Waymo Says Its Self-Driving Tech Is More Advanced Than Tesla FSD, Others

[Daniel in SD]

I am just saying from the FSD video's I've seen so far, Tesla doesn't have those same restrictions

I'm sure that when Waymo has a safety driver in the car they take more challenging routes...

This is what I gathered from the video. Especially when used in the context of an entirely geofenced area with remote operators monitoring.

What about this video from 2010?

Unfortunately it's very easy to create the perception that FSD is right around the corner. The "capabilities" in a demo video from 2010 and 2020 look the same. The difference is the reliability.

 

[mark95476]

It's amazing how the narrative has changed just this year--

- Google/Waymo saying they're more advanced than Tesla's self-driving tech
- Toyota saying Tesla doesn't have a real product line
- VW saying they're bringing real competition to Tesla
....

 

[diplomat33]

30 mn video of Waymo's driverless rides:

 

[ZeApelido]

Can't wait until someone gets FSD beta in and just drives it around Chandler, Arizona repeatedly over the course of the next year.

 

[The Duke]

Once you buy FSD is there any location it gets disabled?
I have used mine in almost all the lower 48 and Canada, never a problem with FSD being available except for a few odd weather conditions.

 

[diplomat33]

Wayno just announced 2 new testing facilities:

 

[diplomat33]

Here is some more info on Waymo's new testing facilities:

Waymo plans to open a new testing site in Ohio for its driverless autonomous vehicles that will focus on dense, urban areas, the company said Tuesday.

The new site being built at the Transportation Research Center near Columbus will allow the company to work on motion control testing, heavy-duty truck testing and testing in varying weather conditions, the company said.

Waymo's main testing facility is near Merced, California. While it has conducted testing in several other states, this will be its first permanent location at a third-party test site.

It plans to open the new center in Ohio sometime in the middle of next year.

The company has been working on snow and wet roads in the Detroit area and heavy rain and fog in Florida and San Francisco, Patrick Cadariu, Waymo’s head of supply chain operations, said during the summer.

The testing site at the Transportation Research Center in Ohio will be designed with various types of terrain, including hills, along with dense, urban environments, the company said.

The research center in East Liberty, about 40 miles (64 kilometers) northwest of Columbus, opened a $45 million testing ground for self-driving vehicles last year.

It features roads and structures intended to replicate high-speed intersections, rural roads and urban areas normally encountered by drivers. Officials say the facility is among the most advanced in North America.

https://www.thenewstribune.com/news/business/article247525985.html

 

[mt09]

This is one of the most interesting Waymo rides I've seen with a busy parking lot and a pretty decent amount of traffic. Most of the action is at the beginning at the front of a busy parking lot followed by the exit with a lane merge. The rest of the drive is pretty typical and uneventful. Interesting interaction in a residential neighborhood with pedestrians crossing though.

1:10-3:32 Waiting for an opening start moving
3:32-5:14 Exiting parking lot
5:21 Exits and chooses right lane even though lane merges past traffic light
6:21 Attempts to merge left and stops in lane waiting for an opening
7:35 Interesting interaction with pedestrians crossing

I think the parking lot scenario shows how close Tesla FSD beta is in these uncontrolled situations. I feel current FSD beta doesn't handle merge conditions well though. It doesn't react to merge arrows on the road and just sees two lanes becoming one big lane. In this scenario I think FSD beta would continue and attempt to force it's way in since there isn't logic to negotiate a merge of two lanes into one.

 

[EVNow]

The research center in East Liberty, about 40 miles (64 kilometers) northwest of Columbus, opened a $45 million testing ground for self-driving vehicles last year.

It features roads and structures intended to replicate high-speed intersections, rural roads and urban areas normally encountered by drivers. Officials say the facility is among the most advanced in North America.

I thought Waymo was beyond testing in artificial environments at this point ... I guess they want to continue testing new versions on artificial testing grounds before using on public roads.

 

[diplomat33]

I thought Waymo was beyond testing in artificial environments at this point ... I guess they want to continue testing new versions on artificial testing grounds before using on public roads.

Yes, I think Waymo uses these testing grounds mostly to validate new hardware before deploying to public roads. I could see plenty of situations where this would be useful. For example, Waymo might want to test how a new lidar or new radar improves detection of cross traffic at an intersection. They can set up a fake intersection and test the new sensors without risking it on public roads.

 

[diplomat33]

I would add that the testing grounds are also useful for testing new software versions. Say Waymo wants to test new planning or new driving policy algorithms. It makes sense to test them in a controlled environment before deploying it to the fleet on public roads.

 

[ChrML]

Having the most advanced FSD system is actually a bad thing though. A simpler system is always better than more complex systems, providing it solves the same tasks.

A better phrased question is, which FSD solves most problems and is most reliable?

 

[ChrML]

I would add that the testing grounds are also useful for testing new software versions. Say Waymo wants to test new planning or new driving policy algorithms. It makes sense to test them in a controlled environment before deploying it to the fleet on public roads.

