An approach to multiclass mesoscopic simulation based on individual vehicles for dynamic network loading

Abstract

Dynamic network loading problem is crucial to perform dynamic traffic assignment. It must reproduce the network flow propagation taking into account the time and a variable traffic demand on each path of the network. In this paper, we consider the simulation-based approach for the dynamic network loading as the best suited option. We present a multiclass multilane dynamic network loading model based on a mesoscopic scheme that considers continuous-time link-based approach with a complete demand discretization. A well-known classification of the dynamic network loading models based on simulation represents models in a 3D space with time, space, and demand axis. Based on that, we propose a new representation scheme which serves as a base of a more detailed classification. We show how our model is displayed in this new classification. Considering disaggregated treatment of each individual vehicle allows to use different vehicles classes in the problem. Moreover, our aim is to reproduce transversal movements described by vehicles changing lanes which can considerably augment the link congestion. Therefore the proposed model allows longitudinal discretization of links in lanes. We computationally tested it on the network of Amara (Spain), and compared the results with those obtained from a microsimulator. The obtained results look promising, showing a good quality in the proposed model. Furthermore, the results show model's ability to reproduce multilane multiclass traffic behaviors for medium-size urban networks.