The latest chapter for the self-driving car: mastering city street driving • Hacker News Board

Re: The latest chapter for the self-driving car: mastering city street driving

#61

Post by dalke » Mon, Apr 28, 2014, 1:37 PM UTC

cbhl wrote: post #31

nabla9 wrote: post #18

Earlier quoted context omitted.

Somebody correct me if I'm wrong, but Google cars still can't drive in the rain or in the snow. Not even on a wet street. They have driven all those thousand miles in Mountain View, Calif. in good weather. Generally lidar's don't work well with rain or reflective surfaces.

My understanding is that it uses vision processing algorithms on a camera nowadays, and works just fine in rain. Snow is an issue because the car can't see the lane markings through the snow -- a human drives through snow by remembering where they were or making a guess.

"... works just fine in rain". For some levels of rain. ;) I've been in some downpours so strong I couldn't see the lane markings.

Re: The latest chapter for the self-driving car: mastering city street driving

#62

Post by icebraining » Mon, Apr 28, 2014, 1:38 PM UTC

shekhar101 wrote: post #11

I saw an interesting comment on the blog post. Is anyone keeping tab on the project knows if Google in working on some sort of communication between vehicles on road. That will make the self driving cars much more accurate. Lot of 'sudden' circumstances will be predictable by the car.

Many manufacturers are working on it, and in fact the Feds are already planning to standardize and mandate them:

http://www.usatoday.com/story/money/cars/2014/02/03/nhtsa-ve...

https://www.dot.gov/fastlane/v2v-cars-communicating-prevent-...

Re: The latest chapter for the self-driving car: mastering city street driving

#63

Post by mikepurvis » Mon, Apr 28, 2014, 1:38 PM UTC

troels wrote: post #22

arikrak wrote: post #10

It won't be able to drive around New York if it follows the rules so strictly. It will never be able to turn. They'll have to add a setting for aggression...

That's funny, but probably a very real challenge. As a (human) driver, it happens that we come across ambiguous, contradictory or just plain impossible constraints and we overcome them by deliberately breaking the rules. I wonder how they would cope with that?

This is part of the larger problem of roboethics—prioritizing and balancing a set of conflicting objectives and requirements. There's a neat introductory article with some video of example scenarios here:

http://footnote1.com/what-should-a-robot-do-designing-robots...

Re: The latest chapter for the self-driving car: mastering city street driving

#64

Post by logikblok » Mon, Apr 28, 2014, 1:38 PM UTC

Genuine question, can anybody answer how these cars will: 1) Allow for situational aspects of driving eg I need to get there quickly so might drive faster because I'm running late? 2) Protect from people intentionally take advantage of safety mechanisms eg cutting in front, stepping out onto the road?

Re: The latest chapter for the self-driving car: mastering city street driving

#65

Post by cake » Mon, Apr 28, 2014, 1:38 PM UTC

bertil wrote: post #27

antr wrote: post #3

I love how the car understands the cyclist's hand signals. The self-driving car is going to have a gigantic impact on the world, I can't wait to see it commercialised.

As a cyclist, I was shocked to see the cyclist change his mind like that —unless he was a tester too, that’s how you get bumped– but I really hope Google Cars learn that most cyclists actually don’t signal their turn that well. Cars don’t bother either, and the last frame show how most non-car drivers get killed: don’t notice what passing just being, shift in gear, swerve and… bang. Seeing how Google taught its car t…

It really looks like the cyclist is a tester.

I agree, most cyclist don't signal at all, sometimes they will just subtly point with their finger.

Not the best idea to rely on the cyclist's signals.

Re: The latest chapter for the self-driving car: mastering city street driving

#66

Post by barrkel » Mon, Apr 28, 2014, 1:39 PM UTC

gamegoblin wrote: post #21

rwmj wrote: post #5

This is hardly city driving. Try driving in a European city. (I love this project and can't wait to buy or rent a self-driving car)

Harder yet, Southeast Asia. Hundreds of pedestrians, tons of vehicle variety (some entirely makeshift), uncontrolled intersections, very little regard to any regulations.

I think swarm-like logic might work better at that level.

Re: The latest chapter for the self-driving car: mastering city street driving

#67

Post by jaredmcateer » Mon, Apr 28, 2014, 1:39 PM UTC

cbhl wrote: post #31

nabla9 wrote: post #18

Earlier quoted context omitted.

Somebody correct me if I'm wrong, but Google cars still can't drive in the rain or in the snow. Not even on a wet street. They have driven all those thousand miles in Mountain View, Calif. in good weather. Generally lidar's don't work well with rain or reflective surfaces.

My understanding is that it uses vision processing algorithms on a camera nowadays, and works just fine in rain. Snow is an issue because the car can't see the lane markings through the snow -- a human drives through snow by remembering where they were or making a guess.

And if you ever look at a large parking lot after a thaw when it had been snowing earlier in the day you can see we're also horrible at predicting where the lines are and, for the most part, use other drivers as our guide.

Re: The latest chapter for the self-driving car: mastering city street driving

#68

Post by lojack » Mon, Apr 28, 2014, 1:40 PM UTC

lmg643 wrote: post #9

somehow the current scuffle between uber and regular cabs seems to miss the joke. we're going to need a plan for the eventual unemployment of a lot of cab drivers, regardless of how the cab is hailed.

Someone proposed some food for thought a little while ago that intrigued me. One day in the not too distant future it'd be completely reasonable for cabs to drive themselves and automatically accept payment via bitcoin or some other digital currency. It'd also be reasonable for them to refill their fuel tanks when empty, paying with the money they received, and additionally go into the shop whenever their diagnostics detect an issue.

At this point, they may not be self aware, but they are completely autonomous, capable of making their own living wages, and can look after their own well-being.

Re: The latest chapter for the self-driving car: mastering city street driving

#69

Post by maxerickson » Mon, Apr 28, 2014, 1:40 PM UTC

seren wrote: post #52

maxerickson wrote: post #36

Earlier quoted context omitted.

Even New York City seems to only have tens of thousands of cab drivers ('seems' because I just looked at wikipedia for a minute), I don't think they are that big a percentage of the population.

I think the real markets are long haul trucking and small delivery van for the last miles. All in all it is probably a significant portions of job.

There are several million truckers in the U.S., so it's somewhere around 1% (but it's probably less than 2%).

I guess there are probably lots of delivery drivers, but those vehicles will still need tenders (at least until robots can do the last 100 feet of the delivery).

Re: The latest chapter for the self-driving car: mastering city street driving

#70

Post by yardie » Mon, Apr 28, 2014, 1:41 PM UTC

It's amazing how new technology is meeting old technology again. For example, the horse drawn carriage.

Typically, the carriage driver doesn't steer the horse but points it in the direction they want to go. The horse and it's little horse brain negotiates the terrain and immediate obstacles.

You never had to steer the horse to avoid driving over the cliff, it was smart enough to know self preservation. The modern mechanical car doesn't even have that level of avoidance systems. This left the driver to manage more strategic tasks. There were downsides obviously. Horses could freak out and run over a crowd of people.