Author(s)

N. T. Ratrout

Abstract

Many countries are still using pre-timed signal controllers. Although this type of controller is technically inferior compared to actuated and computerized controllers, they provide an acceptable level of performance, especially for cyclic traffic. They are also cheaper and reliable since they do not depend on traffic sensors. The efficient use of pre-timed controllers depends primarily on selecting the optimum number of signal timing plans within a day and the starting and ending time of each plan which is commonly known as Time-of-Day (TOD) breakpoints. An optimal timing plan is then developed for each time period using optimization software such as Synchro. When the traffic pattern is cyclic and repeats itself from one day to another, these plans can be used for several days. This paper suggests a procedural framework for selecting the optimum number of timing plans and TOD breakpoints through the use of K-means statistical clustering and traffic simulations. This study also intends to reduce \“outliers” of clusters. Outlier means that the statistical clustering technique proposes a short cluster interval between two relatively long intervals. K-means clustering technique based approach to find out optimal TOD breakpoints usually suffers from \“outliers” of clusters. In this proposed procedure, Z-score of approach volume, in lieu of raw approach volume, is used simultaneously with time variable to solve the outlier problem. The recommended procedure was tested on a major arterial in Al-Khobar city, Saudi Arabia. It eliminated the outlier problem and provided a systematic and reasonably efficient procedure for timing pre-timed signal controllers. Keywords: optimal signal plans, K-means clustering, pre-timed controllers, TOD breakpoints, cyclic traffic.

Keywords

optimal signal plans, K-means clustering, pre-timed controllers,TOD breakpoints, cyclic traffic