Log in

No account? Create an account
04 February 2008 @ 10:46 pm
Random thought after using public transportation for the day. I just finished "reading" Vernor Vinge's "Rainbows End" (have i mentioned before that that is a very awesome book?) Anyways, the robotic cars in there are very cool. However when the subject came up awhile ago due to some actual advances in the field some other people suggested that A: getting such cars to interact safely with human-driven cars would be difficult and B: humans wouldn't feel comfortable with such cars. (I forget if the argument was riding in them or driving on the road with them, or both.) Personally i think that the first one is true, but not so difficult compared to getting the things driving safely with each other as some people think, and the second is definitely true.

So wouldn't it be a good middle step to initially introduce them as an addition to public transportation, specifically on rail lines? You could add a bit of sidetrack at each station where the cars would pull over to drop off and pick up passengers. Theoretically you'd be able to grab a car as soon as you got to the station, if things were well organized that is, and it would then move back onto the main tracks and accelerate until it was trailing the next train down the line. (Or any other cars that were already trailing it of course.) Assuming it was successful enough you could then start eliminating trains over time until it was just the cars, which would let you speed things up since you wouldn't ever have to slow down or stop except at your destination. Obviously running them on tracks at first would make the programming a lot simpler and it would be a good way to get people used to them before they were more widely introduced.

Anyways, enough about that, i should go read up some more on the Propositions for tomorrow =P
jon_leonardjon_leonard on February 5th, 2008 07:28 am (UTC)
I think a more likely path is a sequence of driver-assist features. I think I've seen a reference to automatic braking on detected obstacles, and gradually installing such things could lead to public acceptance. The big obstacle right now is liability: If you can convince a jury that someone died because the automatic driver had a problem, then you can't afford to install them. But if they're shown to improve safety, then it could work.

But the software really isn't ready yet anyway.
DonAithnendonaithnen on February 5th, 2008 02:43 pm (UTC)
"But the software really isn't ready yet anyway."

Yes, however i think it _is_ ready to handle a rail version of the car.

Realistically though i expect the cost of adding the additional infrastructure (a hundred or so feet of rail per station plus a way for people to access that rail) is more of a roadblock than the AI itself. (To use an _almost_ appropriate metaphor =)
Johndolohov on February 5th, 2008 02:50 pm (UTC)
This is what the Urban Grand Challenge was all about, right? Getting a robot car to play nice with human drivers?

And the system described in your second paragraph already basically exists: the PRT at West Virginia University. You can't often grab a car immediately upon arriving, though, because the cars are designed for groups of a half dozen or so people at a time. It's not as efficient as it could be (having been designed in the 70s) but the basics are pretty much as you describe them.
Kirinkirinn on February 5th, 2008 05:51 pm (UTC)
I've had a similar thought which is "robot lanes", almost exactly analagous to current carpool lanes. You have cars that can be driven in either human mode or robot mode, and lanes on the right-hand side designated *only* for robot mode. With reliable software, the robot lanes could theoretically go significantly faster than the human lanes (better reaction time, smaller following distance, etc.) And you could give the navigation software an advantage by embedding easy-to-follow strips in these lanes too. The tricky part would be the routines for merging into and out of the robo-fast lane, though standardized inter-car communication could help tons ("hey, robo car A merging in", ok, a distributed algorithm determines the best way is for robo-car B to gradually slow down until there's space between it and robo-car C).

Anyway, once this was in place, I imagine it would help *encourage* people to give there cars over to robo-navigation on the freeway, because hey, fast lane.
Andrewneonelephant on February 5th, 2008 08:47 pm (UTC)
The argument for B) was definitely driving in them, specifically because many people are control freaks to some degree and/or don't entirely trust computers.

Taking the rider out of the "driver" mindset (by doing what you suggest) might help with the first one of those, at least.