Waymo

[Supcom]

Bosch is withdrawing from the development of lidar sensors​

Handelsblatt

app-handelsblatt-com.translate.goog

That won't help the cost of Lidar.

Perhaps they watched the V12 demo.

 

[diplomat33]

Bosch is withdrawing from the development of lidar sensors​

Handelsblatt

app-handelsblatt-com.translate.goog

That won't help the cost of Lidar.

Waymo builds all their lidar in-house. So this should not affect Waymo.

 

[hooty]

Waymo builds all their lidar in-house. So this should not affect Waymo.

So after the FSD class action lawsuits, will it be Waymo or Luminar that Tesla goes cap in hand to?

 

[diplomat33]

So after the FSD class action lawsuits, will it be Waymo or Luminar that Tesla goes cap in hand to?

If Tesla ever does do lidar, my guess is that they would go to Luminar or Velodyne. AFAIK, Waymo does not sell their lidar to outside parties.

 

[hooty]

If Tesla ever does do lidar, my guess is that they would go to Luminar or Velodyne. AFAIK, Waymo does not sell their lidar to outside parties.

So Bosch is irrelevant. Just be careful to whisper it around the echo chamber. Wilful ignorance is bliss.

 

[diplomat33]

* Marina to Fisherman's Wharf*

00:00 - Marina pickup
00:59 - Twitter video in same vehicle: https://twitter.com/mayaw/status/1694...01:25 - Twitter poll
02:40 - Fisherman's Wharf
02:56 - Sanitation trucks
03:48 - Pedicab
05:10 - People people people!
06:37 - North Beach
08:08 - Emergency vehicle
08:36 - Recent emergency vehicle nearby: • North Beach SF: Waymo detects sirens ...

** Chinatown and beyond **

09:30 - Entering Chinatown at Stockton St
10:15 - View of Broadway Tunnel;

Previous: • Daytime Robert C Levy (Broadway )Tunn... , • Waymo in SF's Broadway Tunnel at night

11:52 - Turning onto Eddy and then Grant
12:30 - Pedestrians in the street, ignoring lights
13:00 - Tight squeeze by stopped Tesla #1
13:32 - Tight squeeze by stopped Tesla #2
14:55 - Navigating by Woman loading van
15:25 - Downtown, North Beach
17:00 - Shade
17:31 - Broadway Tunnel
18:51 - Passing Lafayette Park
19:16 - Dropoff

 

[diplomat33]

https://twitter.com/x/status/1699407807500468424

The findings indicate that in comparison to the Swiss Re human driver baseline, the Waymo Driver — Waymo’s fully autonomous driving technology — significantly reduced the frequency of property damage claims by 76% (a decrease from 3.26 to 0.78 claims per million miles) when compared to human drivers. Furthermore, it completely eliminated bodily injury claims, a drastic contrast to the Swiss Re human driver baseline of 1.11 claims per million miles.

 

[flutas]

https://twitter.com/x/status/1699407807500468424

Furthermore, it completely eliminated bodily injury claims, a drastic contrast to the Swiss Re human driver baseline of 1.11 claims per million miles.

How can they claim they "completely eliminated bodily injury claims" when reports (from Waymo themselves) have explicitly stated injuries?

https://www.dmv.ca.gov/portal/file/waymo_08022023-pdf/

A passenger in the Waymo AV reported injuries.

It may be the other drivers fault, but it's still an associated injury claim with the Waymo driver.

 

[diplomat33]

How can they claim they "completely eliminated bodily injury claims" when reports (from Waymo themselves) have explicitly stated injuries?

Because the study uses data up to Aug 1, 2023 but the injury happened on Aug 2, 2023. So the injury happened after the data cut off date.

 

[flutas]

Because the study uses data up to Aug 1, 2023 but the injury happened on Aug 2, 2023. So the injury happened after the data cut off date.

