An aspect of road traffic safety that is less intensely researched is the rerouting or network effect. When measures are taken to improve traffic safety, usually the effect is measured at the location where the measure was activated. However, because most measures also affect the throughput and hence the attractivity of the route, measures might in addition cause rerouting effects either away from or towards the location where the safety measure was taken. While doing so, also safety effects can be exported to or imported from alternative roads. Dynamic Traffic Management (DTM) measures can be aimed at improving safety, efficiency or both. Either way, they will have cause rerouting, intended or as a side-impact. It is thus important to ex-ante evaluate the impact of DTM on the network level. Given that DTM is usually aimed at mitigating congestion effects, an adequate modeling of the dynamics of traffic in networks is crucial in such evaluation. However, existing modeling instruments do not always comply with this requirement. Static models lack the time dimension, whereas microsimulators are impractical (if not unable) for modeling larger networks. That is why the use of macroscopic dynamic traffic assignment (DTA) models is being explored. There are several components affecting the outcome of dynamic network evaluations such as dynamic traffic demand known as dynamic origin-destination, Dynamic Network Loading (DNL) which propagates traffic demand given network characteristics, resulting among others in route travel times and Route Choice (RC) component that uses the route travel times to calculate route choice fractions. The combination of last two components known as Dynamic Traffic Assignment (DTA) that is consistent with both DNL and RC and obtained by an iterative process between these two components. The inability to provide high quality in one of these components makes the prediction simply impossible. Therefore, it is essential to measure and in-depth analysis the capability of DTA modeling software that can adequately describe traffic dynamics and behavioral processes in a network. It is vital to have an accurate DTA model that would realistically describe network flow distributions, to capture the effects of congestion as well as route choices. To achieve this, a benchmarking of the latest DTA modeling software of PTV AG and Sistema called TRE (Traffic Real-time Equilibrium) is conducted (in a complementary report, the StreamLine software of DAT.mobility is benchmarked using the same methodology). The main focus of this benchmarking was done for Dynamic Network Loading and Route choice components. Since dynamic traffic demand is an exogenous input for DTA, this is not relevant for this study. In this benchmarking, several test cases have been defined, where DNL and RC components were scattered into several sub problem and different aspects of the DNL and RC have been analyzed. In the first two cases, we explore the propagation model under different conditions. These test series brought depth analysis the functionality of propagation model and influences of the demand and supply interaction with flow propagation. Delay is the most important measure of effectiveness (MOE) at a signalized intersection because it relates to the amount of lost travel time, fuel consumption, and the frustration and discomfort of drivers. To understand it, modeling of the traffic flow at signalized intersections is explored and analyzed under different demand and supply conditions. This test enables us to understand how signalizations are taken into account in the signalized intersection also the calculation and definition of the delay in non-congested condition or when the queue spillovers the intersection. Modeling route choice behavior is essential to appraise traveler’s perceptions of route characteristics, to forecast traveler’s behavior under hypothetical scenarios, to predict future traffic conditions on transportation networks and to understand traveler’s reaction and adaptation to sources of information. Therefore, route choice component is analyzed where stochastic and deterministic models are activated under different scenarios. The sensitivity of the RC model to the overlapping of paths and functionality of the model under gridlock condition are also explored. Beside these technical analyzes, robustness, consistency, accountability and ease of use of TRE within each test case are explored. The analyses show that, although the software can be improved in several aspects, it is sufficiently mature and evolves sufficiently quickly to recommend configuring practical networks in it. (Author/publisher)
Abstract