Hacker News Reader

Comment by noonespecial | Hacker News Reader

Its not really important what speed driverless cars actually settle on. What matters is that they are consistent. The most dangerous thing about human drivers is that they are erratic.

I would not be at all surprised in the coming world of robo-cars to find that humans who still wanted to drive themselves will need special licensing and training, much like pilots of today, if nothing more than to learn to be predictable.

View on HN

Actually it is very important from a fuel economy perspective. Air resistance is quadratic in velocity, so a car that gets 35mpg at 65mph will get, say, 20mpg going 85. On the other hand, the possibility of convoy driving (http://phys.org/news/2011-05-drivers.html) means you might be able to have your cake and eat it too: rapid travel with little efficiency loss from wind resistance.

I'm not sure I'd feel safe about convoy driving unless there were some kind of way to manage emergencies (like a tire blowout, engine problem, deer) which under normal circumstances isn't very dangerous because you have a second or two to manoeuver. But in a close convoy system, there is little room for error and even an automated system may be unable to compensate for the emergency in a timely fashion thus resulting in a chain reaction crash --the kind we see in foggy conditions with human drivers.

I'm sure that convoy driving is another 15-20 years in the future from the point where self-driving takes off. However it's not like those dangers you mention cannot be managed. Self-driving cars already require a wide array of sensors - together in a convoy you get such an amount of data that the computers should be able to get out of hairy situation at any point.

Vehicle 'platoons' have long been a subject of traffic studies.

Here is a cool video from CAR 2 CAR... http://www.car-to-car.org/index.php?id=132&L=mbgiqdmn

One option to reduce convoy risk is have cars attach to one another such that no single car can cause a failure in the overall system (similar to trains).

We shouldn't think of automated cars as independent as we do. They are in fact simply compartments.

With that in mind, imagine cars attaching to each other to form trains and break apart as they begin to reach the desired destination.

This also has impacts on fuel efficiency since drag can be reduced across the entire system.

(Cars don't have to be the same model, they just need a formalized attachment system on the front / rear).

Yeah, I think buffers (like in trains) could be an option, maybe magnetic. If it requires physical contact, though, that would mean a whole train of cars (maybe cars would be grouped into sets of ten or so?) would have to speed up and down as cars joined and left the caravan --I know we do this in everyday driving, but it's "softer".

[edit, just thinking, if we knew destination in advance, the first car would be the one going furthest and the last car in the group going the shortest so that the interruption would be inconsequential]

Back to accidents, there will the edge cases. And because it's programmed (and not a human "accident") people might find more fault in the system than they would in a human. So even though accidents and deaths would/could be reduced considerably, those that would occur might be seen as more negligent than accident. Just thinking out loud.

Physics suggests otherwise.

Reply on news.ycombinator.com