Estimación de la demanda de transporte público en área metropolitana a partir de los registros de entrada de usuarios

Laburpena

This thesis deals with some of the most common problems that arise from the use of the data from the operations support system of a public transport operator, especially when combining information from different subsystems. It presents a flexible methodology to improve the definitions of bus runs and boarding events, ameliorating the distortions that stem from the simultaneous use of data from the automated vehicle location, automated fare control, and scheduling subsystems. Then, for the quite common case of a system that only registers the access of the passengers to the buses, the trips they carried out and which ride did they choose to materialize them are inferred. To this end, a trip chaining model is defined, in- corporating several enhancements from the available literature, and making use of the previously obtained enhanced accounts of passengers’ validations and vehicles movements. Firstly, the definition of each distinct vehicle run and boarding event carried out in the transport system is improved, integrating the information provided by stop-level events from its automated vehicle location and fare collection systems, and scheduling subsystem information at the initial stop of planned runs, if avail- able. The data are structured; and then corrected and completed utilizing several criteria, which include identifying and combining all entries that are linked to the same call of a bus to a stop, and applying a probabilistic approach based on the distributions of travel and dwell times to both event filtering and the reconstruc- tion of incompletely or wrongly registered runs, following a procedure akin to dead reckoning, utilizing as the initial fix the longest sequence of calls compatible with the configuration of the route the vehicle should be following. In turn the origin and destination stops of the trips performed by transit users are deduced, as well as which runs offered by the public transport system they rode to do so. The trip chaining model implemented to do so benefits from the improved definitions of bus runs and boarding events and includes several contributions from the state of the art to better identify the end of each ride, find the last destination of the last ride of each day, and detect short activities that may be incorrectly be labeled as transfers. Finally, the passenger trips thus obtained are aggregated and expanded to provide origin-destination matrices for different periods of the year and days of the week, considering the mobility patterns in the city. A case study is discussed with one year of records from the automated vehicle location, fare collection, and scheduling subsystems in Santander City, Spain, em- ploying captures from different interactive web visualization tools that has been developed for this work. This document has been written using LATEX. Its figures were created with TikZ, PGFPlots, Gimp, Inkscape, QGIS, and several tools developed for this work, based on Bokeh (a Python library). Part of the work of this research has been submitted as the article ‘Integration of automated vehicle location, fare control, and schedule data for improved public transport trip definition’ (Juan Benavente, Borja Alonso, Andr ́es Rodr ́ıguez and Jos ́e Luis Moura) to a noted journal and is currently under review.