Waymo

[spacecoin]

If we don't get eyes-off on highways by 2027-28, it won't be because the cameras, radar or lidar are not good enough, it will be because the software is not good enough. I don't believe the problem with autonomous driving is that the sensors are not capturing enough data or the right data.

It might also be the case that OEMs can’t justify 6kUSD to their BOM until there is solid demand and legislation.

Most consumers won’t pay $8k extra for hw and then a hefty monthly subscription fee each month.

 

[diplomat33]

It might also be the case that OEMs can’t justify 6kUSD to their BOM until there is solid demand and legislation.

Most consumers won’t pay $8k extra for hw and then a hefty monthly subscription fee each month.

Sure. There are lots of reason that eyes-off on highway may not happen by 2027-28. Cost, regulation, legislation, liability, supply issues, lack of demand etc... could all be factors that prevent OEMs from deploying eyes-off on highways to consumer cars. I am just saying the cameras, lidar or radar not being good enough, won't be one of the reasons.

 

[Supcom]

Yes, I believe that the $6000 is what Mobileye will charge to OEMs. And OEMs will likely charge more to the consumer in order to increase their profit margin. So consumers will likely pay a lot more than $6000. But I was talking about the cost of the hardware to the OEMs, not the cost to the consumer. The question was when will the hardware be affordable to the OEMs to put on the cars.

OK, but if the hardware costs the end user too much, then it costs too much for the OEM because it won't sell. And the markup to consumers is not just profits. Every item installed on the car requires installation provisions like sheet metal cutouts, brackets, clips, screws, etc. They all need wiring harnesses to be built and everything needs to be installed. Then, there's the cost of the warranty that has to be passed on to the consumer. And since the cost is going to be quite - even if it was 'only' $6000 - the system will almost cerrtainly only appear on the highest end cars, making it even less affordable and probably allowing for an even higher markup.

So, that's why a Mobileye 'target price' means nothing and is almost certainly not going to be affordable unless you are a high end car buyer. But, since it still doesn't exist, it's all a moot point.

 

[spacecoin]

So, that's why a Mobileye 'target price' means nothing and is almost certainly not going to be affordable unless you are a high end car buyer. But, since it still doesn't exist, it's all a moot point.

I agree. But at least ME is aiming for autonomy, and not just marketing its L2 as "autonomy capable"...
HW costs will go down by 10% per year at least and all new tech is starting out in the high end cars.

 

[spacecoin]

Sure. There are lots of reason that eyes-off on highway may not happen by 2027-28. Cost, regulation, legislation, liability, supply issues, lack of demand etc... could all be factors that prevent OEMs from deploying eyes-off on highways to consumer cars. I am just saying the cameras, lidar or radar not being good enough, won't be one of the reasons.

No one is arguing with you apart from the cost point, which is the main thing that drives OEM adoption rate and their lobbying efforts.

Since Waymo has been driverless for many years at lower speed and they have also proved that their hardware works on the highway now, we know that the needed hardware exists.

As of today only the flagship models from two OEM:s (MB and BMW) have barely shipped super limited L3. As @Supcom points out, it is yet to be seen if someone can deliver the hw and sw suite that's needed for highway speed L3 at a resonable price point by 2025. If they can, we'll probably see a few flagship models capable of useful autonomy in some form by 2027-2029, otherwise later.

 

[kabin]

Waymo to begin testing driverless on highways! This is big!

https://twitter.com/x/status/1744373965672321054

Wow! They continue to be the real deal.

It didn't take long for Waymo to absorb Phoenix area highways and city streets. And it's strange, I haven't heard Waymo complaining about needing to build new supercomputers and training chips before performance improves.

I hope they accelerate work on consumer grade driverless systems.

 

[diplomat33]

Wow! They continue to be the real deal. It didn't take long for Waymo to absorb Phoenix area highways and city streets.

Yeah. Waymo is taking a no-fuss, slow but steady approach and it seems to be really working for them. Their autonomous driving continues to get more capable each month and they continue to steadily expand the ODD.

And it's strange, I haven't heard Waymo complaining about needing to build new supercomputers and training chips before performance improves.

Well, Waymo has access to Google's supercomputers. So they did not need to build their own.

I hope they accelerate work on consumer grade driverless systems.

Me too. I am optimistic based on the progress I am seeing. With driverless testing starting on Phoenix highways, I am thinking we should see Waymo open driverless on highways to the public in a few months and likely expand to other highways by the middle of this year. The Geely vehicle is entering mass production this year so we should see some testing of the Geely vehicle later this year as well.

Waymo will likely not do consumer grade driverless until they are fully confident their driverless is safe in a very wide ODD (Waymo has said that their target ODDs are: urban, suburban, highway and weather) since they won't do any consumer grade system until the human can strictly be a passenger. But With Waymo able to do safe driverless on urban, suburb, rain and now highways, the ODD is starting to get pretty good. They are almost at all 4 ODDs they listed as a goal. So I am hoping we see Waymo do some type of consumer grade eyes-off or driverless system in 2-3 years.

 

[kabin]

Waymo will likely not do consumer grade driverless until they are fully confident their driverless is safe in a very wide ODD (Waymo has said that their target ODDs are: urban, suburban, highway and weather) since they won't do any consumer grade system until the human can strictly be a passenger. But With Waymo able to do safe driverless on urban, suburb, rain and now highways, the ODD is starting to get pretty good. They are almost at all 4 ODDs they listed as a goal. So I am hoping we see Waymo do some type of consumer grade eyes-off or driverless system in 2-3 years.

It would be interesting to see how much complexity and cost they can squeeze out without significant performance/safety degradation.

