Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Waymo

EVNow

Well-Known Member
Sep 5, 2009
10,749
30,953
Seattle, WA
About 3.5 years.

But in Phoenix, they had no experience with ride-hailing and just starting their 4th Gen FSD. And covid delayed things a bit too. I doubt it will take that long in SF. Waymo now has the benefit of all the experience in Phoenix and the 5th Gen FSD which is much better than what they have in Phoenix. In the Autonomous Progress thread, I predicted "6-12 months" before they open it up wide to everybody with no NDA. I could be too optimistic. But I still think it will take much less than 3.5 years.
Yes - even for Waymo anything more than 6 to 12 months would be dragging their foot for too long.

Infact when I just read the headlines about Waymo starting in SFO, I thought it was already driverless and paying customers.

What we really need to know is - how long does it take Waymo to expand to n+1 city. That will determine whether they can ever be successful.
 
  • Like
Reactions: rxlawdude

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,407
11,746
Terre Haute, IN USA
A data point not showing that Waymo can cut it in difficult driving situations. Witness the gerrymandered geofence in SF for evidence to that effect.

I am talking about a data point that shows how quickly Waymo launched a driverless service in SF after starting limited testing. The data point will help see if Waymo is able to accelerate their expansion or not.

And I think deploying robotaxis, even in geofenced areas, is pretty cool. But you see everything about Waymo as a glass half empty.
 

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,407
11,746
Terre Haute, IN USA
Waymo isn't selling their Lidars anymore.


Probably just another bullish sign, just like losing their top execs. ;)

Waymo wants to focus on deploying their robotaxis rather than just selling the sensor:

Waymo confirmed the decision to Reuters, adding that it’s now focusing on deploying its Waymo Driver tech across its Waymo One ride-hailing and Waymo Via trucking divisions.

Solving autonomous driving and deploying robotaxis is a much better focus than selling hardware IMO. I am happy that Waymo will focus more on deploying robotaxis.
 

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,407
11,746
Terre Haute, IN USA
Yeah. Moved to Waymo thread.

Thanks, and a couple of follow-up questions. Perhaps this should move to the Waymo thread, but oh well:
  1. Is there any indication, from conferences or interview answers etc., that Waymo has rethought any major part of the AV technical strategy or architecture over the last several years? In the way that e.g. Tesla has evolved their neural network architecture, examined the fusion problems of disparate sensors and the phased conversion to temporal persistence, and also the intent to move more of the stack to NN over time? Or like MobilEye with their evolved dual-perception / late fusion architecture? The thrust being that, over years of development far past the original expectation, one would expect that some pretty major rethinking would happen along the way.

Short answer is definitely, I think Waymo has evolved a lot in terms of their sensors, hardware and software stack over time. They've undergone some big changes.

In terms of hardware, they went from a crude lidar on the roof of the first Google car to a much more sophisticated array of long range and short range cameras, lidar and radar on the 5th Gen I-Pace, including perimeter sensors for seeing around corners. In fact, just from the 4thGen in Chandler to the 5th Gen in SF, we see some important changes. For example, the 5th Gen has cameras, lidar and radar in the front bumpers to see around corners where the 4th Gen only has the lidar and radar. I suspect Waymo encountered cases on the narrow streets in SF like what we see in FSD beta videos where the car has to creep forward and realized they needed more/better perimeter sensors than what they were using on the wide streets in Chandler. The sensors have also improved in quality a lot.

In terms of the software stack, they started with mostly lidar machine learning back when computer vision was still rudimentary. They've added a lot more advanced computer vision now. They've also added more ML for prediction and planning. They also developed much more advanced sensor fusion as they went from basically 1 lidar on the first Google car to almost 30 sensors now on the I-Pace. Waymo has also evolved on how to represent their perception and how to interface between perception and prediction. They started with just an image representation but then switched to VectorNet which is a NN that converts map and perception into vectors and polylines. Anguelov says that VectorNet significantly improved their prediction and planning. He also says that Waymo has not completely solved yet the question of how to best represent perception in a way that helps prediction and planning. He also describes how they are working on joint prediction/planning, meaning a single NN that handles both prediction and planning. So we could see more changes there in their stack.

I could give a more detailed answer but yeah, I think Waymo has evolved a lot in their FSD since they started.

  1. Your explanation implies that that the last CEO, Krafcik, was too much of a marketer and not enough bound to the engineering reality (probably similar to your and many others' take on Elon). However, at the time of this change a few months back, I seem to recall some people saying that Waymo actually wasn't moving fast enough toward real scaled-up deployment, and was stuck in a money-draining science-fair condition with no end in sight. These two explanations seem to be rather contradictory. I'm not asking you to reverse your position, just trying to understand if it's your own take or the generally accepted story. Perhaps I'm relying too much on one or two things I heard that weren't representative of mainstream Waymo-watcher consensus.

Well, I just gave you my personal take.

I think the general consensus seemed to be "Waymo has these robotaxis in Chandler. Their FSD looks great. So why haven't they expanded yet?" I don't think Krafcik had a good answer to that. He seemed to kind of say "we have plans and when the FSD is safe enough, we will expand". Behind the scenes, we know Krafcik wanted to expand but presumably couldn't because they were encountering edge cases and did not feel confident enough in the FSD.
 
  • Like
Reactions: Microterf

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,407
11,746
Terre Haute, IN USA
Again, going to take this to the Waymo thread so as not to detract from the thread on Elon's tweets.

Tesla has an FSD problem, Waymo has a scaling problem. Nobody has any idea which one will solve their problem first, but I do like Karpathy's presentations showing what they're doing to solve the FSD problem. I have yet to see anything from Waymo showing how they're going to solve the scaling problem.

