Global Speed vs. Mean Travel Time. Polynomial Dependence Analysis in Traffic Signals Optimization Using Genetic Algorithms and Parallel Computing

In our group we have developed a traffic lights programming optimization model based on the combination of Genetic Algorithms and Microsimulation running over a Beowulf Cluster parallel computer. So far, in this architecture we have used a single variable for the fitness function. In this research our aim is to explore any polynomial dependence – up to a 2 degree – between two candidate variables as potential participants in the fitness function: Global Speed and the Mean Travel Time. All tests have been fulfilled using data from a real world scenario located in Saragossa, Spain. We have used the supplied traffic lights provided, and also traffic statistics from the zone.