Driverless cars... how far off are they now?

Forums › Transport (cars, bikes and boats) › Driverless cars... how far off are they now?

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9

20141 posts

Uber Geek

#1952507 6-Feb-2018 16:35

tdgeek:

I feel the tech that would make driverless cars work well is not that far away. Even if a glitch causes an accident or a death that is still far safer than humans as tech will do what its told, humans don't, or make an unintended or careless mistake

The difference is that driverless cars will likely make errors when there are grey area situations that it's rules can't work out, due to operating in a human world. That will likely be their main problem. Also in the case of crashes, it has to decide whether it is better to save the car owner and passengers, or pedestrians, and which will potentially save the most lives. The whole insurance / liability thing will be a bit of a nightmare.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1952535 6-Feb-2018 16:50

mattwnz:
tdgeek:

I feel the tech that would make driverless cars work well is not that far away. Even if a glitch causes an accident or a death that is still far safer than humans as tech will do what its told, humans don't, or make an unintended or careless mistake

The difference is that driverless cars will likely make errors when there are grey area situations that it's rules can't work out, due to operating in a human world. That will likely be their main problem. Also in the case of crashes, it has to decide whether it is better to save the car owner and passengers, or pedestrians, and which will potentially save the most lives. The whole insurance / liability thing will be a bit of a nightmare.

I agree. As I said earlier it has to be right. A car can see what we can see and it should make the best decision rather than a human taking 3/4 second to react then making a good or poor decision . As you imply, lots to sort out

kingdragonfly

11190 posts

Uber Geek

Subscriber

#1952557 6-Feb-2018 18:12

Tesla is level 2, which is still very impressive. This is a level 4 prototype

alasta

6701 posts

Uber Geek

Trusted

Subscriber

#1952570 6-Feb-2018 19:21

I also wonder how they will deal with detecting hazards for speed management purposes. Is that a cat at the side of the road about to run onto the road? That ball that just rolled onto the road, could a child rush out chasing it? What about that poorly loaded trailer attached to the vehicle ahead - is something about to fall off it?

There must be so many scenarios like this that a machine would struggle to interpret.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1952580 6-Feb-2018 19:41

alasta:

I also wonder how they will deal with detecting hazards for speed management purposes. Is that a cat at the side of the road about to run onto the road? That ball that just rolled onto the road, could a child rush out chasing it? What about that poorly loaded trailer attached to the vehicle ahead - is something about to fall off it?

There must be so many scenarios like this that a machine would struggle to interpret.

Yes. I see Infra Red as the solution there

Dratsab

3946 posts

Uber Geek

Trusted

Lifetime subscriber

#1952645 6-Feb-2018 20:52

epr: How will they teach self driving cars who to kill in an accident situation. Do I run over 1 child to save 2 old people or do I kill any one to protect my owner at all costs.

My view is it will try and stop to protect the occupants. Autonomous vehicles (AV's) are not sentient beings and will be programmed to act to the benefit of the occupants in the most rapid fashion. This means not having to figure out the best outcome between a host of choices. If manufacturers have to wait for commercial technology to develop to the point where vehicles can scan everything around them and recognise/differentiate life forms, then have to delay any reaction (possibly compounding problems) by weighing up the probable outcomes of a series of possible actions, the idea of a society with autonomous vehicles becomes even more remote.

frankv: *I* would also have to be confident that your autonomous car wouldn't harm me or my car, whether I was walking or cycling or driving. And of course there's a lot of skeptical, conservative neo-Luddite people out there whose thoughts also need to be considered.

What's your level of confidence a human driven car car won't harm you or your car, whether you were walking or cycling or driving? Mine's pretty low. There's already vehicles out there which slow down and even brake quite heavily while on cruise control. Collision avoidance sensors are not exactly new tech. As for the others you mention, they could form a lobby/protest group :-)

One of the things I'm most interested in is the common communications protocol that's being proposed for autonomous vehicles - how it'll be implemented and what technologies will be used. There's lot's of places in lots of countries where there's no cell signals etc so when they lose signal, will they be happy to disconnect from the cloud and broadcast some form of open connectivity WAN to sense and talk with other vehicles in the area? Then there's the security implications - someone who manages to hack the final protocols could wreak absolute havoc.

Dratsab

3946 posts

Uber Geek

Trusted

Lifetime subscriber

#1952651 6-Feb-2018 21:10

Aredwood: The edge cases will be the difficult ones. That person standing in the middle of the road, waving their arms about. Are they a drunk teenager, or a Police officer directing traffic? Especially as Police officers can override pretty much any road rules. Say a road is blocked due to an accident, and the police are redirecting traffic the wrong way down a 1 way street. The officer can easily tell a human driver - drive up that street, the police blocking the other end will let you through into the main road. How would a Police officer tell those instructions to the computer in a driverless car?

What about roadworks, where you sometimes get sections of road without any lines painted? Or lots and lots of road cones, all of the cones are the same, but the car has to then figure out their pattern.

Or a simple lane closure at an intersection. With traffic being directed to drive in a bus lane, or left and straight ahead traffic to use the right turn only lane.

And what about when the drunk teenagers figure out how to fool the driverless cars into thinking they are police officers. And they direct the driverless cars into a Wilson's parking building. Meaning the cars are now trapped, as the cars are not fitted a robotic arm to put coins into the payment machine.

Great questions/concerns! I think a lot of them could be answered by proper use of the common communications protocol. For example police comms centres updating the cloud which vehicles are [at least in areas with the right coverage] connected to or getting the responsible local authorities to do so where necessary. Road working companies could be required to upload/manage plans within the cloud, or perhaps some sort of localised WAN where there's no cloud capable coverage. Of course, getting the instructions removed post-event would be just as important! Perhaps Police might have to wear some sort of reactive tag that identifies them to AV's...