 

[diplomat33]

Here is area and ODD of Waymo's CA DMV driverless permit:

Area
· San Francisco County
· San Mateo County
· Santa Clara County
· Portions or all of the
following cities:
– Atherton
– Belmont
– Brisbane
– Burlingame
– Colma
– East Palo Alto
– Foster City
– Hillsborough
– Los Alto
– Los Alto Hills
– Menlo Park
– Millbrae
– Mountain View
– Palo Alto
– Portola Valley
– Redwood City
– San Bruno
– San Carlos
– San Mateo
– South San Francisco
– Sunnyvale
– Woodside
· Los Angeles County
· Portions or all of the
following cities:
– Bell
– Bell Gardens
– Beverly Hills
– Carson
– Commerce
– Compton
– Cudahy
– Culver City
– El Segundo
– Gardena
– Hawthorne
– Huntington Park
– Inglewood
– Lawndale
– Long Beach
– Los Angeles
– Lynwood
– Manhattan Beach
– Maywood
– Paramount
– Redondo Beach
– Santa Monica
– South Gate
– Torrance
– Vernon
– West Hollywood

ODD: All times of day and night
· Inclement weather, rain, and fog
· Speed 65 mph

California Licenses for Autonomous Vehicles in Detail

On January 11, 2024, the California Department of Motor Vehicles (DMV) made the details of the licenses issued for driverless autonomous driving more easily accessible and clearer. The DMV website …

thelastdriverlicenseholder.com

It seems to cover quite a lot.

 

[diplomat33]

0:00 Start
0:22 Pull out
0:42 Unprotected left turn with cross traffic
2:11 Indecisive route flicker
2:25 Stepping on yellow line to avoid parked cars
3:30 Unprotected left turn with oncoming vehicle
6:03 Slowing for midblock crosswalk
7:03 Double parked vehicle
10:22 Right on red
11:01 Unprotected left turn with uncomfortable trajectory
11:22 Nudging for double parked vehicle
12:31 Found a destination on Hwy 1 ramp
12:58 Yielding to pedestrian with intent to cross
14:29 Yielding to pedestrians in street
15:21 Two lane changes in a row
15:47 Beach / Highway 1 view
17:03 Yielding to occluded cyclist after fresh green
17:21 Yielding to oncoming right of way violator
17:41 Yielding to cut-in from active police vehicle
18:47 Yielding to pedestrian with intent to cross
19:24 Unprotected left turn stealing oncoming’s right of way
20:59 Tourists are always watching
22:31 Kids react to Waymo
26:28 Indecisive proceed on yellow light
27:43 Pull over

 

[diplomat33]

https://twitter.com/x/status/1748134215265456444

 

[spacecoin]

https://twitter.com/x/status/1748134215265456444

I'd give a few hundred bucks to understand if they identify that the peson if a figure of a authority and some some rando.

 

[diplomat33]

I'd give a few hundred bucks to understand if they identify that the peson if a figure of a authority and some some rando.

I feel like it probably has to identify the figure as a person of authority, otherwise the car could accidentally obey any person that happens to be making hand gestures. But likely, the Waymo Driver is looking at several clues, like the fact that is at an intersection, that the person is standing in the middle of the intersection, that other vehicles are obeying the hand gestures, in order to judge that the hand gestures are relevant. That is probably what Dolgov means by "scene understanding", that the Waymo Driver is looking at many clues from the scene, not just one.

 

[diplomat33]

https://twitter.com/x/status/1748139925449404927

 

[Bitdepth]

I'd give a few hundred bucks to understand if they identify that the peson if a figure of a authority and some some rando.

I feel like it probably has to identify the figure as a person of authority, otherwise the car could accidentally obey any person that happens to be making hand gestures. But likely, the Waymo Driver is looking at several clues, like the fact that is at an intersection, that the person is standing in the middle of the intersection, that other vehicles are obeying the hand gestures, in order to judge that the hand gestures are relevant. That is probably what Dolgov means by "scene understanding", that the Waymo Driver is looking at many clues from the scene, not just one.

They indeed do understand if the person is relevant. I'll use Zoox for example where they discuss pose estimation, classification and semantic understanding.

 

[diplomat33]

Waymo navigates two busy parking lots in LA's Venice Beach. Filmed November 6, 2023.

0:00 What’s in Waymo’s glovebox?
1:05 Discussion about mapping trick for debris
2:22 Hairpin turn with pedestrian in street
4:14 Unprotected left turn
6:47 Unprotected left turn
8:31 Slowing for possible cut-in
11:02 Proceed on yellow
11:52 Yielding to cut-in
13:47 Following a cyclist
14:17 Entering parking lot 1
15:20 Misclassified dog
15:47 Camera crew films Waymo
16:06 Slowly driving through traffic control spikes, foot traffic
16:41 Tight squeeze with oncoming truck
17:20 Right on red
21:10 Entering parking lot 2
22:30 Venice Beach Skate Park

 

[diplomat33]

Waymo has applied for their CPUC permit to extend driverless service to LA:

https://twitter.com/x/status/1748550800958235004

 

[diplomat33]

From the CPUC application, here are the maps of the proposed LA and SF service areas.

CPUC Letter: https://www.cpuc.ca.gov/-/media/cpu...ssenger-safety-plan-update-january-192024.pdf

Both areas are big. The LA area includes a lot of dense areas.

I am surprised by how far down the SF area stretches. It seems to follow the 101. So I am guessing this will include highway driving. It shows that enabling driverless on highways will really help Waymo expand their service areas.

 

[Bitdepth]

That’s a massive expansion.

 

[diplomat33]

@ThomasD You asked about this. Waymo returning to Buffalo, NY for winter testing.

The Buffalo forecast calls for winter weather and Waymo’s return. We’re excited to be back in the Queen City for the next few months with the #WaymoDriver for weather testing.

https://twitter.com/x/status/1749564095295074397