THE ORIGIN-DESTINATION (OD) MATRIX GIVES THE VOLUME OF TRAFFIC FROM EACH OF A REGION'S ORIGINS TO EACH OF ITS DESTINATIONS AND IS A FUNDAMENTAL INPUT TO TRANSPORTATION PLANNING AND NETWORK DESIGN ACTIVITIES. BECAUSE THE TRADITIONAL METHODS OF ESTIMATING THE OD MATRIX--SURVEYS AND TRIP GENERATION/DISTRIBUTION MODELING--TEND TO BE EXPENSIVE, CUMBERSOME, AND INACCURATE, RESEARCHERS HAVE SOUGHT TO DEVELOPMETHODS FOR ESTIMATING THE OD MATRIX FROM OBSERVATIONS OF TRAFFIC VOLUMES ON THE REGION'S ROAD NETWORK. FOR SIMPLE LINEAR NETWORKS, SUCH AS SINGLE INTERSECTIONS OR FREEWAY SECTIONS, OD ESTIMATORS WITH DESIRABLE STATISTICAL PROPERTIES CAN BE DEVELOPED USING LEAST-SQUARES METHODS, BUT FOR GENERAL NETWORKS IT HAS NOT YET BEEN POSSIBLE TO PRODUCE CONSISTENT ESTIMATORS OF OD PARAMETERS USING TRAFFIC COUNT DATA ALONE. IT IS BELIEVED THAT THE LINK COUNTS ON A TRAFFIC NETWORK ARE GENERATED BY A STOCHASTIC PROCESS THAT IS PARAMETERIZED BY THE MEANS AND VARIANCES OF THE SEPARATE OD FLOWS. BY USING A TRACTABLE APPROXIMATION TO THE TRAFFIC-GENERATING PROCESS, IT IS POSSIBLE TO DEVELOP BOTH MAXIMUM LIKELIHOOD AND METHOD OF MOMENTS ESTIMATORS OF OD PARAMETERS, AND THE ESTIMATORS HAVE DESIRABLE CONSISTENCY AND ASYMPTOTIC NORMALITY PROPERTIES. SIMULATION STUDIES SUGGEST THAT THE MAXIMUMLIKELIHOOD ESTIMATOR, THOUGH EFFICIENT IN ITS USE OF DATA, IS COMPUTATIONALLY DEMANDING, WHEREAS THE METHOD OF MOMENTS ESTIMATOR IS NOTCOMPUTATIONALLY DEMANDING BUT IS STATISTICALLY INEFFICIENT. THIS PAPER APPEARS IN TRANSPORTATION RESEARCH RECORD NO. 1328, TRAVEL DEMAND FORECASTINg: NEW METHODOLOGIES AND TRAVEL BEHAVIOR RESEARCH 1991
Abstract