Mini-Workshop on Public Transport: Smart-Card Data Analytics and Modelling
15 Feb 2018, ITS
11:00 – 12:00 Using automated data sources to improve the performance of public transport systems
Prof Nigel Wilson, MIT
13:30 – 14:00 Estimation of OD matrix for bus ridership from smart card and GPS data
Tianli Tang, University of Leeds
14:00 – 14:30 Estimating route choice from observed journey time
Tamas Nadudvari, University of Leeds
14:30 – 15:00 Modelling consideration of alternatives as a function of individuals attitude and perception: an application to airport access mode choice
Mauro Capurso, University of Leeds
15:00 – 16:00 On the effect of real time information on bus bunching
Dr Achille Fonzone, Edinburgh Napier University
Non-University of Leeds participants are requested to confirm attendance by emailing Z.Clough_at_leeds.ac.uk by 14 Feb 2018.
Using automated data sources to improve the performance of public transport systems: a framework and application, Prof Nigel Wilson, MIT
Automatic data sources including automatic vehicle location systems, automatic passenger counting systems and electronic fare payment and ticketing systems are becoming ubiquitous in large public transport systems and are starting to have an impact on the quality and availability of information for both off-line and real-time functions needed for service provision. The off-line functions include service and operations planning, and performance monitoring and measurement, while the critical real-time functions include operations management and control, and customer information. While the impacts of these advances are already apparent in many systems, there is the potential for much deeper impact in the future. The power and cost-effectiveness of information technology continues to advance and will offer opportunities to develop and apply more ambitious models which should positively affect many facets of the performance of public transport systems. This talk will present a framework for assessing the various roles that automated data sources can play in public transport systems and will summarize recent applications of the resulting methods based on research at MIT for Transport for London and other transit agencies. The potential for further enhancement of critical public transport agency functions in the future making even greater use of these data sources will be discussed.
Estimation of OD matrix for bus ridership from smart card and GPS data, Tianli Tang, ITS
In recent years, vast amounts of travel data have been gained because a variety of automatic data collectors are introduced in the public transit. Meanwhile, many researchers are focusing on these big data and trying to observe the travel information from the data to support the public transport planning. The aim of this study is to develop a data fusion methodology for estimating the bus ridership using the smart card and GPS data. Since only boarding time is recorded in the smart card data, the location information from GPS helps to complement the full boarding information including stations and lines. Then, the destinations of a part of trips are inferred due to the assumptions of trip chains and transfer trips. Finally, the unknown trips are classified into the inferred trips by many characteristic variables. In addition, the decision-making tree is applied in the classification process. And the expected result is to achieve a possible destination for each trip. The main contribution of this study is to restore as many trips as possible using automatically collected data. The approaches will be tested using a nine-line public transport network in Changsha from six months in April 2016. The current result shows that the destination only 30% trips can be estimated by the past methods
Estimating Route Choice from Observed Journey Time, Tamas Nadudvari, ITS
Passenger Route Choice (RC) and flow is an important information for operators of metro network. With the widespread of Smart Cards for fare payment, a wealth of data is available on passengers’ actual journeys; however they reveal only entry and exit transactions and hence route chosen remains unobserved. This research examines whether RC can be estimated with confidence in case only the distribution of Observed Journey Times (OJT) is available from Smart Card data. We proposed to extend the work of Fu (2014) on the use of Mixture Model to estimate the distribution of OJT. In this talk we focus on overcoming the issue of the data availability for single station-to-station OD pairs. We propose to group a set of origin and destination stations – we call them superstations – with the property, that for any single station-to-station OD pair whose origin and destination station belongs to the given superstation, same route choice set and similar route choice probabilities are expected. Having this definition and assumption, it is possible to estimate the route choice probabilities for superstation-to-superstation OD pairs from their centroid-to-centroid OJTs, which enables us to applying the MM on larger sample of data.
Modelling consideration of alternatives as a function of individuals’ attitudes and perceptions: an application to airport access mode choice, Mauro Capurso, ITS
For air travellers, trips to the airport are the first “leg” of a longer trip and are associated with a hard constraint (i.e., the departure time of the flight). Hence, the possible consequences of a delay in arriving at the airport may be severe. Given this, we investigate whether travellers actually ‘consider’ all possible access alternatives as feasible - i.e., if they make standard trade-offs between attributes (such as travel time/cost, or frequency for public transport) of all alternatives - or if they discard a priori some alternatives and make choices from restricted choice sets. Assuming that all alternatives are ‘fully considered’ when this is not the case has been proven to have severe implications on parameter estimates and demand forecasting; this becomes extremely important when models’ outcomes affect decisions at other levels, such as decisions on supply and pricing of transport services and infrastructures. In this paper, we specifically model airport access mode choices to Bari International Airport, in Italy, using stated preference data and discrete choice models. In particular, we assume that consideration of alternatives can be modelled as a function of individuals’ personal judgement on the reliability of alternatives and contextual factors.
On the effect of real time information on bus bunching, Dr Achille Fonzone, Edinburgh Napier University
Real time information allows passengers to time their arrival to bus stop so as to reduce their wait time. The aggregate result is a peaked passenger arrival distribution at bus stops, which has been shown to potentially lead to more serious bus bunching. In this paper, we examine the trade-off between these two effects of real time information. We formulate an analytical model of information content (in terms of a predicted bus departure times) and information dissemination (in terms of its update frequency and reliability). We model passenger’s perception and response to information, and derive instantaneous arrival flows based on passengers’ risk-averse disutility to wait time and a penalty for missing the last attractive bus. We show that frequent information updates and moderately fuzzy (none precise) information can be helpful in spreading the passenger arrival times and reducing bus bunching. Crisp more accurate information may not always be the best for overall system performance. The results have implications on the design of real time information dissemination strategies.