Critical enhancements of a dynamic traffic assignment model for highly congested, complex urban network

Abstract

To accurately replicate the highly congested traffic situation of a complex urban network, significant challenges are posed to current simulation-based dynamic traffic assignment (DTA) models. This thesis discusses these challenges and corresponding solutions with consideration of model accuracy and computational efficiency. DynaMITP, an off-line mesoscopic DTA model is enhanced. Model success is achieved by several critical enhancements aimed to better capture the traffic characteristics in urban networks. A Path-Size Logit route choice model is implemented to address the overlapping routes problem. The explicit representation of lane-groups accounts for traffic delays and queues at intersections. A modified treatment of acceptance capacity is required to deal with the large number of short links in the urban network. The network coding is revised to maintain enough loader access capacity in order to avoid artificial bottlenecks. In addition, the impacts of bicycles and pedestrians on automobile traffic is modeled by calibrating dynamic road segment capacities. The enhanced model is calibrated and applied to a case study network extracted from the city of Beijing, China. Data used in the calibration include sensor counts and floating car travel time. The improvements of the model performance are indicated by promising results from validation tests.