Given a good architecture, the driving policy rules should be defined at such a high level that it should be testable without real life testing. Manual testing is extremely slow and unrepeatable, and will inhibit development speed a lot. The automated testing kit should be fully decoupled from the self-driving implementation, and there must be an easy way to record real life scenarios into test cases.

Real life testing is more about uncovering missing test scenarios causing undefined behavior. There will quickly be so much driving between each "new" case that manual testing is impractical, and it becomes more of a way to observe first hand information. At that point deployment to a large fleet will be a major catalyst for reliability, providing problems are caught and reported automatically.

 

[Daniel in SD]

Having the most advanced FSD system is actually a bad thing though. A simpler system is always better than more complex systems, providing it solves the same tasks.

Advanced is not equivalent to most complicated. In this context I think it means closest to the goal of replacing all human drivers with machines. LIDAR is far simpler than vision if that's what you mean though. Humans understand exactly how LIDAR works. We don't understand how vision works.

A better phrased question is, which FSD solves most problems and is most reliable?

I'm betting Cruise but the proof is in the pudding and Waymo is actually driving around passengers without safety drivers.
Supposedly Cruise is going to start their San Francisco service soon.

 

[diplomat33]

Given a good architecture, the driving policy rules should be defined at such a high level that it should be testable without real life testing. Manual testing is extremely slow and unrepeatable, and will inhibit development speed a lot. The automated testing kit should be fully decoupled from the self-driving implementation, and there must be an easy way to record real life scenarios into test cases.

Real life testing is more about uncovering missing test scenarios causing undefined behavior. There will quickly be so much driving between each "new" case that manual testing is impractical. At that point deployment to a large fleet will be a major catalyst for reliability, providing problems are caught and reported automatically.

As I said earlier, I suspect the main purpose of the testing grounds and R&D facilities is developing new hardware since Waymo develops all their FSD hardware in-house. It would make sense to test an improved camera or improved lidar in a controlled environment. You won't develop say an improved radar and just put it on your fleet driving around on public roads with no prior testing.

A better phrased question is, which FSD solves most problems and is most reliable?

IMO, Waymo has solved the most problems and is the most reliable FSD right now. Waymo calls their Driver the most advanced FSD for a reason. For example, the Waymo Driver can read hand gestures of a cop directing traffic at an intersection with no traffic lights. That's not a problem that Tesla's FSD can do yet AFAIK. Waymo Driver is also more reliable since it has a disengagement rate around 1 per 13,000 miles last year and can do FSD with no safety driver. No other FSD company has shown that degree of reliability yet.

 

[diplomat33]

I'm betting Cruise but the proof is in the pudding and Waymo is actually driving around passengers without safety drivers. Supposedly Cruise is going to start their San Francisco service soon.

Cruise appears to be a close 2nd to Waymo. And yes, Cruise plans to deploy driverless robotaxis in SF very soon, now that they got approval from the CA DMV.

I know Waymo also plans to expand Waymo One in SF as well. When both Waymo and Cruise have driverless robotaxis operating in SF, it will be interesting to see how they compare. It will certainly be interesting to see robotaxis from 2 different companies operating in the same city. Competition is good!

 

[EVNow]

A better phrased question is, which FSD solves most problems and is most reliable?

Actually ...

... which FSD solves
- most problems (scenarios)
- most reliable
- for most people
- in most locations ?

We'll see that no one system is the best in all the four dimensions.

 

[diplomat33]

Actually ...

... which FSD solves
- most problems (scenarios)
- most reliable
- for most people
- in most locations ?

We'll see that no one system is the best in all the four dimensions.

The first "dimension" (most problems) is probably superfluous. If you have FSD that is reliable and works in a particular location, then by definition, it can handle all the scenarios in that location. Otherwise, it would not be FSD.

 

[EVNow]

The first "dimension" (most problems) is probably superfluous. If you have FSD that is reliable and works in a particular location, then by definition, it can handle all the scenarios in that location. Otherwise, it would not be FSD.

We actually have 3 dimensions
- Scenario (or feature)
- Location
- Individual (i.e. people)

Reliability is not a dimension - its a "measure". For a given combination of the above 3 dimensions we get a particular reliability (or error probability) number. Ofcourse the reliability can be calculated for a group of individuals/scenarios/locations or any combinations of them.

Waymo has high reliability for a lot of features - but only in small # of locations and only for their own cars.

ps : Elevators work in a lot of locations but in extremely limited number of scenarios

 

[diplomat33]

We actually have 3 dimensions
- Scenario (or feature)
- Location
- Individual (i.e. people)

Reliability is not a dimension - its a "measure". For a given combination of the above 3 dimensions we get a particular reliability (or error probability) number. Ofcourse the reliability can be calculated for a group of individuls/scenarios/locations or any combinations of them.

Waymo has high reliability for a lot of features - but only in small # of locations and only for their own cars.

I like those dimensions a lot better. They make more sense.