Use of Genetic Algorithm for Phase Optimization at Intersections with Minimization of Vehicle and Pedestrian Delays

Abstract

The two objectives of this research were to develop a methodology to optimize signal timing at intersections by minimization of delay for both vehicles and pedestrians and to establish guidance for the selection of the pedestrian crossing phase (two-way or scramble) and the length of the “Walk” phase when the scramble crossing pattern was used. An optimization procedure for signal plans was developed for an isolated intersection. The procedure can provide up to four vehicular phases with either the two-way or scramble pedestrian crossing phase. Genetic algorithms were used to find suitable signal plans because of the existence of a large solution set. Furthermore, the proposed genetic algorithm procedure can generate contour diagrams as selection guides to determine appropriate pedestrian crossing phases at an intersection with different combinations of vehicle volumes and pedestrian volumes. The impacts of relative time values, initial vehicle queues, and geometric designs on the pedestrian phase selection were also determined. According to the contour diagrams, to determine appropriate pedestrian crossing phases, not only do pedestrian volumes and right-turning vehicle volumes need to be taken into account but also through (and left-turning) vehicle volumes should be considered. The scramble crossing reduced delay when the conflicting pedestrian and right-turning vehicle volumes in approaches were high and the through (and left-turning) vehicle volumes in the same approaches were relatively lower. Otherwise, the scramble phase increased delay.