ASSIGNMENT MODEL FOR TIMETABLE BASED SYSTEMS (TPSCHEDULE)

Existing assignment models for public transport primarily use frequency based networks. It is well known in theory and from practice that such deterministic models often have major shortcomings (frequency aggregation corresponding lines etc.). To deal with these shortcomings TetraPlan has developed a new timetable based assignment model. The model allows the user to model differences in travellers' preferences by using distributed weights in the utility function. It is also possible to model travellers' imperfect knowledge to timetable by adding a stochastic error-term to all perceived costs in the network. Finally, departure time may be simulated. The route choice is calculated using an iterative probit-based framework. Thus, problems with overlapping routes can be avoided. The algorithm is implemented using a dual network, where all interchanges are modelled as arcs in the network. This representation gives great flexibility in assigning costs to the network while retaining a linear structure thus making it possible to apply standard shortest route algorithms to the network. The model has been used on a network consisting of approx. 250000 nodes and 1600000 arcs. Calibration has shown a better description of observed route choice than models based on the frequency aggregation principle. Furthermore travel times are calculated with much more detail and accuracy than can be done in a frequency based network. The paper describes the method and results in more detail. For the covering abstract see ITRD E105584.