The demand for travel continues to increase in European cities of today, which results in long car travel times and highly congested road networks, especially during the morning and afternoon peak periods. The congested car system gives rise to high emissions of particles and greenhouse-gases, which is negative for both the local and global environment. Congestion also causes an uncertain travel time, something that in the latest years has been recognised as a major factor in car-users perception of trip disutility. Congestion is time-dependent in nature. Therefore, not only the spatial distribution of trips over the network is important in analysis and prediction, but also the temporal distribution of trips. A traditional congestion-relieving strategy such as a capacity expansion often has an impact on when people travel, since shorter travel times during the peak hour can attract traffic from the peak shoulders. The temporal effects are even more pronounced for the new policy measures gaining ground today, e.g. variable road pricing. Most variable road pricing systems aim at moving traffic from the peak hour to the peak shoulders, so called peak spreading. The means by which this is done is by charge differentiation: it is most expensive to travel at the most congested point in time. However, most large-scale transport planning models in use today are static and changes in the temporal distribution of trips are not considered. It is therefore likely that false conclusions are drawn when using these models to evaluate the ability of different pricing schemes or infrastructure investments to alleviate congestion.SILVESTER – A Dynamic Transport ModelTo better model the temporal distribution of traffic has been the basis in the development of SILVESTER (SImuLation of choice betWEen Sarting TimEs and Routes), which is a dynamic transport model for the Stockholm area. In SILVESTER road network conditions during the extended morning peak period (06:30-09:30) are modelled. The morning is divided into twelve 15-minute time intervals and a departure time choice model allocates trips to each interval depending on their attractiveness. The attractiveness of a time interval is determined by its corresponding travel time, travel time uncertainty, monetary cost and how close it is to the preferred time interval (PDT) of the traveller. The travellers can also choose to start before 06:30 or after 09:30. Mode choice is partially modelled by introducing the possibility for car-users to switch to public transport if it is perceived as a better option than any of the time intervals. It is distinguished between three trip purposes: business trips, trips with fixed schedule and school trips, and trips with flexible schedule and other trips. In SILVESTER iteration towards a general equilibrium between supply and demand is performed. The supply quantities (travel times, uncertainties etc.) are calculated with the mesoscopic dynamic traffic assignment model CONTRAM, whereas the demand for each time interval and public transport alternative is calculated with a mixed logit discrete choice model. A calibrated origin-destination-matrix (OD-matrix) for the Stockholm CONTRAM network exists and is based mainly on traffic counts but also on travel times for some selected OD-pairs. Calibration of Preferred Departure TimesEven though many trip-timing models use the concept of schedule delay, which is defined as the deviation from a preferred departure/arrival time, little work has been done on how to find the PDT-distribution when applying the model. For estimation the survey respondents can be asked to state their preferred time of travel, but for large-scale applications similar studies are expensive and time consuming. Previous work has often assumed a simplified distribution, such as all travellers in a market segment having the same PDT. Without calibration of PDT’s, e.g. using only a simplified exogenous assumption, the predictive capability of the transport model is questionable. Instead of making an exogenous assumption about the PDT-distribution this paper uses a reverse engineering approach to reveal PDT’s from the observed departure times of the reference situation. It is the combination of the estimated departure time choice model, the travel conditions and the observed departure times that can be used to get information about the PDT’s. Once the PDT’s have been calibrated for the reference situation they can be used in the evaluation of a congestion relieving strategy. This paper will present calibration methodology, obstacles overcome on the way and calibration results. It will also discuss future work in which the calibrated departure time choice model will be used to improve the design (charge levels and time periods) of a pricing scheme.
European Transport Conference 2008, 6-8 October 2008, Leeuwenhorst Conference Centre/ The Netherlands
QC 20101015.Tidigare titel: "Deriving Preferred Departure Times of Road-Users in a Real-Life Network".