Waymo

[spacecoin]

Waymo’s Updated Visualization

Earlier this week, a Waymo ride still showed the familiar depiction of people as white dots, but today I saw a new visualization. People are now shown as people, as are cyclists. Cars themselves, w…

thelastdriverlicenseholder.com

 

[diplomat33]

Waymo’s Updated Visualization

Earlier this week, a Waymo ride still showed the familiar depiction of people as white dots, but today I saw a new visualization. People are now shown as people, as are cyclists. Cars themselves, w…

thelastdriverlicenseholder.com

I like the new visualizations.

 

[rentierparasit]

According to teslafsdtracker, the intervention rate is 146 miles per safety critical intervention. Tesla needs to get to at least 10k miles per safety intervention to even start a geofenced robotaxi. Put differently, 96% of drives have no safety critical intervention. Tesla would need to get that number to like 99.999% to do robotaxis. V12.3.3 is a very long way from eyes-off or driverless. The notion that vision-only is almost solved is silly.



IMO, there is no need to remove radar/lidar yet. Cost of radar and lidar are coming down and radar and lidar add huge safety benefits. Comparing stats, Waymo is at 99.999959% in terms of reliability while Tesla FSD V12 is at 96%. So using radar/lidar, Waymo's autonomous driving is way more reliable than Tesla's vision-only. Why would Waymo remove something that is getting cheaper and makes their autonomous driving orders of magnitude more reliable?

Hi, Diplomat --

> Tesla needs to get to at least 10k miles per safety intervention to even start a geofenced robotaxi.

I've been using 100k for robotaxi/10k to start getting serious about seeking regulatory approval. If accidents/critical disengagement are 1/10, 100k gets 1MM miles between "This thing was clearly at fault" accidents. How attached are you to that 10K number? I do note the "at least" qualifier!

Yours,
RP

 

[diplomat33]

Hi, Diplomat --

> Tesla needs to get to at least 10k miles per safety intervention to even start a geofenced robotaxi.

I've been using 100k for robotaxi/10k to start getting serious about seeking regulatory approval. If accidents/critical disengagement are 1/10, 100k gets 1MM miles between "This thing was clearly at fault" accidents. How attached are you to that 10K number? I do note the "at least" qualifier!

Yours,
RP

I am not super attached to the "10k miles per critical intervention". That is why I said "at least". The number is an approximation. It comes from Waymo and Cruise reporting disengagements to the CA DMV in that range when they started driverless testing. So I feel like ~10k miles per critical intervention is a minimum before starting driverless testing.

Of course. the key is what type of intervention. Are we measuring all interventions, only interventions that correlate to any collision, only interventions that correlate to at-fault collisions, only interventions that correlate to a collision with serious injury? I think any company needs to look at all of these metrics compared to humans in same ODD and make a determination when the system is "safe enough" for driverless.

 

[rentierparasit]

I am not super attached to the "10k miles per critical intervention". That is why I said "at least". The number is an approximation. It comes from Waymo and Cruise reporting disengagements to the CA DMV in that range when they started driverless testing. So I feel like ~10k miles per critical intervention is a minimum before starting driverless testing.

Of course. the key is what type of intervention. Are we measuring all interventions, only interventions that correlate to any collision, only interventions that correlate to at-fault collisions, only interventions that correlate to a collision with serious injury? I think any company needs to look at all of these metrics compared to humans in same ODD and make a determination when the system is "safe enough" for driverless.

Hi, Diplomat --

> ~10k miles per critical intervention is a minimum before starting driverless testing.

We're in agreement, then.

> Are we measuring all interventions

I'm going off the tracker, because that's all we've got, and looking at their (somewhat subjective) City Miles To Critical Disengagement.

Yours,
RP

 

[diplomat33]

Winter in Buffalo , Spring in DC , and summer in Atlanta ! We’re taking the #WaymoDriver to the ATL as we continue evaluating how our system generalizes to different environments and new experiences. Read about our cross-country training program: https://waymo.com/blog/2023/11/road-trip-how-our-cross-country-testing-helps-advance-the-waymo-driver/

https://twitter.com/x/status/1780249405821689938

 

[JulienW]

https://twitter.com/x/status/1780249405821689938

What a coincidence and lookie what I just saw in ATL on 17th ST. I'm so pissed. I just got 12.3.4 and installed while at the gym. Got back in my car and put in an across the city Nav to test. Pull down and get Cruse Control unavailable. Pulled onto a side street and checked Autopilot Settings and did a MCU reboot and nada. Decided to cut short for home and when I turned on 17th St THE car was coming and I just decided to wait there. As it got closer I was like WHAT THE HELL is THAT.

SUCKS that FSD wasn't working so we could have had a face off. Still EXCITED to see ATL getting some Waymo .

EDIT: Just reread Tweet and Summer in ATL. Looks like it came early. Too bad DC.

 

[ThomasD]

I wonder if Waymo is also testing on I 75 through Atlanta.

 

[JulienW]

I wonder if Waymo is also testing on I 75 through Atlanta.

You mean The Connector (what we call it) and is actually I75&I85 combined. Doubt it since they don't really do Interstates right now. Although right this minute (5:45PM pic from my condo) it could be going in or out of ATL and no big deal since it is as usual not moving much. This is actually only I85 just north of the I75 intersect (to right of pic). I75 goes below but you can't see it. It only gets worse to the south (right in pic) where they combine and go through town.

