobileye CES talk here is diagram of Mobileye's ADAS and AV products with ODD, sensors and compute for each.

So, Mobileye needs to market their product differently from Tesla. The fact that Tesla does not share a great deal of information outside the company does not mean that they are just winging it.

@Supcom , so do you think their current hardware of 8 cameras (radarless now) at their current locations on the car will do "Full Self Driving" in any major city in the world where they sell Teslas?

Thanks for the perspective on why Tesla is closed lipped but I think regulators giving out "FSD" and robotaxi in the USA, EU (we've seen this aspect), and elsewhere will let Tesla get away with "trust us" and look at the accident rate vs humans stats?

Now, please don't get me wrong, as I'm a fan of FSD, I just think it is way oversold.

It would be interesting to see a Zeekr with Mobileye SuperVision and a Tesla with HW3 FSD do the same routes in different cities.

https://cars.usnews.com/cars-trucks/features/zeekr-automotive-coming-to-america

 

[scottf200]

This was a pretty interesting graphic of the camera placement and degrees of visibility (FOV) of Mobileyes SuperVision requirements on their partner Zeekr.

- Of note: front camera (2 vs 3), Side Front Camera on mirror vs B-pillar.
- "7 long range cameras at 8 MP each" 47:55 seconds into video.
- Side cameras: 100 degree vs Tesla at 80 or 60 degree.
- Parking: 3 cameras at 195 degrees 'down low'. (obvious benefit is when things moved in those areas after car is parked and shut off)

Tesla:

Undocumented – TeslaTap

teslatap.com

 

[diplomat33]

Mobileye has to be more open about their products because they want to sell to OEMs for integration into their vehicles. OEMs are not going to commit to an outside vendor's solution without some assurance that the vendor has a viable product. Tesla integrates into their own vehicles, so they have no need to publicly disclose more than they feel is necessary to sell FSD to consumers and to attract engineering talent.

So, Mobileye needs to market their product differently from Tesla. The fact that Tesla does not share a great deal of information outside the company does not mean that they are just winging it.

What you say is correct but it is not a good excuse. Tesla should still try to convince the public and regulators that they have a viable product. They have not done that. Tesla sold FSD in 2016 with a rigged video, made claims that the cars have all the necessary hardware for FSD and then changed the hardware, put out a half-baked AP software (2016) then did a few rewrites because it turned out the current software approach was not working, the CEO makes claims of a cross country demo that never happened, makes claims that FSD will be "solved" that don't happen year after year, AP had several crashes into guard rails and parked fire trucks and Tesla did nothing to solve. That does not inspire confidence. It does come across as making it up as they go. Compare that to Mobileye which lays ou their roadmap, lays out their approach, defines their ODD, defines precise safety goals, builds maps to support the system, builds safety driving policy, builds redundancy etc... that does inspire confidence that they know what they are doing.

 

[diplomat33]

It would be interesting to see a Zeekr with Mobileye SuperVision and a Tesla with HW3 FSD do the same routes in different cities.

https://cars.usnews.com/cars-trucks/features/zeekr-automotive-coming-to-america

Thanks for that link. So it looks like we might get SuperVision in the US later this year if Zeekr does bring the Zeekr001 to the US as planned.

 

[Supcom]

What you say is correct but it is not a good excuse. Tesla should still try to convince the public and regulators that they have a viable product. They have not done that. Tesla sold FSD in 2016 with a rigged video, made claims that the cars have all the necessary hardware for FSD and then changed the hardware, put out a half-baked AP software (2016) then did a few rewrites because it turned out the current software approach was not working, the CEO makes claims of a cross country demo that never happened, makes claims that FSD will be "solved" that don't happen year after year, AP had several crashes into guard rails and parked fire trucks and Tesla did nothing to solve. That does not inspire confidence. It does come across as making it up as they go. Compare that to Mobileye which lays ou their roadmap, lays out their approach, defines their ODD, defines precise safety goals, builds maps to support the system, builds safety driving policy, builds redundancy etc... that does inspire confidence that they know what they are doing.

I doubt there is anything Tesla could do to change your obvious bias against them.

 

[diplomat33]

I doubt there is anything Tesla could do to change your obvious bias against them.

Not at all. There is lots Tesla could do to change my view on this. And by the way, It is only FSD where I am critical of Tesla. I like Tesla's vehicles, the gigafactories, their lineup of models, the OTA updates, the service centers, the superchargers etc...

 

[emmz0r]

Google today is announcing a HD version of its vehicle mapping solution. Unlike Google Maps, Google’s HD map is not a consuming-facing application, but an additional layer of data that’s served to the vehicle’s L2+ or L3 assisted driving systems through Google Automotive Services.

Google launches HD maps for vehicles, Volvo and Polestar first to integrate | TechCrunch

Google launches HD maps for vehicles, Volvo and Polestar first to integrate.

techcrunch.com

I guess they get the speed limits right

 

[rbt123]

Google today is announcing a HD version of its vehicle mapping solution. Unlike Google Maps, Google’s HD map is not a consuming-facing application, but an additional layer of data that’s served to the vehicle’s L2+ or L3 assisted driving systems through Google Automotive Services.

Google launches HD maps for vehicles, Volvo and Polestar first to integrate | TechCrunch

Google launches HD maps for vehicles, Volvo and Polestar first to integrate.

techcrunch.com

Volvo and Polestar: Both are Geely brands. Interesting.

I wonder if this is a partial payment for their Zeekr vehicles? A data for vehicles exchange rather than paying strictly cash.

 

[diplomat33]

I wish I were at CES to check out the Mobileye tech up close.

https://twitter.com/x/status/1611410641914314774

 

[scottf200]

I doubt there is anything Tesla could do to change your obvious bias against them.

Not at all. There is lots Tesla could do to change my view on this. And by the way, It is only FSD where I am critical of Tesla. I like Tesla's vehicles, the gigafactories, their lineup of models, the OTA updates, the service centers, the superchargers etc...

Well, I for one, appreciate @diplomat33's reserved, well thought out, constructive, open-minded, and fact based responses.

 

[Dewg]

Well, I for one, appreciate @diplomat33's reserved, well thought out, constructive, open-minded, and fact based responses.

Hear, hear.