Waymo

[diplomat33]

Practically speaking, it would mean adding highway service. I don't see that for years.

First, you don't need highway driving to connect downtown Phoenix to Chandler.

Also, why would it take years to add highway driving? Waymo is already doing autonomous driving on highways in TX with their semi trucks now. And Waymo is doing autonomous driving on highways in CA in the I-Pace based on the CA DMV Report. So Waymo can clearly do highway driving. So, I don't see why it would take years to add highway driving. Waymo did remove highway driving from the Chandler ride-hailing because the 4thGen had some issues with taking on ramps. But I would imagine that issue is probably fixed by now with the 5th Gen, especially since we know the semi-trucks with 5th Gen are handling highway driving with no issues. If it is not fixed yet, it would certainly not take years to fix.

 

[ThomasD]

I wonder if Waymo has mapped any 2 lane highways for the Semi

 

[diplomat33]

I wonder if Waymo has mapped any 2 lane highways for the Semi

Yes, I think so since Waymo semis are driving autonomously 240 miles on I45 between Forth Worth and Houston:

Waymo Via is joining the freight-dense Interstate 45 route from Fort Worth, Texas, to Houston in a six-week test of its latest autonomous trucking software for UPS that comes as the package delivery company benefits from extra help for the holiday shipping season.

Seasonal assist: Waymo Via testing autonomous trucks with UPS in Texas

After moving packages for UPS in converted Chrysler Pacificas, Waymo Via is applying its latest autonomous system to Class 8 trucks in Texas.

www.freightwaves.com

 

[Daniel in SD]

Yes, I think so since Waymo semis are driving autonomously 240 miles on I45 between Forth Worth and Houston:

Seasonal assist: Waymo Via testing autonomous trucks with UPS in Texas

After moving packages for UPS in converted Chrysler Pacificas, Waymo Via is applying its latest autonomous system to Class 8 trucks in Texas.

www.freightwaves.com

Not a two lane highway. Two lanes = one lane in each direction (very annoying that the incorrect terminology is becoming more prevalent!). I doubt they'll operate on undivided highways for a very long time. Having the vehicle make decisions about evasive maneuvers when a car is driving on the wrong side of the highway just sounds too difficult at this point...
However, I bet there's enough trucking done on divided highways to make it a viable business.

 

[diplomat33]

Not a two lane highway. Two lanes = one lane in each direction (very annoying that the incorrect terminology is becoming more prevalent!). I doubt they'll operate on undivided highways for a very long time. Having the vehicle make decisions about evasive maneuvers when a car is driving on the wrong side of the highway just sounds too difficult at this point...
However, I bet there's enough trucking done on divided highways to make it a viable business.

I thought @ThomasD was referring to a divided highway with 2 lanes going one direction and 2 lanes going the other direction. That is how I understand a "2 lane highway". What you are talking about is not a highway in my book. One lane, undivided, in each direction is usually a state road, not a highway. Highways are usually multilane, divided interstate highways.

 

[ThomasD]

I was referring to U.S. highways not state roads. Like U S 60 or U S route 66 mostly 2 lane roads with the routes changing to multilane roads and also going on to Interstates. Parts of route 66 is also I 40. U S 60 is a major road Route 60 is over 2,000 miles long.

 

[Doggydogworld]

What you are talking about is not a highway in my book. One lane, undivided, in each direction is usually a state road, not a highway.

The world famous US Route 66 is not a highway in your book? Almost all US highways had two lanes outside of major cities until the 1950s when Dwight Eisenhower pushed for the Interstate system with minimum standards (4 divided lanes with 12' minimum width, 6% max grade, etc.).

There are still 150k+ miles of US highways vs. 50k+ of Interstates. I'd say the majority of US highways are still two lane outside of cities, but that's just a guess based on personal observation. Then you have a couple 100k miles of state highways with two lanes again being very common outside of cities.

Getting back to Chandler-to-downtown Phoenix, it's certainly possible to avoid highways but not, as I said, very practical. And their trucks do run on highways, but will have safety drivers for years. The highway thing is a liability/PR issue, not a technical limitation. Limited access highways are easy for the tech stack. Wrecks at those speeds can be horrific, though. It's a bad look have mangled cars on the evening news, especially if people are killed or maimed. Doesn't matter who's fault it is, "self-driving car in fatal accident" gets top billing.

 

[Daniel in SD]

Wrecks at those speeds can be horrific, though. It's a bad look have mangled cars on the evening news, especially if people are killed or maimed. Doesn't matter who's fault it is, "self-driving car in fatal accident" gets top billing.

The fatality rate on interstate highways is less than half the national average. Two lane highways are a different story, I can understand not wanting to operate there. Perhaps self-driving on interstates is not as easy as it might seem...

https://tripnet.org/wp-content/uploads/2021/06/TRIP_Interstate_Report_June_2021.pdf

 

[EVNow]

Having the vehicle make decisions about evasive maneuvers when a car is driving on the wrong side of the highway just sounds too difficult at this point...

They already do that in the city ...

 

[Daniel in SD]

They already do that in the city ...

