News for hackers. Stuff that matters.
Re:Google's Driverless Car Is Worth Trillions by bcoates on Thursday January 24, 2013 @03:25AM (#5108349)
Driverless cars would be a fantastic improvement to society, but I think you might be underestimating how much people will compensate. Traffic rises to fill roads, and only a small part of public transport budgets are for drivers (it looks like around 10% for LA)
Re:Google's Driverless Car Is Worth Trillions by surrealize on Thursday January 24, 2013 @03:45AM (#5108422)
> only a small part of public transport budgets are for drivers (it looks like around 10% for LA)

Well, you also have to consider that bus systems are built around driver scarcity. What if buses were half as large, and run twice as often? They can't now, because that would double driver labor costs.

If buses were the size of vans, capital costs might even decrease due to having access to van-market economies of scale (the market for vans presumably being significantly larger than the market for buses).

Re:Google's Driverless Car Is Worth Trillions by justincormack on Thursday January 24, 2013 @04:09AM (#5108508)
Many countries with lower labour costs this eg the Dolmus in Turkey https://en.wikipedia.org/wiki/Dolmu%C5%9F
Re:Google's Driverless Car Is Worth Trillions by pekk on Thursday January 24, 2013 @05:46AM (#5108848)
Also all over Africa and India. It's a very common pattern outside the weird US
Re:Google's Driverless Car Is Worth Trillions by fulafel on Thursday January 24, 2013 @04:57AM (#5108685)
The sensible model seems to be called "flex service" or "demand responsive transit" (1): Dynamically routed small shuttles where customers place trip orders to the transit company, and the routing system figures out paths for the shuttles so customers get where they want to go in reasonable time.

(1) https://en.wikipedia.org/wiki/Demand_responsive_transport

Re:Google's Driverless Car Is Worth Trillions by eli_gottlieb on Thursday January 24, 2013 @06:40AM (#5109024)
You mean share-taxis? They're relatively common already in many places.
Re:Google's Driverless Car Is Worth Trillions by Cushman on Thursday January 24, 2013 @12:53PM (#5110943)
Well, there's no force that makes traffic "rise to fill roads"; there's externalities to having increased traffic capacity that may cause transportation needs to increase. We still have a lot of underutilized roads.

The main difference in terms of capacity, I think, is that once we have majority- and fully-autonomous roads, the system will actually not allow the road to exceed capacity. Individual vehicles will respond to signs of lower throughput by getting off of the potentially congested road, while centralized signaling systems will help coordinate routes and departure times to optimize-- and due to the non-Newtonian nature of traffic, keeping throughput high actually increases road capacity.

So if you get into your car at 9am and say, "Take me to work," it's not going to say, "Time with traffic: One hour," it's going to say, "We should depart in twenty-five minutes." Even better, if you are subscribed to a service like most Americans will be, you'll instead tell them the day or week or month before, "I need to be at work by 9am," and they'll say, "There will be a fifteen dollar capacity fee. Would you rather arrive at 8:30 or 10?" And let you and your work decide exactly how much of the economy's money it's worth for you to be there at exactly the same time as everyone else in the time zone.

Re:Google's Driverless Car Is Worth Trillions by bcoates on Thursday January 24, 2013 @02:50PM (#5111753)
That's way beyond driverless car, that's a global optimization problem. If it were possible, you wouldn't even need driverless cars to implement it. System-to-car communication and slightly more complicated street lights would be enough.

For example, many cities already have lit signs that forbid turns from cross-streets onto oversubscribed roads, your nav system could allow you to pay the congestion fee and get authorization to disregard the sign and take the turn. Similarly, you could get out your smartphone and queue yourself to a destination and be told when you get to leave and take the fast route for free. As long as queuing at home got you to your destination faster than queuing in traffic, it would be the individual best strategy to do it and enough people would use it for it to work.

The hard part is probably doing traffic scheduling at an acceptable latency, for a city like LA you'd need to model the behavior of millions of cars however many iterations your optimization algorithm needs every few seconds. These kind of problems are really computationally expensive, it may not be economical with current systems and algorithms.

Re:Google's Driverless Car Is Worth Trillions by Cushman on Thursday January 24, 2013 @03:34PM (#5112035)
> That's way beyond driverless car, that's a global optimization problem. If it were possible, you wouldn't even need driverless cars to implement it. System-to-car communication and slightly more complicated street lights would be enough.

I disagree, I'd say scheduling is the smallest part of it. On most trips I take, there's a single major road that covers most of the distance, and I'm most likely to be delayed by traffic on that road slowing or halting completely. It's the moment-to-moment perverse behavior of other drivers when it comes to merging, changing lanes, accelerating and braking that causes that to happen, and having intelligent drivers on the same roads at the same times would make a much more significant impact.

In other words, you're saying we could do global scheduling with traffic lights, which is true. I'm saying that with sufficiently competent drivers we won't need the lights, and I think that will cause a qualitative shift in the nature of traffic.