The study deals with traffic assignment analysis and O/D-matrix estimation based on traffic counts and other collateral information of importance. The developed traffic assignment model is based on the DSD method that utilises detailed route information for individual O/D-relations when determining the equilibrium solutions. The route information is used directly for O/D-estimation, but is also used for other purposes. Some examples are stochastic assignment models, "queuing" analysis, asymmetric assignment problems, cold start effects on links and location of traffic counting positions. O/D-matrix estimations are made with entropy models (with detailed and aggregated information) and with a least squares method. O/D-matrices have been simulated for one small test example, and for part of the Stockholm network respectively. - true and a priori O/D-matrices. All three models are capable of modifying the initial O/D-matrix towards the true ones and to reach the traffic count values. Reductions of the O/D-matrix errors are substantial, especially when the a priori matrices have biased means. The variances in the errors are also reduced in the same degree, or even better (in percentage terms). In principle, the errors in the traffic count values can be reduced to small values. The complexity of the problem increases with increasing congestion levels. (Author/publisher)
Abstract