Passengersâ€™ Density Impact on Route Choice in Urban Public Transport Networks
Vincent LEBLOND, RATP, Cécile Prat, RATP
Passengersâ€™ density impact in public transport route choice is commonly introduced in transport model with crowding curves. Our work aims to estimate a crowding function based on a large data collection without local network constraints.
Since the beginning of transport modelling in the 1970s, studies requirements have evolved leading to more accurate traffic forecasts with high level of details: simulation of operating schemes, transfer and platform access time optimization, transfer trips extraction on complex multimodal nodes, specific use of stations... Regarding these new requirements, the assignment model of public transportation demand has become the main focus of improvement. Indeed, the route choice model largely determines the quality of traffic forecasts, and therefore the estimation of the impact of future projects or tested scenarios. However, modelling the route choice is a complex problem, given the lack of information about passengersâ€™ route selection criteria. That is why most models are calibrated on volumes and not on routes with standard time penalties.
In 2013, our modelling team began a renewal of the route choice model. It has shown that the EGT household survey can be used for route choice model estimation. This can be done if a model is able to enumerate feasible route alternatives. The final estimated model gave an efficient estimation of usual coefficients: in-vehicle and waiting time, access and egress time, transfer time and number of transfers. It was then further improved with the addition of specific penalties to busses, in access or egress, and in Paris.
Recently the model has been enhanced with another source of observed individual routes. These data were collected directly on railway platforms. This survey offers a large amount of data on subway lines and urban railways. However results regarding bus lines are not representative enough to be considered. Utility functions are now improved with variables specific to each line that quantify the attractiveness of each of them.
However we all know that route attractiveness depends on the perceived comfort. This explanatory variable is commonly introduced with a crowding curve. Stated Preference (SP) surveys are usually considered for the estimation of parameters. More rarely, Revealed Preference (RV) surveys are conducted in specific network situation. However these two methods are relatively distorted: on the one hand SP surveys over-estimate individual preferences, on the other hand RV surveys are limited by the local context of the network (eg: fork).
Our work aims to estimate a crowding function on a large data collection without local network constraints. In practice, a sample of 30 000 observations is used as it covers the entire RATP network. For each feasible route of each observation, observed in-vehicle volumes of passengers per each section and trains capacity have been joined to our database. We have then developed a specific tool in order to estimate the perceived travel time on each railway section due to passengersâ€™ density.
This work is conducted within the overall understanding of passengersâ€™ density impact on peopleâ€™s everyday activities: mode choice, departure / arrival times and route choice in public transport. We expect a better understanding of new projectsâ€™ impact in a context of significant growth of public transport demand and highly congested network.
Association for European Transport