3 ms·

The car has 360 degree awareness of what's going on and cameras everywhere. VR headsets are real. I see no reason why remote drivers couldn't have even better s

The car has 360 degree awareness of what's going on and cameras everywhere. VR headsets are real. I see no reason why remote drivers couldn't have even better situational awareness than a normal driver given the sheer number of sensors in the car. 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.

angry_octet 5y ago
The bandwidth isn't available to transmit even the fused picture, VR has notorious misperception problems which are difficult to eliminate even in lab settings where you know the precise 3D measurements of everything, have carefully aligned the IPD (not even available on prosumer devices), and conduct careful per-session calibration. Is isn't ready for precise navigation tasks, and the delay and potential for jitter and dropout make it unsuited for safety critical tasks like driving. 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.
- thu2111 5y ago
  But Google are claiming that the erratic behaviour was caused by the remote assistance in this case. So they must be able to control the car. 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.
  - tialaramex 5y ago
    > But Google are claiming that the erratic behaviour was caused by the remote assistance in this case. So they must be able to control the car. 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.
    - thu2111 5y ago
      I understand that mechanism just fine, and that counts as a form of remote control. But it's clearly not good enough for this situation which was both totally unacceptable and very, very easily resolved by anyone who could control the steering wheel and accelerator. 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.

angry_octet 5y ago

The bandwidth isn't available to transmit even the fused picture, VR has notorious misperception problems which are difficult to eliminate even in lab settings where you know the precise 3D measurements of everything, have carefully aligned the IPD (not even available on prosumer devices), and conduct careful per-session calibration. Is isn't ready for precise navigation tasks, and the delay and potential for jitter and dropout make it unsuited for safety critical tasks like driving. 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.

thu2111 5y ago
But Google are claiming that the erratic behaviour was caused by the remote assistance in this case. So they must be able to control the car. 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.
- tialaramex 5y ago
  > But Google are claiming that the erratic behaviour was caused by the remote assistance in this case. So they must be able to control the car. 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.
  - thu2111 5y ago
    I understand that mechanism just fine, and that counts as a form of remote control. But it's clearly not good enough for this situation which was both totally unacceptable and very, very easily resolved by anyone who could control the steering wheel and accelerator. 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.