Static traffic assignment models are an important tool for long term traffic planning. A central part of these models is the underlying travel time functions, often called volume delay (V/D) functions. The travel time functions used until 2014 in Trafikverket’s (Swedish Transport Administration) transport modelling system Sampers have been criticized to be out of date and inaccurate in terms of the relation between travel time and travel demand. During 2008—2010 Trafikverket conducted a general review of the Swedish speed limit system. The review resulted in that the Swedish road network now also include roads with speed limit 40, 60, 80, 100 and 120 km/h. The road network has also changes in terms of that oncoming separated roads have become more common. Thus, there was a need to develop travel time functions for the new speed limits and for oncoming separated roads. Trafikverket developed a new description of the Swedish road network in Sampers based on the National Road database NVDB. The purpose was to ensure consistency between Sampers and the effect size calculation methods that Trafikverket use in their other tools, e.g. EVA and Samkalk. As a first step, travel time functions were created based on Trafikverket’s speed-flow relationships (VQ-relationships). However, these VQ-relationships describes only the travel time and delay on road links and do not include delays at intersections. The aim of this project was to apply the methodology from VTI rapport 571 in order to produce a complementary set of intersection delay functions. Development of travel time functions are conducted in several steps. It is neither practical nor doable to calibrate a unique travel time function for each physical link in the road network. The road network need to be divided into road classes. All links within a road class are assumed to have the same relation between travel time and travel demand. The next step is to choose a mathematical function with suitable parameters that can be used to describe the travel time as a function of demand. In this work we used a mathematical function that include separate terms for the free flow travel time, the link delay and the intersection delay. The free flow travel time and the link delay is based on Trafikverket’s piecewise linear VQ-relationships while Akcelik’s intersection delay function is used for the intersection delay part of the travel time function. The intersection delay function includes parameters that need to be calibrated individually for each road class. The calibration requires synchronized data on intersection delay and travel demand, which is both difficult and expensive to collect. An alternative approach was used instead, in which the intersection delay was estimated based on assumptions on intersection design and flow relationships. Based on these assumptions and calculations using the intersection delay model Capcal a set of intersection delay functions were constructed. The travel time functions developed were implemented using a three step process. First the functions we applied in Sampers calculations for the Samm region (Stockholm and Mälardalen). The result from these calculations indicated that adjustments of the travel time functions and their parameter values were needed. The travel time functions and their parameters were revised and then a new set of Sampers assignments were conducted using the revised travel time functions. The validation shows that the travel time elasticity with respect to travel demand is closer to the observed elasticities (in average) when using the new travel time functions. However, there is not a perfect match. The results of the project are satisfying in a short term perspective but a stronger empirical base is desirable. New travel time functions need to be developed both for urban and rural environments. This development also need to investigate which further factors that may influence the route choice, e.g. route guidance, comfort, speed cameras, incidents, etc. (Author/publisher)
Samenvatting