Actually, I think Waymo's scaling problem is a FSD problem. Waymo does not want to use customers to supervise their FSD so they need safety drivers or remote assistance/roadside assistance. But you can't scale effectively with safety drivers and remote assistance/roadside assistance because it costs too much. So Waymo can only scale effectively when their FSD is good enough to not need safety drivers and not need remote assistance or roadside assistance. So basically, it is a FSD problem. If Waymo achieves FSD that does not need safety drivers or remote assistance or roadside assistance, then scaling won't be a problem. They can just deploy robotaxis, turn them on, and let them go.

Tesla does not have the same scaling problem because they don't need to solve FSD. Tesla is ok with customers supervising the system, so they can deploy a driver assist without solving FSD.

Waymo has talked about how they plan to solve scaling.

A year ago, Anguelov presented on machine learning at scale where he addresses some of the machine learning problems that Waymo is tackling:


Waymo has also said that they believe realistic simulation, meaning simulation with both realistic sensor input and realistic agents, is the key to scaling because it will allow them to train and validate their FSD in full trips at large scale.

Starting at 3:27:00:


In the Scale AI talk, Anguelov says he hopes simulation will allow Waymo to scale to "dozens of cities". He talks about leveraging simulation as the key to scaling around 14:20 mn mark:

 

Doggydogworld

Active Member
Mar 4, 2019
1,746
6,794
Texas
Do they mean the city limits or the San Francisco metro area
Service area, compared to old Uber/Lyft ride frequency map:
1632163030472.png
 

diplomat33

Average guy who loves autonomous vehicles
Aug 3, 2017
8,407
11,746
Terre Haute, IN USA
I can assure you that even within that "green" area, there will be many many restrictions.


Don't trust Whole Mars for Waymo info. He is a Waymo FUDster.

Waymo does go everywhere in the geofence area. The blue areas are simply designated pick up spots because the car can't pick you up in the middle of a building or in the middle of a park or in the middle of the road. But of course Whole Mars ignores that to try to take a cheap shot at Waymo.

And frankly it is rude for him to call people morons if they don't agree with him that "waymo is a joke".
 
Last edited:

willow_hiller

Active Member
Apr 3, 2019
3,251
16,492
Maryland
Don't trust Whole Mars for Waymo info. He is a Waymo FUDster.

It's pretty obvious that the blue areas are where Waymo has decided that there exists a safe curb for the vehicle to pull over and pick up a passenger; so it's not indicative of the driving geofence.

But it's worth asking whether the Waymo driver can decide for itself where it's safe to pull over and pick up a passenger, or whether the pick up areas need to be pre-labeled by Waymo employees.
 
  • Like
Reactions: EVNow and Bladerskb

Cheburashka

Active Member
Jan 29, 2018
2,533
3,693
Los Gatos, CA
He's correct though. Waymo has limited pick up and drop off points.
Of course FSD will just stop illegally in the middle of the road which seems worse to me. 🤷‍♂️

I don't get it though, is he implying that Tesla has multiple pick up and drop off points?

I haven't seen any evidence that Tesla can pick you up or drop you off...
 

powertoold

Active Member
Oct 10, 2014
3,036
6,534
USA
But it's worth asking whether the Waymo driver can decide for itself where it's safe to pull over and pick up a passenger, or whether the pick up areas need to be pre-labeled by Waymo employees.

With HD maps, people underestimate what is being manually labeled and human-verified (by driving and/or double checking vs the images / videos).

I wouldn't be surprised if things like "max speed 3 mph down this driveway" are being manually labeled and/or verified. The car isn't smart enough to figure out practical speeds by itself, and even if it does, it still needs to be human-verified.

There are many many things that need to be manually labeled and/or verified, all traffic control lane semantics as another example.
 
  • Like
Reactions: Microterf

Daniel in SD

Well-Known Member
Jan 25, 2018
7,065
10,530
San Diego
With HD maps, people underestimate what is being manually labeled and human-verified (by driving and/or double checking vs the images / videos).

I wouldn't be surprised if things like "max speed 3 mph down this driveway" are being manually labeled and/or verified. The car isn't smart enough to figure out practical speeds by itself, and even if it does, it still needs to be human-verified.
Seems plausible. How much do you estimate that would cost per mile?
What's so dumb about these endless Waymo vs. Tesla arguments is you're comparing an actual (barely) working system to a system that does not exist yet. My take away is that the reason for Waymo's limitations is that it's an incredibly difficult engineering problem and removing limitations wouldn't make it any easier to solve.
 

powertoold

Active Member
Oct 10, 2014
3,036
6,534
USA
Seems plausible. How much do you estimate that would cost per mile?
What's so dumb about these endless Waymo vs. Tesla arguments is you're comparing an actual (barely) working system to a system that does not exist yet. My take away is that the reason for Waymo's limitations is that it's an incredibly difficult engineering problem and removing limitations wouldn't make it any easier to solve.

I don't know anything about the initial upfront cost, but it seems like the ongoing maintenance cost will be high.

The main issue with Waymo is they don't have Tesla's fleet size and diversity.

I don't understand anything Waymo is doing nowadays. Just look at what's going on in Chandler (nothing).

 

Products we're discussing on TMC...

About Us

Formed in 2006, Tesla Motors Club (TMC) was the first independent online Tesla community. Today it remains the largest and most dynamic community of Tesla enthusiasts. Learn more.

Do you value your experience at TMC? Consider becoming a Supporting Member of Tesla Motors Club. As a thank you for your contribution, you'll get nearly no ads in the Community and Groups sections. Additional perks are available depending on the level of contribution. Please visit the Account Upgrades page for more details.


SUPPORT TMC
Top