260 likes | 365 Vues
This study analyzes how transit service frequency and station characteristics affect passenger arrival time distributions using smart card data. Conducted in the Greater Copenhagen Area, the research aims to understand the implications of passenger waiting times influenced by both random and non-random arrivals. The methodology incorporates validation against manual observations and examines the influence of various station attributes. Key findings reveal relationships between frequency, waiting times, and passenger behaviors, offering insights for improving public transport scheduling and real-time information accessibility.
E N D
The influence of transit service frequency and station characteristics on passenger arrival time distributions:A smart card data analysis in the Greater Copenhagen Area Trafikdage, Aalborg – 28 August 2017 Jesper Bláfoss Ingvardson1, Otto Anker Nielsen1, Sebastián Raveau2, Bo Friis Nielsen1 1 Technical University of Denmark 2PontificiaUniversidadCatólica de Chile
Agenda • Motivation & Research objective • Methodologicalframework • Case study & Data • Data cleaning & Validation • Results • Conclusions
Motivation and research objective • Motivation • Passenger waiting times important due to highervalue of time • Increasedavailability of real-time information and on-demand public travel planners encouragepassengers to time their arrival at stations • Traditionalpublishedtimetables vs. frequency-basedtimetable • Research objective • Developmethodology for analysingpassenger arrivals thatexplicitlytakeintoaccountpassengersarrivingrandomly and non-randomly • Analyse the influence of station characteristics and amenities on passenger arrival patterns
Methodologicalframework (I) • Two types of passengers:
Methodologicalframework (II) • Random arrivals • Thosearrivingrandomly, e.g. without knowing the timetable • Adoptedtraditionalapproach to model arrivals as uniformlyrandom
Methodologicalframework (III) • Non-random arrivals • Those timing their arrival according to the timetable • Adda buffer to not miss the departure • Modelled as Beta-distribution • Bounded on interval • Can handle passengers’ access time buffers • General form of the uniform distribution
Case study & data • Smart card data (Rejsekort) • 100 million public transport trips annually (Rejsekort A/S, 2017) • ~1 million used (Sep-Oct 2014, onlyfirst trip leg on weekdays) • Tap-in-tap-out on station platforms (buses excluded) • Tap-in at arrival (?) • Sample bias (fewcommuters and students) • Validatedagainst manual observations of passenger arrivals • Timetable data for trains (Suburban and regional) • Synthetictimetable for metro (nopublishedtimetable)
Data cleaningconsiderations … • Important to takeintoaccountrealisedtimetable
Data cleaningconsiderations … • Important to takeintoaccountrealisedtimetable
Data cleaningconsiderations … • Important to takeintoaccount S-traindwell times
Validation • Comparison of Rejsekort data and manually collected arrival data at Bernstorffsvej station on August 11, 2016
Results • Frequency-based vs publishedtimetable
Results • Frequency-based vs publishedtimetable
Results – Mixture distributions • Headway time: 5 minutes
Results – Mixture distributions • Headway time: 10 minutes
Results – Mixture distributions • Headway time: 20 minutes
Results – Mixture distributions • Headway time: 30 minutes
Results – Mixture distributions • Headway time: 60 minutes
Results – Overview • The lower the frequency the more timed arrivals
Results – Station characteristics • Station layout important for passenger waiting times • Difference between Time-Of-Day • Probablyrelated to travel purpose
Results – Station characteristics • Station layout important for passenger waiting times • Difference between Time-Of-Day • Probablyrelated to travel purpose
Results – Station characteristics • Station layout important for passenger waiting times • Difference between Time-Of-Day • Probablyrelated to travel purpose Percentageuniformly random
Results – Station characteristics • Station layout important for passenger waiting times • Difference between Time-Of-Day • Probablyrelated to travel purpose Avg. waiting times
Conclusions • Conclusions • Waiting time distribution canbe modelled as a mixture of Uniform and Beta distributions • The lower the frequency the more timed arrivals • Implications • Important to provide real timetables to passengers • Framework canimprove waiting time estimations in transport models
Thank you for your attention! Jesper Bláfoss Ingvardson Ph.d.-student, Technical University of Denmark jbin@dtu.dk