Uk to allow Driverless Cars

[pcstru]

Sure, once someone starts making these things in bulk, prices will come down but initially, that is now and next few years, the technology is unlikely to be within the reach of the average car buyer. So lawmakers will likely have to address the issue of "in charge of" as it relates to an autonomous vehicle.

 

I would guess there is a reason Google are using a 64 laser LIDAR capable of resolving features a few mm across. I'm also sure (having built a few robots) that the computing power you need to to real time analysis of that 3d environment is not insignificant. You can make your own LIDAR on the cheap, but you probably wouldn't want to be driven anywhere by a computer which relies on it.

[unixman_again]

I suppose a driverless car won't do this sort of thing Satnav plunges £96k Merc into river. On the other hand, maybe it might happen more often.

[pcstru]

I suppose a driverless car won't do this sort of thing Satnav plunges £96k Merc into river. On the other hand, maybe it might happen more often.

 

It's a bit of a misleading headline - "stupid driver plunges car into river" would be more accurate.

 

I was once directed to drive over a cliff by a Satnav system in Italy, the road it wanted me to take emerged 500ft below from a tunnel in the mountainside.

[tommej]

Railway crossings will be the big headache. They aren't uniform (different angles, gate types and so on), they are often abused by other drivers and in some areas are hard to program to avoid. And if the car gets it wrong just once...

 

 

They're quite intelligent!

[sonofsanta]

I read an interesting article on driverless cars a while ago - all around the ethics.

 

For example - how do you program a car to crash - they will at some point.

 

Say a child runs out from behind a wall. There was no indication what so ever that it could have been there, and the car needs to decide - take evasive action, possibly into oncoming traffic, or hit the child? If we as individuals behind the wheel were put in that decision, we'd make different decisions. Driverless cars quite possibly make the decision based on the person who wrote the program - and in the event of an identical accident it would make the same decision.

 

How about a crash where the driverless car is going to hit one of two motorbikes. One rider is wearing a helmet, the other (stupidly) is not. How do you program a car to make a decision? The rider with the helmet is statistically more likely to live - but should they be punished for wearing the correct equipment and following the laws or should the car be programmed to hit the rider who seems to take less regard of their life?

 

Do we need to humanize the program and add in a random number generator to decide someones fate?

 

I thought it was a fascinating subject.

 

I think I read the same article, or a similar one (seems there were a few of them). Further to all the above was the question: would you car choose to kill you? Imagine your front farside tyre blew on a bridge, steering you towards oncoming traffic. You, in your big modern robot SUV might survive, but the two occupants of the car you'd hit would be killed. Would your car instead choose to steer you off the bridge, certainly killing you but saving two other lives? Mathematically the latter is sounder but it's morally thorny. It's a twist on the Trolley Problem that some poor software engineer is going to have to decide on.

 

Other articles on the topic:

The Robot Car of Tomorrow May Just Be Programmed to Hit You | Autopia | WIRED

Google and the Trolley Problem | Owen abroad

When Should Your Driverless Car From Google Be Allowed To Kill You? - Forbes