OPTIMIZATION OF CONSECUTIVE SIGNALIZED INTERSECTIONS BASED ON COMBINED ALGORITHMS – COMPARING RESULTS WITH MICROSIMULATION

Abstract

The primary objective of this research is to optimize signal timing in consecutive signalized intersections. In this paper, the combination of genetic programming (GP) with genetic algorithms (GA) and neural network (NN) with genetic algorithm (GA) were used and compared in order to optimize signal timing in consecutive signalized intersections. First, genetic programming and neural network were constructed from existing signal timing data to predict the delay of intersections. Then genetic algorithm was applied to optimize these predictive networks (GP and NN). The results and comparisons of timing process and error percentage showed that neural network is more efficient than genetic programming. However, the ability of genetic programming in producing formula is a specific characteristic which makes it more applicable than neural network. Finally, for validating the results, Aimsun and Synchro micro simulation software were used, and accuracy of our models was approved.