 

[ThomasD]

I pass through there when I go to visit relatives in Florida. Also Summit Racing is not far from there.

 

[JulienW]

I kinda thought Waymo used Google Maps to make HD Maps but does Waymo use its cars to build its HD Maps? In the Atlanta Journal Constitution (news paper) it has an article about Waymo in ATL and a Waymo spokesperson says:

"Waymo said it will have a handful of cars in Atlanta for testing over the next few months, driving manually at first and then moving to autonomous testing with a human driver behind the wheel...

The vehicles will initially be driving at all hours of the day, in various neighborhoods around Atlanta, including, Downtown, Georgia Tech, the Westside, Midtown, Piedmont Heights, and Virginia Highland...."

So do they drive manually to map the streets first?

Waymo arrives in Atlanta. Robotaxi company starts testing self-driving vehicles

Waymo is testing its autonomous driving system in Atlanta.

www.ajc.com

 

[diplomat33]

So do they drive manually to map the streets first?

Yes, Waymo drives manually first to map the streets. Waymo makes their own HD maps. And Waymo uses the same I-Paces for mapping and for rides. Driving manually also allows them collect data to train the planner.

 

[JulienW]

Yes, Waymo drives manually first to map the streets. Waymo makes their own HD maps. And Waymo uses the same I-Paces for mapping and for rides. Driving manually also allows them collect data to train the planner.

Then I guess my FSD vs Waymo fantasy face off wouldn't have worked anyway since Waymo, like me was also having Full Self Driving "issues".

 

[JulienW]

Yes, Waymo drives manually first to map the streets. Waymo makes their own HD maps. And Waymo uses the same I-Paces for mapping and for rides. Driving manually also allows them collect data to train the planner.

Just another: I guess they are also always dynamically updating their HD maps with each customer drive to keep them as current as possible?

 

[diplomat33]

Just another: I guess they are also always dynamically updating their HD maps with each customer drive to keep them as current as possible?

Correct.

 

[Doggydogworld]

Just another: I guess they are also always dynamically updating their HD maps with each customer drive to keep them as current as possible?

I wouldn't say with each drive. More likely that cars flag potential map issues and alert HQ, which then has a process to decide if a map update makes sense.

If they dynamically updated maps then six Waymos wouldn't have followed the first one into trouble when construction cones forced all traffic away from Bayshore Blvd onto 101 the other night. And this is an obvious major case - the first car recognized the situation and got out of the way as best it could. The vast majority of times cars flag a map issue will be very minor. Things like scaffolding on a building, a big construction dumpster in a parking spot, graffiti on the street or a sign, a huge pile of sand or dirt off to the side for some project, roadside tree or bush trimming, a mattress or armchair that fell off a truck, a restaurant putting tables or a bar on the sidewalk or street plus thousands of other things that happen every day.

It's nice to imagine AGI-like inference engines in the cars figuring all this out on the fly, like humans do, and dynamically changing the map so other cars know what's coming. But even Waymo is a long way from that. (Tesla will have it in 6 weeks, though

 

[diplomat33]

Just another: I guess they are also always dynamically updating their HD maps with each customer drive to keep them as current as possible?

Here is what Waymo says in their blog about how they update HD maps.

Our system can detect when a road has changed by cross-referencing the real-time sensor data with its on-board map. If a change in the roadway is detected, our vehicle can identify it, reroute itself, and automatically share this information with our operations center and the rest of the fleet in real time.

We can also identify more permanent changes to the driving environment, such as a new crosswalk, an extra vehicle lane squeezed into a wide road, or a new travel restriction, and quickly and efficiently update our maps so that our fleet has the most accurate information about the world around it at all times.

We’ve automated most of that process to ensure it’s efficient and scalable. Every time our cars detect changes on the road, they automatically upload the data, which gets shared with the rest of the fleet after, in some cases, being additionally checked by our mapping team.

Source: The Waymo Driver Handbook: How our highly-detailed maps help unlock new locations for autonomous driving

 

[JulienW]

CRAZY again. There MUST be a fleet of Waymos in ATL. If only 1 then just seeing it yesterday would be a HUGE coincidence. BUT I went for a run this morning and on Spring St heading south (behind me in pic) a Waymo passed me and turned left/east on Linden St. I hazardly crossed Linden St while trying to get my iPhone out of my running pack. Just barely caught a last glimpse of the Waymo as it turned left/north on West Peachtree to take a pic as proof.

If there is only 1 Waymo in ATL that just got here 2 days ago and I have seen it both days.........then I need to buy a lotto ticket.

 

[JulienW]

Almost totally unbelievable but I am now 3 Waymo sightings this week. How is this possible? One of the following must be the answer so which is it?

1. I’m in San Fransisco and just don’t realize it
2. I actually live in the tiny town of Atlanta, Texas so seeing the same car every day is not so unusual
3. Waymo must have brought a fleet a Waymo’s to ATL
4. I'm on a defy the odds roll and should buy a lotto ticket today

 

[diplomat33]

Almost totally unbelievable but I am now 3 Waymo sightings this week. How is this possible? One of the following must be the answer so which is it?

1. I’m in San Fransisco and just don’t realize it
2. I actually live in the tiny town of Atlanta, Texas so seeing the same car every day is not so unusual
3. Waymo must have brought a fleet a Waymo’s to ATL
4. I'm on a defy the odds roll and should buy a lotto ticket today

View attachment 1039971

Likely, Waymo brough a bunch of Waymos to ATL and happen to be testing in you area.