Posts tagged APC

Supervised learning: Predicting passenger load in public transport

For many Public Transport (PT) users, overcrowding in PT vehicles has a major decreasing effect on the comfort experience. However, most online routing applications still not take comfort regarding to crowdedness into account, but provide recommendations based on shortest distance, shortest travel-time, or number of interchanges.
Being able to include certain information on crowdedness, requires knowledge about the current and future level of passenger load. Increasing amount and complexity of data describing public transport services allows us to better explore the detection methods and analysis of different phenomena of PT operations. Some countries or operators provide the possibility to use Smart Card (SC) data for occupancy prediction. However, SC data is not available in real time, which makes it hard to incorporate it into real time recommendation models. In this work, we show that it is possible to predict the passenger load via supervised learning, eliminating the need for fare collection data beyond the set needed for training.

Find the CASPT presentation by Léonie Heydenrijk-Ottens HERE

Driver schedule efficiency vs. public transport robustness: A framework to quantify this trade-off based on passive data

More complex, efficient driver schedules reduce operator costs during undisrupted operations, but increase the disruption impact for passengers and operator once a disruption occurs. We develop an integrated framework to quantify the passenger and operator costs of disruptions explicitly as function of different driver schedule schemes. Since the trade-off between driver schedule efficiency and robustness can be quantified, this supports operators in their decision-making.

Read the CASPT paper by Menno Yap HERE and find the presentation HERE

Assessing disruption management strategies in rail-bound urban public transport from a passenger perspective

This paper provides a framework for generating and assessing alternatives
in case of disruptions in rail-bound urban public transport systems,. The proposed
framework considers the passenger perspective as well as the operator perspective,
for the often-used measures of detouring and short-turning. An application of the
framework demonstrates that currently used disruption management protocols often
do not lead to the optimal solution from the passenger perspective. Furthermore, the
optimal choice between alternatives from passenger perspective shows to be
dependent on the passenger flows.

Read the CASPT paper HERE and find the presentation HERE

Data-driven transfer inference for public transport journeys during disruptions

Disruptions in public transport have major impact on passengers and disproportional effects on passenger satisfaction. The availability of smart card data gives opportunities to better quantify disruption impacts on passengers’ experienced journey travel time and comfort. For this, accurate journey inference from raw transaction data is required. Several rule-based algorithms exist to infer whether a passenger alighting and subsequent boarding is categorized as transfer or final destination where an activity is performed. Although this logic can infer transfers during undisrupted public transport operations, these algorithms have limitations during disruptions: disruptions and subsequent operational rescheduling measures can force passengers to travel via routes which would be non-optimal or illogical during undisrupted operations. Therefore, applying existing algorithms can lead to biased journey inference and biased disruption impact quantification. We develop and apply a new transfer inference algorithm which infers journeys from raw smart card transactions in an accurate way during both disrupted and undisrupted operations. In this algorithm we incorporate the effects of denied boarding, transferring to a vehicle of the same line (due to operator rescheduling measures as short-turning), and the use of public transport services of another operator on another network level as intermediate journey stage during disruptions. This results in an algorithm with an improved transfer inference performance compared to existing algorithms.

Find the paper HERE

A data-driven approach to infer spatial characteristics and service reliability of public transport hubs

Public transport hubs play an important and a central role in public transport networks by connecting several public transport lines from one or multiple network levels. Hubs can be characterized by a large relative and absolute number of transferring passengers between public transport services within the same network level and/or between different network levels. Hubs are especially important with respect to service reliability of passenger journeys, since missing connections at hubs can substantially increase the nominal and perceived passenger journey travel time. The availability of AFC and AVL data allows an in-depth analysis of hub definition, identification, characterization and reliability performance evaluation. Such analysis enables optimisation of synchronisation of schedules, thereby increase the level of service reliability.

Find our TransitData2017 presentation HERE

Insights into door-to-door travel patterns of public transport passengers

Public transport enables fast and reliable station to station journeys. To assess passenger travel patterns and to infer actual quality of service, smartcard and AVL data offer great opportunities. There is, however, an increasing interest in insights into access and egress dynamics of public transport riders as well. What is the size of a stop’s catchment area, which modes are used, and how long and reliable are access and egress times? The answers to these and other questions enable optimization of the total mobility system, thereby also increasing public transport ridership and efficiency. Sufficient biking access of public transport stops (routes and parking), for instance, offer opportunities to increase public transport stopping distances, thereby increasing operational speed and reliability, without compromising accessibility of service areas. We developed a methodology to calculate and demonstrate these dynamics by using new and existing data technologies, namely AVL, survey and new promising app.

Find the Transit Data Conference abstract HERE and our presentation HERE

Optimization of a passenger railway transportation plan considering mobility flows and service quality

This research focuses on designing transportation plan for SNCF Transilien (French railway
operator for the Parisian suburban mass transit). The objective is to develop methods
and decision support tools to propose a timetable adapted to the passenger demand in the
Parisian mass transit system, including comfort and reliability criterias.
This paper aims to present the first step of this research. We propose a graph theoretic
ILP formulation for the Line Planning Problem, minimizing both travelers travel time and
operating cost. We furthermore develop a multi-objective method to solve this problem.
This method offers a pool of solutions in order to let the final designer choose the solution.
We report computational results on real world instances provided from SNCF Transilien.

