> "...the Fleet Response team provided incorrect guidance..."
It sounds like the car automatically pulled over when it detected the cones (as intended), but the remote support team erroneously told it to forge ahead anyway, which resulted in the escalating series of problems. Not sure what the best solution for that particular issue would be -- better training for the remote support teams, I guess?
You can watch the video. It pulls over incorrectly at first and is blocking traffic, then the support team made a mistake, and a hilarious escalation ensues with the vehicle actually attempting to run away from the support team multiple times!
They deliberately can't drive the car remotely. If Waymo wants a human to drive the car, they send the human to where the car is, the human gets inside and drives it like any other car.
One easy thing that can go wrong, just as with any other human activity, is the humans make a mistake and their "clarification" is actively unhelpful. If the Waymo Driver is confused about which lane is coned off, and you "help" by telling it the wrong lane, unsurprisingly things will not be improved.
The best solution is always that the Waymo Driver learns how to do better on its own, if you imagine it sees this situation and just confidently picks the correct lane and uses it, Fleet Response never interacts at all, there is no opportunity for them to "provide incorrect guidance". So a bunch of the invisible effort at Waymo is on making the Waymo Driver just get stuff right on its own.
Letting the passenger drive would unfortunately involve the risk of exiting the vehicle in traffic, which the lawyers probably vetoed, even though it might be safer in some situations.
At any rate, the remote team in this case IS able to control the car, hence it trying to drive into traffic cones and Google blaming the humans. It appears that their interface just can't be very good, especially as it requires people to physically turn up in a truck - you definitely won't have good situational awareness if you're trying to drive a car through a web app whilst looking out the window at the other side of the road.
The remote assistant never does anything to affect the car onboard system. The only instance is when the assist car turns up and the car is put into hold mode for the driver to enter.
Yes, VR is far from perfect, granted. But is it worse than what they appear to be using instead? Apparently they couldn't see the traffic cones properly from whatever system they were using and repeatedly gave it bad instructions, even though what to do was obvious to the person in the car. It seems like VR should be compared to the existing system rather than theoretically perfect vision.
Nope. This was already explained once in this thread, and you ignored it. Here's what Google's statement actually says
> ... the Waymo Driver detected an unusual situation and requested the attention of a remote Fleet Response specialist to provide additional information. During that interaction, the Fleet Response team provided incorrect guidance, which made it challenging for the Waymo Driver to resume its intended route, and required Waymo's Roadside Assistance team to complete the trip.
They can remotely tell a Waymo, stuff like "this lane is closed" or "No, that's not a school bus, it's just a truck" - and it will add that to the information it has, this might enable it to complete the journey. If you phone your dad, asking what a street sign you saw means, and he tells you - is he now driving your car? No, clearly you are still driving the car, he just answered a question. If he says the symbol means "Free on-street parking" but it actually meant "No entry" you're going to get some problems if you try to act on this bad information.
I mean ultimately, you can try to justify this as a feature not a bug, but I would now feel far less safe getting into a Waymo knowing that they have such a poor level of control over their cars. You shouldn't find yourself stalled in the middle of a multi-line highway like that, that's just dangerous, especially when continuing and pulling over is something any human driver could do in seconds. Being limited to ad-hoc map patching is just not sufficient - there's a whole universe of bugs they might not be able to resolve in that way.
Not sure if Waymo is reading Waze info, those kind of alerts could have been updated there (for roadworks, etc)