Self Driving Cars: How well do they work in areas with haphazard driving rules?

[coppice]

Also, each one of us has to learn on our own and when we die, the knowledge we amass dies with us.

What a, frankly, silly thing to say. The fine points of epistemology and philosophy are one thing, but there's also what, least-ways around here in the East End of London, we call the bleedin' obvious. As soon as our species evolved language the problem of lost knowledge largely went away. All of us here have benefited from the accumulated knowledge of Isaac Newton (dead), Albert Einstein (dead), James Clark Maxwell (dead) and many others.

Of course, what cdev said is silly. However, I think what he is really talking about is amassed skill. That does die with us. In fact, it generally goes downhill badly well before we die.

[CatalinaWOW]

Also, each one of us has to learn on our own and when we die, the knowledge we amass dies with us.

What a, frankly, silly thing to say. The fine points of epistemology and philosophy are one thing, but there's also what, least-ways around here in the East End of London, we call the bleedin' obvious. As soon as our species evolved language the problem of lost knowledge largely went away. All of us here have benefited from the accumulated knowledge of Isaac Newton (dead), Albert Einstein (dead), James Clark Maxwell (dead) and many others.

I like Coppice's point.  Knowledge and skill are different things.  Knowledge is saying "rotate the steering wheel counterclockwise to turn left", or "apply carefully modulated pressure to the brakes".  Skill is the combination of that knowledge with muscle memory, vision, and other sensory input to do those things in a useful manner.  Knowledge is easy to transfer.  The knowledge part of self driving cars is a done deal.  The skill part is that tough problem that is only partly solved - by humans or robots.   Skill is harder to transfer between humans, but whatever passes for skills in robots should transfer pretty easily.  Robot skills are likely to be quite different from human skills.   Different data sets, different accuracies, different actuators and most likely different logic.

One thing that will lead to differences between robot skills and human skills is different goals and metrics.  Don't kill or injure anybody should be a common goal for both, but have fun will have no part of the robots goals.  One difference that is obvious from some comments in this thread is that robot drivers may have a goal to minimize total human travel time, while most human drivers have a goal of minimizing their personal travel time.  Similarly it is likely that the overall fleet of robot drivers will have the same or similar goals in terms of weighting fuel efficiency and speed, while these goals vary widely among human drivers.

[donotdespisethesnake]

Also, each one of us has to learn on our own and when we die, the knowledge we amass dies with us.

What a, frankly, silly thing to say. The fine points of epistemology and philosophy are one thing, but there's also what, least-ways around here in the East End of London, we call the bleedin' obvious. As soon as our species evolved language the problem of lost knowledge largely went away. All of us here have benefited from the accumulated knowledge of Isaac Newton (dead), Albert Einstein (dead), James Clark Maxwell (dead) and many others.

That is also a very silly thing to say, in the context of driverless cars i.e. the topic of the thread, and not the "how many angels fit on a pinhead" debate.

"Hello, this is your Captain speaking, welcome aboard. I've never flown a plane before, but rest assured I have read every book on the subject."

REALLY?

[Cerebus]

Also, each one of us has to learn on our own and when we die, the knowledge we amass dies with us.

What a, frankly, silly thing to say. The fine points of epistemology and philosophy are one thing, but there's also what, least-ways around here in the East End of London, we call the bleedin' obvious. As soon as our species evolved language the problem of lost knowledge largely went away. All of us here have benefited from the accumulated knowledge of Isaac Newton (dead), Albert Einstein (dead), James Clark Maxwell (dead) and many others.

That is also a very silly thing to say, in the context of driverless cars i.e. the topic of the thread, and not the "how many angels fit on a pinhead" debate.

"Hello, this is your Captain speaking, welcome aboard. I've never flown a plane before, but rest assured I have read every book on the subject."

Not the same, and you know it.

[Marco]

For the moment these system train, they don't learn. When they start being able to learn they'll be able to drive a car.

I found the philosophizing aspect of AI research a bit annoying in the past, but at least the philosophers used terms like strong AI and human level intelligence and drove research to chase it. Whereas modern "AI" research mostly pretend that their little toy systems deserve to be called intelligent, while they try to manipulate the language use in the field to make the words which suggest something better should be the aim unacceptable to utter.