So even the Study agrees, Waymo has not "completely eliminated bodily injury claims", unless you remove any and all cases where there was a test operator in the car, and only include data up until the day before a collision with injury, and even then their error bar is almost to the "baseline human driver". Yup. Sounds about PR.

And before someone comes in here claiming I must be a boomer who just hates AVs and stupid and blah blah blah. I like AVs, I don't like PR spin.

 

[diplomat33]

View attachment 971480
View attachment 971481

So even the Study agrees, Waymo has not "completely eliminated bodily injury claims", unless you remove any and all cases where there was a test operator in the car, and only include data up until the day before a collision with injury, and even then their error bar is almost to the "baseline human driver". Yup. Sounds about PR.

And before someone comes in here claiming I must be a boomer who just hates AVs and stupid and blah blah blah. I like AVs, I don't like PR spin.

If you look at the graphs with the errors bars, you can see the entire bar is below the human data, with no overlap of the error bars. So the Waymo safety was better than humans. The "completely eliminated bodily injury claims'" might be spin but the overall conclusion is correct. And the top tweet says " the Waymo Driver significantly reduced bodily injury ".

 

[flutas]

If you look at the graphs with the errors bars, you can see the entire bar is below the human data, with no overlap of the error bars. So the Waymo safety was better than humans. The "completely eliminated bodily injury claims'" might be spin but the overall conclusion is correct. And the top tweet says " the Waymo Driver significantly reduced bodily injury ".

From the paper:

Valid “apples-to-apples” comparisons must overcome differences in collision reporting standards between autonomous and human driven vehicles, correct the underreporting in police-report data, use operational-design-domain-specific human driver comparison data, apply a statistical method to measure uncertainty, and should account for crash causation contribution. Improperly controlled variations across collision datasets can lead to inflated or deflated collision statistics, and thus incorrect interpretations regarding safety.

Note their "baseline" isn't ODD specific like they claim it needs to be, and they even admit this. Because their method for getting a baseline is "who lives in Waymo's service area." Crash on the freeway across the country while on the freeway and on vacation? Well, guess you're a valid baseline for Waymo's crash rates.

In this study, the baseline for comparison was derived from a human population of insured drivers that reside in the same zip codes as the Waymo's service. Benefits of this population include its size and robustness, which lends itself to narrower confidence intervals. In addition, the selected baseline population is likely the population that may use Waymo services instead of driving themselves.

A limitation of the selected human baseline is that the location of crashes that generate claims is not known, which limits the ability to filter claims based on Waymo's Operational Design Domain (ODD). As a result, whereas the Waymo ODD largely does not include freeway driving, the human database includes miles driven and claims which occur on freeways. Due to variations in collision frequency per million miles between freeways and non-freeways, this may have led to a baseline which may be more conservative than a roadway-matched baseline. Due to the fact that in territorial ratemaking the frequency observed for residents in a specific area is considered to be the best proxy estimate of the claim frequency in that area, the impact of these differences is expected to be negligible.

They also convert years driven by the insured into "VMT" or Vehicle Miles Traveled, which is 100% an estimate.

Since Waymo's claims exposure is measured using mileage, in order to produce a valid human baseline for comparison, we convert the number of exposure years contained in the human driven dataset to the number of exposure miles. To do so, we estimate the annual vehicle miles traveled (VMT) for one vehicle8. Annual VMT per vehicle is estimated at a yearly and regional (state or city) granularity.

They state that this is estimated out with "regional (state and city) granularity" but then the footnote on this is literally:

A generally accepted estimate for annual VMT per vehicle in the US is approximately 12,000 miles. However, due to variations in driving patterns across different US cities and states, we separately estimate annual VMT per vehicle for each region (city or state) within the baseline.

So estimated numbers that aren't even listed in the paper is what we're comparing everything to...

This whole thing feels like a PR "let's pay for a study to show how safe it totally is" move.

 

[Doggydogworld]

How can they claim they "completely eliminated bodily injury claims" when reports (from Waymo themselves) have explicitly stated injuries?

