From: Dan Clemmensen (dgc@cox.rr.com)
Date: Sat Feb 09 2002 - 17:11:26 MST
Anders Sandberg wrote:
> Just a thought (to get back to extropian matters): suppose drivers could
> affect lights through some technological means. Would there be any
> system that would make the traffic throughput more efficient than the
> current timer-based lights?
>
> Suppose each car could "vote" for the light (similarly for the
> pedestrians at crossings). At a crossing this would likely mean that the
> biggest queue would get green first, and decrease until the votes of the
> other queue weighed over. With a bit of inertia in the system this would
> likely ensure a fair division. There is likely a bit of need for
> weighing votes by the time the light has been red too, in order to ensue
> that cars crossing a very busy street along a far less busy street ever
> will get past. At night, when there are few cars at all, all lights
> would turn green since there are just single votes.
>
> However, traffic flow theory is filled with surprises and paradoxes, so
> I wouldn't be surprised if there are conditions where this kind of
> solution wouldn't be optimal. Any thoughts?
>
>
This turns out go be a non-trivial problem. Sure, each traffic light
could attempt to maximize its "profit" locally, but if their agents are
intelligent, they will quickly realize that they can make a lot more
money by cooperating with each other or with "traffic arbitrage agents."
A customer would contract with the traffic arbitrage agent
rather than attempting to contract with individual traffic lights.
The arbitrage agents would quickly realize that they can make higher
profits by negotiating among themselves to build packages of cars that
share routes, If the agents are all smart enough the result will
approach a global optimization for the metropolitan area as a whole.
Like any routing problem, this is only useful if the cost of the more
sophisticated routing algorithms are cheaper than the cost of additional
bandwidth (streets and highways in this case.) My area of expertise is
routing in data networks. In data networks, even the crudest algorithms
work perfectly well when the traffic is at 50% of the available
bandwidth, And even the best theoretical algorithm cannot achieve 95%
utilization with acceptable performance. In most data networks it
is cheaper to build extra capacity than it is to deploy better routing
algorithms. The reason the Internet works at all is that its core is
massively underutilized.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:38 MST