Assignment Model for Timetable-based Systems
MOLLER-PEDERSEN J, Tetraplan A/S, Denmark
Until few years ago assignment models for public transport used frequency based networks in practical applications. It is well known in theory and from practice that such models often have major shortcomings (frequency aggregation, corresponding lines etc
Until few years ago assignment models for public transport used frequency based networks in practical applications. It is well known in theory and from practice that such models often have major shortcomings (frequency aggregation, corresponding lines etc.). By using the whole time-table it is possible to describe travel behavior more accurately as it is e.g. possible to use exact interchange times, to model. hidden waiting time consistently and apply capacity restraints within each vehicle. Furthermore it is possible to distribute fares to every vehicle making it possible to predict economic feasibility of new lines and/or new fare structures.
To exploit these advantages TetraPlan has developed a public transport assignment model called TPSchedule, which is based on a network representing an entire time-table and estimated for use in fvll-case studies. Timetable-based assignment models exists in commercially available traffic planning software e.g. Emme/2 and VISUM. These models are deterministic in nature whereas TPSchedule is a stochastic model making it possible to model differences in passenger preferences and incomplete knowledge of the complete timetable. Furthermore, TPSchedule contains a Windows-based, user-fiendly interface for handling public transport networks.
TetraPlan has worked with timetable based assignment for several years. In 1997 TetraPlan presented a timetable-based model, which included modeling of hidden waiting time, risk of delay and distributions of time-of-departure. The model was refined by introducing capacity restraints using an iterative framework. Recently the model has been used to evaluate urban public transport. The inclusion of urban traffic enhanced the problems with overlapping routes ,and differences in preferences, and therefore the model was enhanced to accept error components and distributed coefficients in the utility function following the principles for car traffic assignment presented by Nielsen (1996).
The paper concentrates on the recent developments. The methodological concepts are described in section 2, while section 3 shows how distributed utility functions are integrated using a probit-based framework and sampling techniques. Section 4 describes how various costs are implemented in the model. In section 5 the practical applications of the assignment model are illustrated while section 6 presents the conclusions of the paper.
Association for European Transport