Estimation of turning movement proportions from partial sets of traffic counts at intersections

Estimated turning movement proportions are used in a number of traffic simulation and traffic control procedures in order to predict the turning movement flows at intersections. Historically, these proportions have been estimated by manual counting, but the ongoing deployment of real-time, adaptive traffic control strategies indicates that automatic estimation of these proportions from traffic detector data is becoming increasingly important. When it is possible to count the vehicles both entering and exiting at each of an intersection's approaches, methods based on ordinary least squares can produce usable estimates of the turning movement proportions, but when the number and/or placement of the detectors does not support complete counting, these methods fail. This paper has two objectives, the first being to assess the feasibility of estimating turning movement proportions from less than complete sets of traffic counts, and the second being to compare the statistical properties of less than complete count estimates to those estimates generated from complete counts.