AI research was more interesting without all the billions of VC bucks to create slightly better toy systems.

[stj]

anybody who wants to see where this leads should watch an anime called Ex-Driver.

a squad of people who go after out of control cars in a time when nobody else is allowed to drive!!
http://www.anime-planet.com/anime/ex-driver

given that was made 17 years ago, the japanese sure have a way of seeing what's coming!!

[Kalvin]

https://www.theguardian.com/technology/2017/jul/01/volvo-admits-its-self-driving-cars-are-confused-by-kangaroos

Probably they need to add more cameras and use Lidar.

About the lane markings: During the winter time there won't be necessarily any visible markings anyway. Using a Lidar for environmental feature set detection would possibly provide a solution as the road-side features typically stay in place no matter the weather conditions are, so basically no visible road markings are needed. The car could download the new feature set data and new navigation data from the net.

[KE5FX]

https://www.theguardian.com/technology/2017/jul/01/volvo-admits-its-self-driving-cars-are-confused-by-kangaroos

Probably they need to add more cameras and use Lidar.

About the lane markings: During the winter time there won't be necessarily any visible markings anyway. Using a Lidar for environmental feature set detection would possibly provide a solution as the road-side features typically stay in place no matter the weather conditions are, so basically no visible road markings are needed. The car could download the new feature set data and new navigation data from the net.

Ground-penetrating radar may prove helpful there.  Subsurface features of the road bed are unique and stable over time and changing weather conditions.  Some people at MIT have done some interesting work on that.

[BrianHG]

UK TV 9pm tonight.

Horizon: Dawn of the Driverless Car

http://www.bbc.co.uk/programmes/b006mgxf/broadcasts/upcoming

Just watched it.  Ok, simple explanations but covers the goals and 1 of the companies shown was crap while the others are advancing.
The biggest problem I see is that maybe these self driving cars, (with too many sensors in my opinion, they got to get at least as good as a human purely visually interpreting, then worry about extra improved sensing as a powerful +), only seem to be worried about the cars software being aware of itself exclusively.  IE, Stay on the road.  Avoid accidents.  Navigate to set destination.  The software isn't going anywhere near putting itself in the mind of the other cars and drivers on the road.  If you want proper avoidance, the software should look out for thing like stay extra far away from someone who appears to be driving slightly drunk or making driving errors on the road like a student driver, or they the other drivers in oncoming traffic is making bad choices in poor weather, or, a different self driving AI from another vehicle appears to be not as good at merging or might not see you properly.
Also, here in Montreal, during rush hour, just to make some highway exits, or, left turns at some intersections, you must drive with forceful attitude or intent, otherwise, you will never get anywhere.  How will self driving cars express this, or, take advantage of another drivers slightly slow advancing habits and forcefully squeeze itself into the merging lane.  Human drivers will learn very quickly to never give that 'safe space' the self driving cars would require before they would perform a lane change and the self driving cars will always loose out or miss intersections or stop in one lane waiting for an opening which will never come during rush hour before they could get to where they were instructed to go, blocking the lane they are in, making rush hour even worse for everyone else behind them.  These cars will be despised under these circumstances.

[KE5FX]

Human drivers will learn very quickly to never give that 'safe space' the self driving cars would require before they would perform a lane change and the self driving cars will always loose out or miss intersections or stop in one lane waiting for an opening which will never come during rush hour before they could get to where they were instructed to go, blocking the lane they are in, making rush hour even worse for everyone else behind them.  These cars will be despised under these circumstances.

This is indeed a good point.  The hardest part of designing a self-driving car will be making it compatible with human drivers.  If I have reason to believe that the person who's signaling to move in front of me will fanatically adhere to a speed limit which has been set below the 85th-percentile goal for revenue-collection purposes, of course I'll try to keep them from merging.  And obviously no one will build the capacity to break traffic laws into a self-driving car.  It'd be professional suicide.

[hamdi.tn]

Autonomous Vehicles vs. Kangaroos. 

http://spectrum.ieee.org/cars-that-think/transportation/self-driving/autonomous-cars-vs-kangaroos-the-long-furry-tail-of-unlikely-events