John Berry

Notice that DMV allows remote operators to perform DDT & MRC, which mean drive and stop the vehicle. CA DMV AV driverless permits, Test and Deployment, by default allow remote driving by "remote operators" registered by the permitted mfg w DMV.

FWIW, current California DMV AV regulations only define one telops role: "remote operator."

DMV's proposed revised AV regulations introduce the terms "remote driver" and "remote assistant."

DMV defines remote operator in their Test section 3.7, but also uses it in their Deployment section 3.8.

By stopping the vehicle even though in traffic, the robot achieved the ~minimum CA DMV required Minimal Risk Condition (MRC).

Waymo telop attempted to "move"/drive/"reposition" the vehicle bc the MRC was unsafe.

DMV does not require safe MRC or reporting of MRC events or their safety/consequences.

"bc you are kinda pulled over somewhat in the lane, I am going to attempt to move your vehicle to a safe location"
Waymo telop

The telop performed the very definition of the fallback DRIVER role, per SAE J3016, and indicative of Level 3 automation w a very long handoff time.

OP: tiktok.geicolizxrd

A Waymo robot illegally stopped in road in front of SFFD fire engine E21 displaying emergency lights:

300 block of Buena Vista Ave East, San Francisco

Looks similar to two recent videos of Waymo robots illegally stopping in traffic for school buses w flashing yellow lights.

OP: reddit.misterfuss

Its doppler radar almost certainly detected the motion before the lidar and camera confirmed it.

DMV gave Waymo permission to remove safety drivers in San Diego. If they haven't yet they prob will soon, though gradually.

Waymo requested CPUC's permission to expand ridehail without drivers to San Diego.

thread w Waymo's DMV approved & CPUC requested map:

bsky.app/profile/anic...

A Waymo robot drove away unaware this man was in the trunk/hatchback and delivered him to the next Waymo customer, who called LAPD.

Prob a reason Waymo is eager to get DoorDashers to close doors left open by their passengers. He closed a hatch door left open, from inside.

bsky.app/profile/anic...

It would drive away. They do all the time unaware of obvious things any human driver would notice. A Waymo robot drove away w this woman's dog:

bsky.app/profile/anic...

Perhaps if they were reliable & resilent, but they are not. And not by a factor of at least 100 compared to the human drivers they'd replace. Waymo shut down their entire Austin fleet for <inch of snow. They make bad/emergency situations worse, eg in San Francisco during a power outage.

Yet another (82+) Wrong Way Waymo:

W 35th St at Exposition Blvd, Austin, TX

Waymo robots have a bad/illegal tendency to cross double yellow lines to get around queues of cars. I've posted many similar videos, inc ~4 from one intersection in downtown San Francisco.

OP: tiktok.luxi.editing

Apparently, Waymo's robot never knew it was in trouble.

When Waymo & Cruise were beginning uncrewed/driverless operations, they said their robots recognized situations needing telops help so far in advance riders usually wouldn't notice.

How oblivious does a "driver" have to be not to notice this?

"let me escalate this to our special team who will maneuver the car," telop1.

telop1 prob only had customer service privileges, while telop2 had driving/"repositioning" privileges, and the robot didn't triage the situation or alert/inform telops.

He walked away from a hapless dumbot.

OP's text:

Yet another (81+) Wrong Way Waymo:

"turned into an active construction zone...then...into oncoming traffic...pressed rider support...When [non-US telops] couldn’t assist...transferred...to [US telops]... took more than five minutes to reach someone"
OP: tiktok.tylerlcrawford

He bailed. In Austin.

One of the best explanations of the legal conundrums of the AV driver.

State laws reference SAE J3016 for terminology, but it has problems, esp wrt human v robot roles. Kinda like they loosened just enough threads in a tangle for ppl to get a false sense of order/reason, but the tangle remains.

It also says remote assistants can generate new DDTs and "goals" for an ADS, which inverts their presumed roles IMO.

If an 'assistant' can generate (decide) for and impose on a robot a DDT, then which is the assistant in getting from A to B?

Potentially more of a horse & rider/driver relationship.

Really good. The practical legal issues are mind bending.

FWIW, SAE J3016 says remote assistants can do "DDT subtasks," but not a "complete DDT."

IMO, a distinction without much utility, as there isn't a standard set of DDTs & subtasks, nor would it be feasible to create them in the detail needed.

It's a general problem w 1st responders and emergency scenes in general, not specific to FD. LAPD & Phoenix PD have alotta stories too. My favorite of the robot stopped as it should then went as it shouldn't is this LAPD motorcycle cop:

bsky.app/profile/anic...

This is so common that San Francisco FD created an entirely new form just for robot incidents years ago. Robot mishaps had been making up most of their incident reports and need some extra reporting info/fields. They have alotta stories.

This is one w a creep.

bsky.app/profile/anic...

Yeah, Waymo robots have had this problem intermittently forever, AFAICT. Usually by just stopping in the middle of the road, sometimes add a creep & stop weirdness. One blocked a fire engine then backed up nearly into it.

They also blow right by them or across their path:

bsky.app/profile/anic...

And yet still not high enough.

They can eliminate all the LLM hallucinations with that one sure method: turn it off.

Waymo's next (6th) generation robot that is supposed to

"allow [Waymo] to safely broaden our footprint into more diverse environments, including those with extreme winter weather"

awkwardly stuck in a Philly snowbank being a road hazard.

There's always the next next generation, Alphabet willing.

Also another example of how, unlike human drivers, Waymo's robots do not learn or pickup clues from other drivers in realtime; making them a particular, peculiar, and persistent hazard in unusual conditions and emergencies where humans w less skill/knowledge succeed by following the lead of others.

Waymo has been specifically trying to fix their robot's behavior wrt stopped school buses for at least 4 months, and has made at least 3 distinct sw updates, one being a NHTSA recall. With a normal human driver, a brief conversation should suffice. Waymo has certainly spent millions not fixing this.

"need to fix that"
OP: tiktok.zonieliving, AZ

This is the fix in two ways.

Evidently, Waymo "fixed" their stopped school bus problem by literally fixing their robots in place when they detect a stopped bus w flashing lights. But flashing yellow lights mean warning, not stop.

Kludge upon kludge.

a man is talking on a phone and saying hello it have you tried turning it off and on again ALT: a man is talking on a phone and saying hello it have you tried turning it off and on again

That doesn't mean when they exercise control that it always works, but then who hasn't had a failing computer fail to respond to their control inputs?

Thus the famous appeal to the ultimate control which Waymo telops have:

FWIW, Waymo's remote ops very much do "control" their robots in

- the strict technical sense of a system controller w the ultimate on/off control; as well as stop/go, left/right, forward/back controls

- the common language sense of who is the first person responsible to control/fix a failed robot

Yeah, the crash reporting requirements do not cover all crashes, differ between DMV and NHTSA, have changed over time, and are a little complicated to explain. They also are not entirely reliable bc if a company doesn't know of a crash or relevant detail then it doesn't have to be included.