Davydov I.A.   Tolstykh D.   Legkih I.   Kononova P.  

Genetic based approach for Novosibirsk traffic light scheduling

Reporter: Davydov I.A.

Congestion, derived from the permanent increase in road traffic, is a pressing problem in the big cities all around the world nowadays.
Thus, the methods of the intelligent control of vehicles traffic answer the growing demand. Optimization of traffic signal plans is an important step in this direction. Well-tuned traffic lights schedule augments the efficiency of vehicles flows processing. The research in intelligent traffic signal control helps to significantly improve a traffic situation, to reduce the average vehicles waiting time and to increase the average speed in the network.

In this study, we consider different configurations of genetic algorithm with the intention of finding an effective traffic lights schedule on the real road network. This heuristic evolutionary algorithm copes well with such kind of optimization problems.
For application of our research, we chose a complex segment of the street network in Novosibirsk, big Russian city. Using a microscopic traffic simulator, we modeled a corresponding road map fragment. Obtained model served to evaluate solutions of traffic scheduling problem. We analysed genetic algorithm work with different parameters and obtained improvements in terms of three different objectives reflecting traffic congestion.

To reports list