Do they? I see this example in their safety report:

In simulation, the Waymo Driver detected the wrong way vehicle, initiated full braking, and was simulated to come to a complete stop in its lane prior to impact. The simulated collision assumes that the wrong way vehicle would have continued on the same path as observed in the actual event. The absence of simulated collision avoidance movement by the other vehicle reflects our assumption based on driving behavior and circumstances that the other driver was significantly impaired or fatigued. The resulting simulated collision shows the other vehicle traveling 29 mph when it strikes the stationary Waymo vehicle (S1 severity with expected airbag deployment).

They don't say how their safety driver was able to avoid the collision...

 

[EVNow]

Do they? I see this example in their safety report:

There is no big difference between 60 mph highway vs 50 mph city driving when it comes to head-on collisions. Either they can do evasive maneuver or just come to a stop in both cases.

 

[EVNow]

The world famous US Route 66 is not a highway in your book?

I had to chuckle at "world famous". Not that Route 66 is not famous - but I've heard people say so many things are world famous but barely anyone outside that city/state/country (depending on what that thing is) have heard of it.

 

[Daniel in SD]

There is no big difference between 60 mph highway vs 50 mph city driving when it comes to head-on collisions. Either they can do evasive maneuver or just come to a stop in both cases.

44% more kinetic energy. I would say that's a big difference.
But do they have the capability of doing evasive maneuvers? What level of safety can be achieved by simply slamming on the brakes and staying in the lane? I feel that to achieve greater than human safety on a two lane highway they would have to be able to do evasive maneuvers. And that seems like a very difficult problem because it requires prediction of the actions of the other driver and evaluation of normally non-drivable space.

 

[EVNow]

44% more kinetic energy. I would say that's a big difference.
But do they have the capability of doing evasive maneuvers? What level of safety can be achieved by simply slamming on the brakes and staying in the lane? I feel that to achieve greater than human safety on a two lane highway they would have to be able to do evasive maneuvers. And that seems like a very difficult problem because it requires prediction of the actions of the other driver and evaluation of normally non-drivable space.

Nobody said FSD was easy - though some people here claim Waymo has "solved" FSD.

All I'm saying is some city roads are almost as fast as highways. I've a local road here that is 55 mph. Cars normally go around 60/65 mph there. I use FSD regularly. If and when Waymo comes to my city, they will have to use it.

No different than a 60 mph US highway.

 

[Daniel in SD]

Nobody said FSD was easy - though some people here claim Waymo has "solved" FSD.

All I'm saying is some city roads are almost as fast as highways. I've a local road here that is 55 mph. Cars normally go around 60/65 mph there. I use FSD regularly. If and when Waymo comes to my city, they will have to use it.

No different than a 60 mph US highway.

All the roads like that around here have medians at least. Another thing is that this was in the context of semi trucks which operate much closer to their performance limits. I doubt Waymo will drive semi trucks in cities any time soon.
I challenge you to define what FSD is. haha.

 

[ThomasD]

Waymo may solve FSD in small geographical areas. I don't think they will ever solve it everywhere. Two way traffic on a one lane road with no paved shoulders. A one lane bridge on a down hill slope on a two lane road. Two vehicles get there at the same time. Who goes first? County and state roads that don't have any markings and look more like someone's long driveway than an actual road.

 

[Daniel in SD]

Waymo may solve FSD in small geographical areas. I don't think they will ever solve it everywhere. Two way traffic on a one lane road with no paved shoulders. A one lane bridge on a down hill slope on a two lane road. Two vehicles get there at the same time. Who goes first? County and state roads that don't have any markings and look more like someone's long driveway than an actual road.

Seems safe to bet against any company solving the most difficult engineering problem ever attempted. haha.
The only thing I can think of with so much money spent relative to results is nuclear fusion research.

 

[diplomat33]

Waymo may solve FSD in small geographical areas. I don't think they will ever solve it everywhere. Two way traffic on a one lane road with no paved shoulders. A one lane bridge on a down hill slope on a two lane road. Two vehicles get there at the same time. Who goes first? County and state roads that don't have any markings and look more like someone's long driveway than an actual road.

Why not? It's silly to think that Waymo cannot solve them. And it is silly to think that a problem cannot be solved just because you think it is a difficult problem. Plenty of problems seem very difficult until someone does solve them.

Waymo has camera vision and lidar to see the road and a HD map for added redundancy to tell the car where the road is, even without any markings. Waymo also has sensors to detect the other vehicles and a prediction stack to know how the other vehicles will drive. Waymo also has planning stack to tell the car how to drive. So Waymo has what it needs to handle those cases. They would not be difficult cases for Waymo.

 

[diplomat33]

https://twitter.com/x/status/1516134724598738944

The sessions are behind a paywall but I will share any interesting tidbits if/when they make the video public.

 

[EVNow]

All the roads like that around here have medians at least. Another thing is that this was in the context of semi trucks which operate much closer to their performance limits. I doubt Waymo will drive semi trucks in cities any time soon.

But the original context is why Waymo city service is not going to the city center from Chandler ... ?

I challenge you to define what FSD is. haha.

Stupid Level 5 ?