An Alternative Approach to Centroids and Connectors Pattern: Random Intra-Zonal Travel Time

In traditional traffic assignment procedure, each traffic analysis zone is represented by one point in its geometric center which is connected to the network by several connectors. Results of studies show that different connector patterns would result up to 10% change in estimated volume and up to 20% change in total travel time. Also the different patterns of connectors can change the priority of network development projects. Because of numerous possible scenarios for connection patterns, finding the optimum pattern is very difficult. The main goal of this research is proposing a substitute approach and proper solution algorithms to increase the accuracy of traffic assignment results. In this approach, the travel time inside the zone is considered as a random variable and is used in traffic assignment process. At first, all possible candidate nodes in each zone are selected and then between each pair of these nodes in different zones, the shortest path is determined. After that, the share of each route is estimated using discrete choice models and then share of them from demand is calculated. The proposed approach is tested using a real-size network with two route choice models of Logit and Probit. The results show significant changes in total network indexes (15%) and the average error in estimating of links volumes, has been reduced by 20%. Although the calculation time may increase 10 times than before, the proposed approach leads the traffic assignment results more accurate and close to real observed links volume.