4 ms·

Until you get to the end of the field. If a malfunction means you don't turn or stop, the next thing in front of you might be a house, a school, or a highway. T

Until you get to the end of the field. If a malfunction means you don't turn or stop, the next thing in front of you might be a house, a school, or a highway. Today's farm tractors are powerful, heavy machines that can plow right through a lot of obstacles. I'd want to have a couple extra layers of fail-safes, and probably a human overseer with a kill switch. That could be someone watching a bunch of tractors remotely on monitors.

Pet_Ant 1y ago
I mean a GPS geo-fence is pretty simple and fail safe. I'm more worried about something like humans unexpectedly in the field. Imagine a migrant crossing a field and getting run over by a combine.
- aaronbaugher 1y ago
  Yeah, that's a concern too. I'm not opposed to the idea of autonomous (or better near-autonomous) tractors working in fields. They'd definitely be safer than autonomous vehicles on roads. I just think people, especially at a tech-loving forum like this, are a little too quick to assume safety concerns are fully covered. Perhaps they can be, but will they be when large corporate farms cut corners on things like maintenance? I started driving tractors when I was 10, and I've been in a couple situations where equipment failure required some quick action to get stopped before serious damage was done. Shit happens, and should be expected to happen and prepared for.
- ccozan 1y ago
  But I expect a autonomous tractor - like an auto - to have a radar and be able to stop if an obstacle is there - not only migrants, how about wild animals, etc.
pfdietz 1y ago
A GPS-driven cutoff seems like an obvious thing to have in this situation, no?
- nradov 1y ago
  GPS (or GNSS in general) is hardly sufficient for anything safety critical. The signals are very weak and can easily be spoofed or jammed. Even overhead obstructions like trees and buildings nearby can cause significant offsets due to multipath reflection issues.
  - zdragnar 1y ago
    RTK driven tractors are not using just standard GPS. They have positional accuracy down to one inch.
  - immibis 1y ago
    Simple solution for that. If you jam GPS and it makes an autonomous tractor run over a school, you're liable for the manslaughter of the school kids. Trains can be derailed, but they're still allowed to operate and we just put the derailer in prison. And knives can be used to stab people. And someone can even grab your arm and force you to stab someone with the knife, and the law understands this.
    - nradov 1y ago
      What a stupid "solution". It is almost never possible to catch GNSS jammer users after the fact. GPS alone can never be sufficient, regardless of the legal issues. There are automated technical solutions to this problem involving sensor fusion and/or ground beacons.
      - pfdietz 1y ago
        Simply jamming the GPS would cause the vehicle to stop. What would be needed would be spoofing the GPS to make the vehicle think it was still operating normally, in the field it was supposed to be in, without any detectable quirks, a much harder task.
        nradov 1y ago
        Jamming or not isn't a binary condition, nor is there a clear line between jamming versus spoofing. Certain types of signal interference will still allow for a position fix but with greatly degraded accuracy. There are algorithms to detect signal quality, but again that alone isn't considered adequate for anything safety critical. Commercial aircraft aren't legally allowed to fly with only GPS for navigation, and for good reason.
tonyhart7 1y ago
seems like someone didnt playing autonomous drone before, its a solved problem and Yes we do have OSS tool for that
- bryanlarsen 1y ago
  We even had solutions for that in the 70's when the law in question was written. But the solutions available at the time were far from fool-proof.
JumpCrisscross 1y ago
> I'd want to have a couple extra layers of fail-safes, and probably a human overseer with a kill switch This is the sort of feel-good rule making that stifles an economy. Driverless tractors exist. They’ve been deployed across the world. We have real-world data about their safety and precisely zero cases where they ran into schools because they got lost. There is a legitimate safety debate that can be had. But it should pit data against data, not hypotheticals.

Pet_Ant 1y ago

I mean a GPS geo-fence is pretty simple and fail safe. I'm more worried about something like humans unexpectedly in the field. Imagine a migrant crossing a field and getting run over by a combine.

aaronbaugher 1y ago
Yeah, that's a concern too. I'm not opposed to the idea of autonomous (or better near-autonomous) tractors working in fields. They'd definitely be safer than autonomous vehicles on roads. I just think people, especially at a tech-loving forum like this, are a little too quick to assume safety concerns are fully covered. Perhaps they can be, but will they be when large corporate farms cut corners on things like maintenance? I started driving tractors when I was 10, and I've been in a couple situations where equipment failure required some quick action to get stopped before serious damage was done. Shit happens, and should be expected to happen and prepared for.
ccozan 1y ago
But I expect a autonomous tractor - like an auto - to have a radar and be able to stop if an obstacle is there - not only migrants, how about wild animals, etc.

pfdietz 1y ago

A GPS-driven cutoff seems like an obvious thing to have in this situation, no?

nradov 1y ago
GPS (or GNSS in general) is hardly sufficient for anything safety critical. The signals are very weak and can easily be spoofed or jammed. Even overhead obstructions like trees and buildings nearby can cause significant offsets due to multipath reflection issues.
- zdragnar 1y ago
  RTK driven tractors are not using just standard GPS. They have positional accuracy down to one inch.
- immibis 1y ago
  Simple solution for that. If you jam GPS and it makes an autonomous tractor run over a school, you're liable for the manslaughter of the school kids. Trains can be derailed, but they're still allowed to operate and we just put the derailer in prison. And knives can be used to stab people. And someone can even grab your arm and force you to stab someone with the knife, and the law understands this.
  - nradov 1y ago
    What a stupid "solution". It is almost never possible to catch GNSS jammer users after the fact. GPS alone can never be sufficient, regardless of the legal issues. There are automated technical solutions to this problem involving sensor fusion and/or ground beacons.
    - pfdietz 1y ago
      Simply jamming the GPS would cause the vehicle to stop. What would be needed would be spoofing the GPS to make the vehicle think it was still operating normally, in the field it was supposed to be in, without any detectable quirks, a much harder task.
      - nradov 1y ago
        Jamming or not isn't a binary condition, nor is there a clear line between jamming versus spoofing. Certain types of signal interference will still allow for a position fix but with greatly degraded accuracy. There are algorithms to detect signal quality, but again that alone isn't considered adequate for anything safety critical. Commercial aircraft aren't legally allowed to fly with only GPS for navigation, and for good reason.

tonyhart7 1y ago

seems like someone didnt playing autonomous drone before, its a solved problem and Yes we do have OSS tool for that

bryanlarsen 1y ago
We even had solutions for that in the 70's when the law in question was written. But the solutions available at the time were far from fool-proof.

JumpCrisscross 1y ago

> I'd want to have a couple extra layers of fail-safes, and probably a human overseer with a kill switch This is the sort of feel-good rule making that stifles an economy. Driverless tractors exist. They’ve been deployed across the world. We have real-world data about their safety and precisely zero cases where they ran into schools because they got lost. There is a legitimate safety debate that can be had. But it should pit data against data, not hypotheticals.