Waymo

[MP3Mike]

We don't know. Waymo has not said what happened yet.

Apparently fog defeated the Waymos, but it didn't seem to be a problem for humans...

Waymo says dense S.F. fog brought 5 vehicles to a halt on Balboa Terrace street

Five self-driving vehicles blocked traffic early Tuesday morning in the middle of a...

www.sfchronicle.com

 

[JHCCAZ]

Apparently fog defeated the Waymos, but it didn't seem to be a problem for humans...

To be fair, we probably shouldn't use the fact that humans continue to drive in deteriorating-visibility conditions, as a benchmark for early-generation Robotaxis .

My suggestion to Waymo would be that they put a status message on the interior screen, similar to Tesla Dog Mode, so that people can peer in and see if the car knows why it's just sitting there. If the screen said
"Waymo paused for safety - Fog condition detected"
It might elicit a few snarky comments but overall people would appreciate it, compared to just a mystery stoppage.

This actually goes to my view of the often debated L4 ODD (latest installment found on the previous page of this thread). It's not just a map geofence. A practical L4 vehicle, whether Robotaxi or personally owned, will evaluate the requested route and will accept, refuse or something in between, based on a number of factors including weather, road classifications, reported traffic issues, construction, police information etc.

The "something in between" would usually be a notice to the requestor that there is an X percent likelihood of delay or even failure to complete the requested route. For a personal vehicle, the owner may have a higher risk tolerance for delays or return home scenarios; the Robotaxi operator may apply a more conservative standard.

This is fundamentally not really different from human-driven cars and taxis. But back to the SF fog example, humans will tend to continue operating the vehicle even when the risk is unusually high - because usually they get away with it, and if they don't it's not national news.

 

[cwerdna]

Apparently fog defeated the Waymos, but it didn't seem to be a problem for humans...

Waymo says dense S.F. fog brought 5 vehicles to a halt on Balboa Terrace street

Five self-driving vehicles blocked traffic early Tuesday morning in the middle of a...

www.sfchronicle.com

Five driverless Waymo cars block residential SF intersection is a copy if you get hit by a paywall.

 

[diplomat33]

Apparently fog defeated the Waymos, but it didn't seem to be a problem for humans...

Waymo says dense S.F. fog brought 5 vehicles to a halt on Balboa Terrace street

Five self-driving vehicles blocked traffic early Tuesday morning in the middle of a...

www.sfchronicle.com

The Waymos were able to resume driving in 6 minutes. I would hardly call that a defeat.

"He estimated the self-driving cars were halted for at least six minutes before he saw the first one glide away."

Waymo says that there was dense fog that limited visibility. Yes, human drivers might still try to drive in dense fog but humans often drive in poor visibility when they shouldn't. In fact, we often see crashes due to human drivers driving in fog or snow when visibility is too low. So that is hardly something AVs should imitate. AVs like Waymo will prioritize safety and achieve a minimum risk condition where visibility is too low. Heck, if the Waymo had tried to drive in the dense fog and hit a kid, the story would be a Waymo injuring a child. I think it is better to pause for 6 minutes than to potentially hit a pedestrian.

AVs like Waymo also have the advantage that with the sensors and compute, they can measure exactly how bad visibility is. They can quantify how degraded their perception is and they can be programmed to reach a minimum risk condition when perception is degraded by X%. So unlike human drivers which guesstimate and often try to drive when it is unsafe, AVs can be very reliable in knowing when weather conditions make it unsafe to drive.

Having said that, Waymos will continue to improve in adverse weather conditions. Waymo engineers are undoubtedly adding this fog data to their ML training to further improve their perception in dense fog. In the future, Waymos will handle bad weather better and won't have to stop as much in dense fog or other bad weather.

 

[Doggydogworld]

Apparently fog defeated the Waymos, but it didn't seem to be a problem for humans...

Yeah, humans are undefeated vs. fog.

