It's not the
ultimate solution to traffic problems, but it's a step in the right direction: the forthcoming issue of IEEE Transactions on Evolutionary Computation includes
a paper on regulation of traffic lights in the main city of the island of Tenerife, Santa Cruz, by three Spanish university researchers. Here's the abstract:
In previous research, we have designed and successfully tested a Traffic Light Cycles Evolutionary Optimization Architecture. In this paper, we attempt to validate those results with a real-world test case. For a wide area in Santa Cruz de Tenerife city—Canary Islands—we have improved traffic behavior, using our optimized traffic light cycle times in a simulated environment. Throughout this paper, we present some of the experiences, knowledge, and problems encountered.
It's interesting to see evolutionary computation used in a real life problem, and in my own country. Congratulations to the researchers. I haven't had time to read it yet, but it will be a nice topic for one of the next paper seminars.
Update: I have been trying to read this paper and been utterly unable to undestand it. For instance, after reading the part on chromosome encoding (III.A.2) several times I still don't understand how it works. And that's the foundation of the paper.
I also don't agree that Gray coding makes any sense. It does if you use a "circular" mutation, changing only from a gene to the next one in a row, which they don't (part 6 of the same subsection). If normal bitflip is used, there's no difference between Gray code and normal binary codification. Which means the phrase
when mutation occurs the stage length [the number it codes] only increases or decreases one unit
is simply not true. For instance, take 011 (2 in Gray code), and flip the first bit. It becomes 111, which is 6 in Gray code. Stage length changed by 4.
Etiquetas: genetic algorithms, evolutionary computation, traffic
