Authors:
Ali Hajbabaie
and
Rahim F. Benekohal
Affiliation:
University of Illinois at Urbana Champaign, United States
Keyword(s):
Traffic Signal Optimization, Oversaturated Network, Evolution Strategies, Genetic Algorithms.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Computational Intelligence
;
Evolutionary Computing
;
Evolutionary Graphics
;
Genetic Algorithms
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Soft Computing
Abstract:
This paper compares the performance of Evolution Strategies (ES) with simple Genetic Algorithms (GAs) in finding optimal or near optimal signal timing in a small network of oversaturated intersections with turning movements. The challenge is to find the green times and the offsets in all intersections so that total vehicle-mile of the network is maximized. By incorporating ES or GA with the micro-simulation package, CORSIM, we have been able to find the near optimal signal timing for the above-mentioned network. The results of this study showed that both algorithms were able to find the near optimal signal timing in the network. For all populations tested in this study, GA yielded higher fitness values than ES. GA with a population size of 300, and selection pressure of 10% produced the highest fitness values. In GA for medium and large size populations, a lower selection pressure produced better results while for small size population a large selection pressure resulted in better fi
tness values. In ES for small size population, larger µ/λ yielded better results, for medium size population both µ/λ ratios produced similar results, and for large size population smaller µ/λ provided better results.
(More)