They studied liability claims. They only count accidents when the insurance adjusters assign liability to Waymo vs. the other driver. It's a way to identify fault without limiting the study to the small percentage of wrecks reported to the police.

The initial goal is to reduce at-fault wrecks ~100%. You can't do much about the other wrecks, so it makes sense to exclude them.

 

[ThomasD]

Waymo says it is safer than Human drivers. But that is only in select locations that have been mapped and driven over hundreds of thousands of times. A Human can drive on any road such as dirt, paved or otherwise. People have stated that these vehicles don't need mapping to function. So why not do an Apples-to-Apples comparison. Put several Waymo vehicles on unmapped routes. A combination of City, Suburban and Rural areas. Then compare the safety of the cars with Human drivers.

 

[diplomat33]

Waymo says it is safer than Human drivers. But that is only in select locations that have been mapped and driven over hundreds of thousands of times. A Human can drive on any road such as dirt, paved or otherwise. People have stated that these vehicles don't need mapping to function. So why not do an Apples-to-Apples comparison. Put several Waymo vehicles on unmapped routes. A combination of City, Suburban and Rural areas. Then compare the safety of the cars with Human drivers.

Because they are not testing if waymo is safer than humans everywhere. it only matters if waymo is safer than humans in the service area where waymo is available to the public to ride. There is no point in testing if waymo is safer in areas where the public can't get a waymo.

 

[EvilNuff]

They studied liability claims. They only count accidents when the insurance adjusters assign liability to Waymo vs. the other driver. It's a way to identify fault without limiting the study to the small percentage of wrecks reported to the police.

The initial goal is to reduce at-fault wrecks ~100%. You can't do much about the other wrecks, so it makes sense to exclude them.

That's simply not true. If another car runs a red light and hits me they are at fault. I might have been able to stop and avoid the accident entirely. So what if (purely hypothetically) the not at-fault accidents revealed that waymo cars are terrible at avoiding accidents even when not at-fault?

 

[Doggydogworld]

That's simply not true. If another car runs a red light and hits me they are at fault. I might have been able to stop and avoid the accident entirely. So what if (purely hypothetically) the not at-fault accidents revealed that waymo cars are terrible at avoiding accidents even when not at-fault?

That's a second order effect, and much harder to measure. Waymo ran their simulator for a bunch of Phoenix human-human wrecks and found they would have avoided many of them even when put into the role of the not-at-fault car. But that's a biased approach with pretty useless conclusions.

If all drivers stop causing wrecks we won't have wrecks. So that's by far the top priority. The best way AVs can reduce human-caused wrecks is by deploying widely and cheaply. Then cities and states can get serious about forcing bad drivers off the road, instead of the current halfhearted approach. So widespread deployment is the second priority.

Trying to save bad drivers from themselves is a distant third priority.

 

[enemji]

That's simply not true. If another car runs a red light and hits me they are at fault. I might have been able to stop and avoid the accident entirely. So what if (purely hypothetically) the not at-fault accidents revealed that waymo cars are terrible at avoiding accidents even when not at-fault?

that is why I slow down when going through a green signal, especially when I am the first car going through. I don't care if it is my right and that I will be a no-fault. I will slow down.

 

[diplomat33]

https://twitter.com/x/status/1700242426089136232

 

[OxBrew]

How can they claim they "completely eliminated bodily injury claims" when reports (from Waymo themselves) have explicitly stated injuries?

https://www.dmv.ca.gov/portal/file/waymo_08022023-pdf/

It may be the other drivers fault, but it's still an associated injury claim with the Waymo driver.

Waymo said, in the tweet " the Waymo Driver significantly reduced bodily injury & property damage claims in comparison to the human baseline."

Not "entirely eliminated." Very different. Did you make that up, or was it somewhere else?

(I did not watch the video, so if you got that somewhere else, OK).