How Autonomous Cars' Safety Features Clash With Normal Driving
An anonymous reader writes: Google's autonomous cars have a very good safety record so far — the accidents they've been involved in weren't the software's fault. But that doesn't mean the cars are blending seamlessly into traffic. A NY Times article explains how doing the safest thing sometimes means doing something entirely unexpected to real, human drivers — which itself can lead to dangerous situations. "One Google car, in a test in 2009, couldn't get through a four-way stop because its sensors kept waiting for other (human) drivers to stop completely and let it go. The human drivers kept inching forward, looking for the advantage — paralyzing Google's robot." There are also situations in which the software's behavior may be so incomprehensible to human passengers that they end up turning it off. "In one maneuver, it swerved sharply in a residential neighborhood to avoid a car that was poorly parked, so much so that the Google sensors couldn't tell if it might pull into traffic."
Ban human drivers.
"One Google car, in a test in 2009,..."
One would think that in 6 years some improvements would have been made. Do we have a more current example?
09 F9 11 02 9D 74 E3 5B - D8 41 56 C5 63 56 88 C0 45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
Millions of people on the road today deserve to have their license taken from them because they can't follow simple rules like signaling, not parking halfway out into the street and leaving enough room to brake in case the car in front of you brakes.
The article summary isn't very good. If the software is programmed in a way that causes a car to behave in a way that's dangerous, it IS the software's fault.
That's trivial but true.
It becomes interesting when the software has the car behaving in a way that is SAFE, but unexpected.
bickerdyke
This. All the studies that I've seen boasting about the enormous time advantages of self-driving cars ignore the fact that most human drivers tend to cruise from 5 to 15 MPH over the posted speed limit on many interstates and highways. I can't imagine a self-driving car being designed so as to operate above the posted speed limit in self-driving mode. Unless a second set of roads or a second set of rules is created for autonomous vehicles, you're going to have a difficult time convincing people of the advantage of being slower than anyone else on your morning commute.
Computers follow rules. Humans (a.k.a every other asshole on the road) do not.
This is a no win situation. If you program a car to drive safely and follow rules, then it won't be safe on roads because of all the assholes who don't. If you program the car to behave more like an asshole ( a human driver), then it won't be safe since there's a good chance it will make the wrong call. If you program the car to just account for assholes but still drive safely, then it will basically choke in situations like a four way stop in southern California where every other asshole will just muscle or roll their way through the stop.
The long pole in the tent isn't developing an AI capable of driving. It's developing an AI that can deal with assholes.
~X~
If all the cars were autonomous the morning commute times could be cut in 1/2 or 1/3rd without changing the speed limit since rush hour style rubber band stop and go traffic would be a thing of the past.
Program to take account of these things, or don't plan on driving on the road.
Duh.
Technology in development is imperfect. Big surprise. These issues are why Google hasn't yet started selling them to the public. None of them are insurmountable, but it takes a lot of time and effort to build sophisticated systems.
What if that was a cardboard box and it swerved heavily in case that box "pulled out"?
The cars can easily distinguish between a cardboard box and a vehicle. Determining whether or not the vehicle has a driver in the seat and might move... that's often impossible. Likely the reason that the car swerved sharply rather than braking earlier is because the badly-parked car was obscured by other obstacles.
If it can't make it's way through a junction where the drivers are following the rules, that's bad programming.
Six year-old programming, note. The article mentions that the current version of the software inches forward to establish intent to move.
and potentially weighs up collision with non-hazard vs collision with small child and gets it wrong
Google cars recognize pedestrians (of all sizes) and regularly notice them even when no human could. I'm sure the car would choose to hit another vehicle over a pedestrian or cyclist.
Really, your whole comment is a mixture of outdated information buttressed by invalid assumptions and layered over with a veneer of blindingly obvious conclusions.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
It's not a no-win situation. It just means that self-driving cars have to know when to break the rules. They can and should behave like the best of human drivers.
If you program the car to just account for assholes but still drive safely, then it will basically choke in situations like a four way stop in southern California where every other asshole will just muscle or roll their way through the stop.
The current programming of the car handles that situation. Less aggressively than a human would, but aggressively enough to assert its intention to go, and go.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
with ABS, airbags and selectable 4WD. Window cranks, inert key and levers on the seat. Cheaper, less to break, and I think I can still manage.
"Win treats sysadmins better than users. Mac treats users better than sysadmins. Linux treats everyone like sysadmins."
A slow commute isn't such an issue if you can spend it relaxing or working instead of driving... and even speeding by 15 MPH only saves a few minutes on a commute.
This space intentionally left blank