Equinox IT

Cloud spending continues to surge globally, but most organisations haven’t made the changes necessary to maximise the value and cost-efficiency benefits of their cloud investments. Download the whitepaper From Overspend to Advantage now.

frankv

5680 posts

Uber Geek

Lifetime subscriber

#1952673 6-Feb-2018 21:45

kryptonjohn:

frankv:

jpoc:

I would be confident that I would have a safer ride.

It's not all about you ;)

*I* would also have to be confident that your autonomous car wouldn't harm me or my car, whether I was walking or cycling or driving. And of course there's a lot of skeptical, conservative neo-Luddite people out there whose thoughts also need to be considered.

But maybe there's a niche for a small autonomous vehicle (e.g. a largish scooter or golf cart)... light so that if it did make a mistake, it couldn't kill or seriously injure someone. This kind of thing might be useful for an automated courier or lab samples or whatever delivery system, for example.

Like jpoc I suspect the self drive cars will be safer than the human driven ones. Although they are subject to design errors and software bugs, they are not subject to drunken, incompetent or boy-racer control.

You're probably right. But you're going to need to be a whole lot more certain than "suspect" and "probably" and "confident". And so are the majority of the population.

Dunno about excluding boy-racer control though... I suspect that there's going to be a community of people who hack their cars to get extra performance.

frankv

5680 posts

Uber Geek

Lifetime subscriber

#1952674 6-Feb-2018 21:47

networkn:

I'd suggest a golf cart running into you at 30kmph is very likely to cause you serious injury.

Maybe. But it's not going to be as bad as a 40 tonne truck running into you. So small, light autonomous vehicles might be acceptable where larger ones aren't.

blakamin

4431 posts

Uber Geek
Inactive user

#1952675 6-Feb-2018 21:58

People that think Chevy will release one to the masses next year are dreaming... they might make about 500 a year, and then recall and destroy them... GM are good at that.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1952740 7-Feb-2018 07:16

Dratsab:

If manufacturers have to wait for commercial technology to develop to the point where vehicles can scan everything around them and recognise/differentiate life forms, then have to delay any reaction (possibly compounding problems) by weighing up the probable outcomes of a series of possible actions, the idea of a society with autonomous vehicles becomes even more remote.

I believe All that already exists and is not new technology

Detection of hard items already exists, and its in cars that have auto parking. Radar was in WW2

People and animals is infraRed

Numberplates, thats at airports and metro refuse stations

Mapping is old

Modern computers, that's circa 1980

All that can be enhanced by special paints or additions to make it better and in bad weather. Reaction time of a computer blitzes humans who are 3/4 second. Decision making is humans best part, AI, but we often fail. If a looming accident required quick reaction, a decision on what to avoid to minimise injury, and how to do that, a computer will ace a human all day

There isn't 25 million situations a car needs to know, it's probably 1000, or less.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1952741 7-Feb-2018 07:20

Dratsab:

One of the things I'm most interested in is the common communications protocol that's being proposed for autonomous vehicles - how it'll be implemented and what technologies will be used. There's lot's of places in lots of countries where there's no cell signals etc so when they lose signal, will they be happy to disconnect from the cloud and broadcast some form of open connectivity WAN to sense and talk with other vehicles in the area? Then there's the security implications - someone who manages to hack the final protocols could wreak absolute havoc.

That's old too, cellular and others. If there was an accident or an issue that was not covered by the software, the car pulls, over. It calls home to send a video, and the logs. Doesn't need to be real time, just when it can. That exists now in OS's

You last point is very valid. We see robbing banks be replaced by internet theft, spies replaced by cyberspies, so a very valid point.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1952743 7-Feb-2018 07:23

frankv:

networkn:

I'd suggest a golf cart running into you at 30kmph is very likely to cause you serious injury.

Maybe. But it's not going to be as bad as a 40 tonne truck running into you. So small, light autonomous vehicles might be acceptable where larger ones aren't.

If larger ones aren't and smaller ones are, that defines that autonomous is unsafe.

MikeAqua

7773 posts

Uber Geek

#1952837 7-Feb-2018 11:09

I see the ethical-choice dilemmas posited for autonomous cars as as a red herring.

IMO no company will programme a car to deliberately risk sacrificing its occupants, for any reason. This is because the first time a car implemented that approach guidance, its manufacturer would be sued into oblivion, by survivors or their relatives. I reality such situation will arise rarely. Some reasons (all conjecture):

Autonomous cars will be operated more slowly and highly defensively (in particular when sharing road space with non-autonomous vehicles).

Cars will eventually share their locations wirelessly.

Autonomous cars will have greater ability to anticipate risk of collisions and avoid them.

Autonomous cars will have greater ability to decelerate and manoeuvre and could be built to undertake mitigating actions like deploying external air-bags.

IMO where an autonomous car cannot avoid collision, it will be programmed to stay on the roadway (moving laterally only if it can do so without risking another collision) and to undertake avoidance and mitigation operations. This happens to a limited extent now.

If I brake very hard in my 2015 Mazda 3 (not autonomous), it activates SRS, activates the hazard lights, and pre-tensions the seatbelts. Some cars will apply the brakes themselves.

As to the cat ... an autonomous car could from a safety perspective disregard very small animals like cats. From a safety perspective, why even risk whiplash from heavy braking to avoid fluffy?

Mike

MikeAqua

7773 posts

Uber Geek

#1952842 7-Feb-2018 11:15

tdgeek:

Radar was in WW2

People and animals is infraRed

Radar will pick up animals. It's used to detect flocks of birds by recreational fisherman (as an indicator of fish activity).

For example

Mike

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9