4 ms·

The problem you are stating is called the 'handoff' problem which I believe will be a serious challenge for manufacturers of driverless cars. The handoff proble

by spuz 11y ago

The problem you are stating is called the 'handoff' problem which I believe will be a serious challenge for manufacturers of driverless cars. The handoff problem is when an automated system identifies a scenario in which it needs to hand over control to a human. How will that work in a vehicle that is not designed to be occupied by someone that knows how to control it? This problem has been written about in the book Our Robots, Ourselves [0] and there is an Econ Talk podcast about it as well [1]. To me, it makes sense to make use of the brain power provided by the occupant of the vehicle. For example, if one of the main vision sensors in the car breaks, why not let the human occupant drive the car manually? [0] http://www.amazon.co.uk/Our-Robots-Ourselves-Robotics-Autonomy-ebook/dp/B00SI02AWK http://www.amazon.co.uk/Our-Robots-Ourselves-Robotics-Autono... [1] http://www.econtalk.org/archives/2015/11/david_mindell_o.html http://www.econtalk.org/archives/2015/11/david_mindell_o.htm...

bryanlarsen 11y ago
As far as I understand, the current behaviour of self-driving cars is to stop the cars safely in a safe place, to the best of their degraded ability. Once it's stopped, the human can either take over. If a human can't, you're stuck on the side of the road and you call a tow truck. Which isn't really any different than any other sort of major mechanical failure.
- 0xffff2 11y ago
  That's all well and good if you make it to the side of the road, but what happens when the car decides it can't make it that far and stops in the middle lane of an Interstate?
  - matthewowen 11y ago
    What happens if your car breaks down such that it can't make it off the interstate? It's the same principle. You try to do the safest thing you can, but yes, at some point a car will fail and be stuck in the middle lane of an interstate because the autopilot fails, just as can happen because of mechanical failure.
  - geofft 11y ago
    I'd expect major mechanical problems with robots, like "the vision sensor died in the middle of an interstate," to be about as rare as major mechanical problems with humans, like "they had a heart attack in the middle of an interstate." If not rarer. We're already okay with the possibility that a human might drop dead in the middle of an interstate and lose control and more humans might be injured or killed as a result. As long as it is not more likely that a self-driving car will drop dead, I don't see why it's any different. (And this isn't even getting into things like the possibility of catastrophic failure being lower and the options for recovery being much higher. You can run two robots with redundant hardware; you can't run two humans quite as easily. A human who loses vision will lose their concentration; a robot that loses vision can use its last known positions of nearby cars to attempt to pull over immediately in an orderly fashion.)
  - icebraining 11y ago
    The other self-driving cars will cope with the detour much faster and graciously than humans could.
  - jonknee 11y ago
    The same thing that happens when any other car breaks down in the middle of the road--a traffic jam. Cars (especially ICE ones) break down all the time, we know how to deal with it.
a-priori 11y ago
Two options for how a driverless car without human controls might handle these sorts of hand-off scenarios: 1) It hands control to an out-of-car human driver, similar to how military drones and driverless trains are operated now. 2) It turns on the hazard lights and pulls to the side of the road. Another vehicle is dispatched to pick you up.
perlgeek 11y ago
From a safety engineering standpoint, handoff is a very, very bad idea. When you automate the common operation of a system, how is the operator supposed to get a feeling for the system? And when the operator doesn't handle the system regularly, how is he supposed to handle a situation that's so dangerous and unusual that the automation can't handle it? If you have years of experience driving a car, maybe it'll work. For some time. But what if your last manual intervention was five years ago? Or twenty years ago? The rarer handoff happens, the more stress it puts on the manual operator, and the more likely things go wrong. It's much better to have a safe default reaction (stop the car, shut down the reactor, ...) that kicks in when normal, automatic operation can't continue.
- spuz 11y ago
  You probably misunderstood what I meant by 'handoff'. It refers to the range of programmed interactions between an automated system and a human, for example, a warning light on the dash is a handoff. A warning sound or an automatic request for user action is a handoff. Pulling over and shutting down is a handoff. You say it's a bad idea. It would be a very very bad idea to build systems that do not have these responses programmed into them. It would be like throwing an exception without any exception handlers. If you haven't already, I highly recommend listening to the Econ Talk podcast for much more depth from someone who has been thinking about these kind of problems for years (I can't really do it justice in a HN comment).
- taneq 11y ago
  Thankyou! I've been saying this for a while, the current approach of 'driver assistance' or 'autopilot' is terrible. The more reliable the autopilot, the less attention the human 'driver' is going to be paying when the autopilot actually does say "okay turkey, you fly it" and shut off. And that is far more likely to happen in an unusual (read: hazardous) situation. You're self-selecting for a human panic reaction followed by a crash.

