A genetic algorithm based process for computing signal timing diagrams has been presented. The characteristic features of the process are: a compact diagram coding in order to reduce the free parameters; specific optimization functions either to maximize the green sum or to minimize delay times; a slightly modified reproduction (crossing over and mutation); and an additionalfine tuning process as an intermediate step between generations. The first version of the process has already turned out powerful enough to challenge the predecessor process ('blocked signalgroup algorithm'). The performance as well as the quality of the results are absolutely comparable. As far as offset times (Ob, Oe are involved the genetic algorithm even delivers superior results. Besides this the outstanding advantage of the new process is its flexibility with respect to any optimizing strategy. Two strategies (Qm and Qw have been introduced in this paper, but it is easy to define and implement further ones. Further work will concentrate on extending the capabilities of the process. For instance one could take into consideration given priorities of some flows or given position probabilities for green intervals. And, of course, finding new appropriate optimizing strategies might as well be an interesting future task.
Computing of signal timings with a genetic algorithm
1997
5 Seiten, 3 Bilder, 5 Quellen
Aufsatz (Konferenz)
Englisch
VisSim-Based Genetic Algorithm Optimization of Signal Timings
Online Contents | 2007
|VisSim-Based Genetic Algorithm Optimization of Signal Timings
Transportation Research Record | 2007
|Group-based optimisation of signal timings using parallel computing
Online Contents | 1997
|Optimizing Signal Timings from the Field
Transportation Research Record | 2009
|