 

[ThomasD]

Sometimes the Fog just rolls up on you. Go up a hill and around a corner and the fog is so dense you can hardly see the Deer on the edge of the road

 

[diplomat33]

A fully autonomous (no human driver) ride from the Haight District to Stonestown Mall. Taken Sunday April 16 in the afternoon.

0:00 Video start
0:40 Yield to peds
1:20 Unprotected left
1:50 Yield to ped
2:40 Left turn at T junction
4:49 Yield to ped in road
9:23 Yield to peds
10:46 Unprotected left onto 19th Ave
11:03 Driving on 19th Ave
17:20 Right lane change
18:48 Right turn
20:19 Entering parking lot
22:27 Dropoff

A fully autonomous (no human driver) ride from the Mission District to the Richmond District. Taken Sunday, April 16 in the morning — San Francisco is pretty sleepy then.

 

[diplomat33]

Good podcast with Waymo Head of Research. Lots of good technical stuff.

Drago Anguelov: Waymo and Self-Driving

On the development and challenges of self-driving, Waymo's innovations, and what's next for the field.

thegradientpub.substack.com

Outline:

• (00:00) Intro
• (02:04) Drago’s background in AI and self-driving, work with Daphne Koller + Sebastian Thrun, computer vision / pose estimation
• (14:20) One- and two-stage object detectors
• (15:15) Early experiences and thoughts on self-driving and its prospects
• (21:00) An introduction to the “self-driving stack”: mapping & localization, perception, behavior modeling & planning, simulation
• (29:25) From Stuart Russell’s comments on early Waymo’s “old-fashioned” approach
• (37:34) Scaling 3D Detection: challenges and architectural innovations
• (43:20) Behavior modeling: making decisions and modeling interactions in multi-agent environments
• (52:42) Distributional RL (+ imitation learning) in self-driving?
• (54:10) The Waymo Open Dataset
• (1:01:48) Looking forward in self-driving
• (1:04:36) Outro

 

[diplomat33]

https://twitter.com/x/status/1649790187856097281

 

[ThomasD]

Is 90 seconds to respond to a situation too long

Video shows passengers in driverless car reacting to police officer gesturing to pull over: 'We're sorry, this car won't let us move!'

The passenger told Insider it was his first time in a self-driving car and he was actually impressed by how the vehicle handled itself.

www.yahoo.com

 

[kabin]

A fully autonomous (no human driver) ride from the Haight District to Stonestown Mall. Taken Sunday April 16 in the afternoon.

0:00 Video start
0:40 Yield to peds
1:20 Unprotected left
1:50 Yield to ped
2:40 Left turn at T junction
4:49 Yield to ped in road
9:23 Yield to peds
10:46 Unprotected left onto 19th Ave
11:03 Driving on 19th Ave
17:20 Right lane change
18:48 Right turn
20:19 Entering parking lot
22:27 Dropoff

A fully autonomous (no human driver) ride from the Mission District to the Richmond District. Taken Sunday, April 16 in the morning — San Francisco is pretty sleepy then.

Steering seems more confident with much less high frequency noise input. And unlike 11.3.6 wanting to almost rub side mirrors with parked cars, Waymo gives a safer more realistic separation. You never know when parked car doors will swing open or a kid runs out from between parked cars. Of course it's the driver's fault if FSD plows them over.

 

[diplomat33]

A (driverless?) Waymo was spotted in Scottsdale, AZ.

Just wild to me…this WAYMO car does NOT have a human driver. In fact, this vehicle didn’t have ANY human inside. 😮 😮 😮 | By Curtis HoekstraFacebook

Just wild to me…this WAYMO car does NOT have a human driver. In fact, this vehicle didn’t have ANY human inside. 😮 😮 😮

www.facebook.com

Could be sign that Waymo is planning to expand to Scottsdale and connect the downtown Phoenix area to the Chandler area.

 

[Iron]

When I look at a Waymo robo taxi bristling with sensors all over it and then Teslas reduced to only Vision/cameras. I can't help but wonder how far away Tesla is from achieving it's goal of FSD.