by spuz 11y ago

bryanlarsen 11y ago

As far as I understand, the current behaviour of self-driving cars is to stop the cars safely in a safe place, to the best of their degraded ability. Once it's stopped, the human can either take over. If a human can't, you're stuck on the side of the road and you call a tow truck. Which isn't really any different than any other sort of major mechanical failure.

0xffff2 11y ago
That's all well and good if you make it to the side of the road, but what happens when the car decides it can't make it that far and stops in the middle lane of an Interstate?
- matthewowen 11y ago
  What happens if your car breaks down such that it can't make it off the interstate? It's the same principle. You try to do the safest thing you can, but yes, at some point a car will fail and be stuck in the middle lane of an interstate because the autopilot fails, just as can happen because of mechanical failure.
- geofft 11y ago
  I'd expect major mechanical problems with robots, like "the vision sensor died in the middle of an interstate," to be about as rare as major mechanical problems with humans, like "they had a heart attack in the middle of an interstate." If not rarer. We're already okay with the possibility that a human might drop dead in the middle of an interstate and lose control and more humans might be injured or killed as a result. As long as it is not more likely that a self-driving car will drop dead, I don't see why it's any different. (And this isn't even getting into things like the possibility of catastrophic failure being lower and the options for recovery being much higher. You can run two robots with redundant hardware; you can't run two humans quite as easily. A human who loses vision will lose their concentration; a robot that loses vision can use its last known positions of nearby cars to attempt to pull over immediately in an orderly fashion.)
- icebraining 11y ago
  The other self-driving cars will cope with the detour much faster and graciously than humans could.
- jonknee 11y ago
  The same thing that happens when any other car breaks down in the middle of the road--a traffic jam. Cars (especially ICE ones) break down all the time, we know how to deal with it.

a-priori 11y ago

Two options for how a driverless car without human controls might handle these sorts of hand-off scenarios: 1) It hands control to an out-of-car human driver, similar to how military drones and driverless trains are operated now. 2) It turns on the hazard lights and pulls to the side of the road. Another vehicle is dispatched to pick you up.

perlgeek 11y ago

From a safety engineering standpoint, handoff is a very, very bad idea. When you automate the common operation of a system, how is the operator supposed to get a feeling for the system? And when the operator doesn't handle the system regularly, how is he supposed to handle a situation that's so dangerous and unusual that the automation can't handle it? If you have years of experience driving a car, maybe it'll work. For some time. But what if your last manual intervention was five years ago? Or twenty years ago? The rarer handoff happens, the more stress it puts on the manual operator, and the more likely things go wrong. It's much better to have a safe default reaction (stop the car, shut down the reactor, ...) that kicks in when normal, automatic operation can't continue.

spuz 11y ago
You probably misunderstood what I meant by 'handoff'. It refers to the range of programmed interactions between an automated system and a human, for example, a warning light on the dash is a handoff. A warning sound or an automatic request for user action is a handoff. Pulling over and shutting down is a handoff. You say it's a bad idea. It would be a very very bad idea to build systems that do not have these responses programmed into them. It would be like throwing an exception without any exception handlers. If you haven't already, I highly recommend listening to the Econ Talk podcast for much more depth from someone who has been thinking about these kind of problems for years (I can't really do it justice in a HN comment).
taneq 11y ago
Thankyou! I've been saying this for a while, the current approach of 'driver assistance' or 'autopilot' is terrible. The more reliable the autopilot, the less attention the human 'driver' is going to be paying when the autopilot actually does say "okay turkey, you fly it" and shut off. And that is far more likely to happen in an unusual (read: hazardous) situation. You're self-selecting for a human panic reaction followed by a crash.