Check the RAIL Lille paper of Lucile Brethome et al. HERE

Improving predictions of the impact of disturbances on public transport usage based on smart card data

The availability of smart card data from public transport travelling the last decades allows analyzing current and predicting future public transport usage. Public transport models are commonly applied to predict ridership due to structural network changes, using a calibrated parameter set. Predicting the impact of planned disturbances, like temporary track closures, on public transport ridership is however an unexplored area. In the Netherlands, this area becomes increasingly important, given the many track closures operators are confronted with the last and upcoming years. We investigated the passenger impact of four planned disturbances on the public transport network of Den Haag, the Netherlands, by comparing predicted and realized public transport ridership using smart card data. A two-step search procedure is applied to find a parameter set resulting in higher prediction accuracy. We found that in-vehicle time in rail-replacing bus services is perceived ≈1.1 times more negatively compared to in-vehicle time perception in the initial tram line. Besides, passengers do not seem to perceive the theoretical benefit of the usually higher frequency of rail-replacement bus services compared to the frequency of the replaced tram line. At last, no higher waiting time perception for temporary rail-replacement services could be found, compared to regular tram and bus services. The new parameter set leads to substantially higher prediction accuracy compared to the default parameter set. It supports public transport operators to better predict the required supply of rail-replacement services and to predict the impact on their revenues.

Read our TRB paper HERE

Find the poster HERE

Investigating potential transit ridership by fusing smartcard and GSM data

The public transport industry faces challenges to cater for the variety of mobility patterns and corresponding needs and preferences of passengers. Travel habit surveys provide information on the overall travel demand as well as its spatial variation. However, it often does not include information on temporal variations. By means of data fusion of smartcard and Global System for Mobile Communications (GSM) data, spatial and temporal patterns of public transport usage versus the overall travel demand are examined. The analysis is performed by contrasting different spatial and temporal levels of smartcard and GSM data. The methodology is applied to a case study in Rotterdam, the Netherlands, to analyze whether the current service span is adequate. The results suggest that there is potential demand for 10 extending public transport service span on both ends. In the early mornings, right before transit operations are resumed, an hour-on-hour increase in visitor occupancy of 33-88% is observed in several zones, thereby showing potential demand for additional public transport services. The proposed data fusion method showed to be valuable in supporting tactical transit planning and decision making and can easily be applied to other origin-destination transport data.

Read our TRB paper HERE

Find our presentation HERE

Waar liggen kansen voor OV: datafusie GSM en chipkaart

De grootste uitdaging van de openbaar vervoer sector is om tegemoet te komen aan de verscheidenheid aan reispatronen, en de bijbehorende behoeften en preferenties, van reizigers. Het beter matchen van vraag en aanbod levert zowel een kwaliteitssprong als kostenreductie op en heeft dus alle aandacht. Bestaande databronnen helpen, maar zijn nog niet afdoende. De combinatie van nieuwe bronnen biedt echter hoopgevende resultaten. Door een innovatieve methodiek kunnen GSM- en anonieme chipkaartdata gecombineerd worden om de OV potentie in kaart te brengen.

Bestaande onderzoeken (zoals OViN) geven informatie over de totale reisbehoefte en de ruimtelijke spreiding hiervan. Deze huishoudsurveys bieden veelal echter geen inzicht in de spreiding van deze reisbehoefte over de tijd. Een nieuwe methodiek om GSM- met anonieme OV chipkaartdata te koppelen, geeft die inzichten wel. Door middel van deze datafusie kunnen zowel de ruimtelijke als temporele patronen van OV gebruik vergeleken worden met de totale ruimtelijke en temporele reispatronen. Dit geeft inzicht in de (mis)match van vraag en aanbod in ruimte én tijd. Ideaal dus als eerste stap voor het verbeteren van deze match: OV potentie kan zo worden opgespoord.
Deze methode is toegepast in een case study in Rotterdam om te onderzoeken of het huidige OV bedieningsinterval voldoende aansluit bij de latente vraag. De resultaten demonstreren dat er potentie is om het OV bedieningsinterval zowel in de vroege ochtend als in de late avond uit te breiden. In de vroege ochtend, juist voordat het OV wordt opgestart, kan een uur-tot-uur toename in bezoekersaantallen van 33% tot zelfs 88% worden waargenomen in diverse delen van de Rotterdamse regio. Dit illustreert de potentiële vraag voor extra openbaar vervoer aanbod in de vroege ochtend. Op vergelijkbare wijze is deze analyse uitgevoerd voor het OV aanbod in de late avond.
Deze innovatieve methode van datafusie is van grote toegevoegde waarde te zijn gebleken ter ondersteuning van OV planning. Deze datafusie methode kan ook eenvoudig worden toegepast op andere herkomst-bestemmingsdata.

Lees het CVS paper HIER

© 2011 TU Delft