A mesoscopic simulation model for dynamic network loading and spillback queuing assessment in a multiclass environment

Traffic flow modelling is the most significant component undertaken by the static and the dynamic network loading (DNL) models in the traffic assignment. Dynamic network loading models represent a non-linear relationship between each link flow and its path flow, as they are the fundamental element in estimating the dynamic interaction between demand and supply in oversaturation condition. Moreover, the solution for dynamic network loading problems is necessary for generating the dynamic traffic assignment (DTA) models. Dynamic models can be characterized according to the simulation details level: microscopic, macroscopic or mesoscopic models. Accordingly, microscopic simulation models fit well in small-scale planning determinations with interest addressed to entities’ interactions as these models describe the interaction between vehicles, and between vehicle themselves and transportation infrastructure. Instead, macroscopic models are capable of the general planning purposes adopting large-scale simulations. As they assume the traffic as a continuous fluid and the flow is subject to the congruency and to the continuity constraints. Finally, the mesoscopic approach simulates most of the entities at a high level, but activities and interactions at a low level of details. In this context, for a reasonable level of details, coupled with entities interaction information at once, mesoscopic models simulate each link considering the traffic as a set of continuous or discrete packets: a continuous packet is defined by its head and tail, conversely to the discrete packet which is defined by its head, regardless of the tail position. Many different aspects can be included within the dynamic network loading models such as the multiclass property. It includes the vehicular type in the mesoscopic simulation, which generates different dynamics on the same link considering more than one vehicle type at the same time. With this complication, once the supply becomes unable to meet the demand (oversaturation condition), evaluation of the queuing spillback is necessary to prevent excessive delays and to forecast the new trip travel time. For this aim, this thesis proposes a new dynamic network loading model which simulates traffic dynamics (speed, density, flow, queue, etc.) explicitly, through modelling the traffic flow considering a discrete mesoscopic simulation model in a multiclass environment. The proposed model is capable of using two speed-density relations to simulate flow dynamics: the Greenshields and the triangular-shaped fundamental diagram. FIFO rule holds between the vehicles in the same class and creeping speed is assumed to avoid circulation blockage in oversaturation conditions. Moreover, three vehicle classes (private car, bus, truck) have been considered in the simulation. The proposed model has been validated in undersaturation conditions by comparing model estimations with real observations collected by ATC sensors for Maliha Highway in the United Arab Emirates. For assessing the dynamic queue spillback, the proposed model has been applied to a simple network for easily assessing its capabilities in oversaturation conditions. Moreover, a comparison with a commercial traffic simulation software, Aimsun Next, has been carried out to evaluate the performance of the proposed model. The comparison has shown a relatively similar behaviour and simulation time for all classes in the case of using the triangular fundamental diagram relation but with much more fluctuation for the Aimsun model. On the contrary, using Greenshields relations provided the same behaviour but with much longer simulation time. As a result, the proposed model has presented the mesoscopic simulation in a more reliable way since Aimsun seems to include other microscopic characteristics in the mesoscopic traffic simulation like start-and-stop behaviour. Finally, it can be used with confidence as a tool to quantify the traffic dynamics of each class in oversaturation conditions including queue spillback. Keywords—congestion, flow propagation, Greenshields model, queue spillback, multiclass mesoscopic simulation.