 

[Doggydogworld]

A (driverless?) Waymo was spotted in Scottsdale, AZ.

Just wild to me…this WAYMO car does NOT have a human driver. In fact, this vehicle didn’t have ANY human inside. 😮 😮 😮 | By Curtis HoekstraFacebook

Just wild to me…this WAYMO car does NOT have a human driver. In fact, this vehicle didn’t have ANY human inside. 😮 😮 😮

www.facebook.com

Could be sign that Waymo is planning to expand to Scottsdale and connect the downtown Phoenix area to the Chandler area.

Car was going east on E. Indian School Rd and turned south on N. Hayden Rd. This intersection is 6.5 miles due north of their Chandler service area (which has always included bits of Tempe and Mesa) and 4.5 miles east of their downtown service area.

As a practical matter they can't really connect these areas without highways. Google estimates no-traffic Sunday morning travel time from southern Chandler to the NW part of their downtown area at 35-40 minutes using highways vs 1 hr 15 minutes on surface streets.

 

[diplomat33]

Car was going east on E. Indian School Rd and turned south on N. Hayden Rd. This intersection is 6.5 miles due north of their Chandler service area (which has always included bits of Tempe and Mesa) and 4.5 miles east of their downtown service area.

Thanks, I know. I looked up the intersection before posting. I am correct tat it does represent a possible expansion if the car was 6.5 mi outside the current Chandler area and 4.5 miles outside the downtown area.

As a practical matter they can't really connect these areas without highways. Google estimates no-traffic Sunday morning travel time from southern Chandler to the NW part of their downtown area at 35-40 minutes using highways vs 1 hr 15 minutes on surface streets.

I agree. That is why I believe Waymo will add highways "soon".

 

[rbt123]

Guesses about Waymo expansion to Scottsdale were right.

​

"Driving the news: Waymo's Phoenix expansion will connect its existing downtown and East Valley territories, the company said Thursday.​

• The larger operating zone will include Scottsdale for the first time — and will cover nearly all of Tempe, with additional access to Chandler and Mesa.
• The company is also making it easier to take a robotaxi to or from the airport with a second Phoenix Sky Harbor International Airport pickup spot at the new 24th Street PHX Sky Train Station."

https://www.axios.com/2023/05/04/waymo-phoenix-expansion

 

[diplomat33]

Guesses about Waymo expansion to Scottsdale were right.

​

"Driving the news: Waymo's Phoenix expansion will connect its existing downtown and East Valley territories, the company said Thursday.​

• The larger operating zone will include Scottsdale for the first time — and will cover nearly all of Tempe, with additional access to Chandler and Mesa.
• The company is also making it easier to take a robotaxi to or from the airport with a second Phoenix Sky Harbor International Airport pickup spot at the new 24th Street PHX Sky Train Station."

https://www.axios.com/2023/05/04/waymo-phoenix-expansion

Yep.

Here is official tweet from Waymo.

https://twitter.com/x/status/1654108714377281536

Here is new service area. It covers 180 square miles!

New service area is HUGE. Very impressive!

 

[diplomat33]

Waymo is now doing 10,000 driverless public (non-employee) rides per week (SF and Phoenix).

We’re already serving 10,000 fully autonomous trips with public riders every week,” Panigrahi said. “I want to clarify that that’s external riders; that does not include the rides we offer to our own employees

Waymo barreling ahead with bigger robotaxi coverage in San Francisco and Phoenix

Waymo is out to prove that robotaxis are more than a fad.

www.theverge.com

 

[Doggydogworld]

Waymo is now doing 10,000 driverless public (non-employee) rides per week (SF and Phoenix).

And plans to do 100k/week one year from now.

 

[DanCar]

I'm excited about Waymo covering most of Phoenix metro. I wonder if Waymo still avoids freeways? That would make trips very long in Phoenix.