Predictive decision support platform and its application in crowding prediction and passenger information generation

https://doi.org/10.1016/j.trc.2021.103139Get rights and content

Highlights

  • We propose a real-time predictive decision support platform which addresses both, operations control and customer information needs.

  • The system provides crowding predictions on trains and platforms, communicates this information to passengers

  • Using a fixed-point formulation, the methodology accounts for the expected passenger behavior in response to the predictive crowding information

  • It is demonstrated through a case study that providing predictive information to passengers can potentially reduce denied boarding and lead to better utilization of train capacity.

Abstract

Demand for public transport has witnessed a steady growth over the last decade in many densely populated cities around the world. However, capacity has not always matched this increased demand. As such, passengers experience long waiting times and are denied boarding during the peak hours. Crowded platforms and the subsequent customer dissatisfaction and safety issues have become a serious concern. The COVID-19 pandemic has dramatically reduced passengers’ willingness to board crowded trains, causing a surge in demand for real-time crowding information. In this paper, we propose a real-time predictive decision support platform which addresses both, operations control and customer information needs. The system provides crowding predictions on trains and platforms, communicates this information to passengers, and takes into account their response to it. It is demonstrated through a case study that providing predictive information to passengers can potentially reduce denied boarding and lead to better utilization of train capacity.

Introduction

Rail transit systems in major metropolitan areas (e.g., London, Hong Kong) often operate close to capacity during the peak hours. As such, it is not uncommon for passengers to experience denied boarding and long waiting times. Crowded platforms and the subsequent customer dissatisfaction and safety issues have become a serious concern. Various studies have shown that passengers are willing to pay for less crowding on the trains. Douglas and Karpouzis (2006) estimated that 1 minute of uncrowded seating on a train is equivalent to 1.34 minutes standing for up to 10 minutes, 1.81 standing for over 20 minutes or longer, and 2.04 for crush standing up to 10 minutes. They also estimated that one minute of waiting on the platform under high-crowding conditions was the same as 1.7 to 2.5 minutes of medium-crowding waiting time. Börjesson and Rubensson (2019) studied passenger satisfaction over 15 years in Stockholm, where they find overcrowding to be one of the main drivers of travel dissatisfaction on public transit. Li and Hensher (2011) provide a comprehensive review of passengers’ willingness to pay for less crowded conditions.

Another development is the rapid pace of technological innovation. Smartphones have enabled companies to interact with their customers directly and seamlessly, providing personalized information in real-time. The prevalence of such applications has raised the expectation among passengers to receive a similar type of information about their expected trip conditions. This presents a unique opportunity for transit system operators. Real-time information can potentially influence passengers’ trip making decisions. Some users might decide to shift their arrival time to avoid overcrowded stations. Similarly, some passengers might decide to wait for a train after the one that is currently arriving at a station, if they are informed that this action will result in a better trip experience (if, for example, the following train is not as crowded). In a recent study (Drabicki et al., 2020), it is reported that 50 percent of survey respondents stated that they would consider waiting up to 10 minutes for a less-crowded train.

The global COVID-19 pandemic has made passengers wary of boarding crowded trains in lieu of the associated health risks. As such, many transit agencies have started to provide real-time crowding information to passengers, but these are mainly based on the historical values.

This paper proposes a predictive decision support platform that addresses both, operations control and customer information needs. It enables the operators to foresee expected overcrowded platforms and trains. It also generates information on the likelihood of being able to board upcoming trains, which can be communicated to passengers inside the system and those about to travel. Using this information, passengers can make better informed decisions as to which train to board.

The proposed decision support platform comprises two major components: a demand prediction module, and an on-line simulation. The first module provides short-term (e.g., 15 minutes) prediction of the number of passengers arriving at each station and their destinations. Subsequently, the on-line simulation captures the interactions of trains and passengers for the duration of the prediction horizon. It is assumed that the predictive crowding information is displayed at platforms and passengers base their boarding decisions based on it. The system directly incorporates the predicted passenger response to the information. Outputs from the simulation include the predicted status of trains and platforms during that time period, and expected number of passengers who will experience denied boarding.

This paper makes a number of contributions to the current literature. First, it provides predictive crowding information, by integrating demand prediction models into the decision support platform. Second, it assumes that this predictive crowding information is available to passengers and affects their boarding decisions. The problem is treated using a fixed-point approach, where the crowding information affects boarding decisions, which in turn affects the predicted crowding. This is an important feature of the proposed approach, since ignoring this interaction can result in inaccurate information. This can lead to erosion of trust in the information provided, and is studied in detail in this paper.

The rest of this paper is organized as follows. Section 2 provides an overview of previous work on decision support platforms and their application in crowding prediction. Section 3 describes the simulation-based framework and the predictive information generation process. Section 4, presents the application of the decision support and discusses the accuracy of the crowding predictions as well as the impact of real-time information. Section 5 concludes the paper.

Section snippets

Literature review

Over the last decade, Automated Fare Collection systems (AFC), Automatic Passenger Counter systems (APC), Automated Vehicle Location systems (AVL), and many other technologies have become increasingly adopted by transit agencies. Many cities currently provide real-time information on the arrival times of trains to passengers. However, information about the expected quality of service (e.g., crowding, seat availability) are still rare. For traffic applications, many papers have studied the value

Methodology

The decision support adopts a rolling-horizon approach (Fig. 1). At the start of the updating phase, e.g. 8 am, the system makes predictions for the next 15 (1-step) and 30 (2-steps) minutes based on the available information. The predicted OD flows are used to predict future system performance and generate predictive crowding information that is used by passengers to make a decision about which train to board. Fig. 2 illustrates the structure of the computational engine of the decision support

Application: prediction of train and platform crowding in the presence of information

We present an application of the decision support platform using a subset of a subway network in a major city. The study area consists of 46 stations on two lines for a total of 2070 OD pairs. The analysis is focused on the morning period, with the decision support updating every 15 minutes. This 15 minute time window is long enough to observe AFC records (which are typically available with a lag of a couple of minutes). It also allows the positional correction for more severe real-time

Conclusion

In this paper, we presented an on-line, self-correcting, simulation-based decision support platform. It utilizes real-time train locations to frequently update its state representation of the transit network, and AFC data to predict origin-destination demand for the next few time periods. The simulation models the interaction between supply and demand, predicting the near-future state of the transit network. Expected number of passengers waiting on platforms, load of each train, and expected

References (41)

  • A. Nuzzolo et al.

    A mesoscopic transit assignment model including real-time predictive information on crowding

    Journal of Intelligent Transportation Systems: Technology, Planning, and Operations

    (2016)
  • G.E. Sánchez-Martínez et al.

    Real-time holding control for high-frequency transit with dynamics

    Transportation Research Part B: Methodological

    (2016)
  • P. Shang et al.

    Equity-oriented skip-stopping schedule optimization in an oversaturated urban rail transit network

    Transportation Research Part C: Emerging Technologies

    (2018)
  • L. Sun et al.

    Demand-driven timetable design for metro services

    Transportation Research Part C: Emerging Technologies

    (2014)
  • J. Yin et al.

    Dynamic passenger demand oriented metro train scheduling with energy-efficiency and waiting time minimization: Mixed-integer linear programming approaches

    Transportation Research Part B: Methodological

    (2017)
  • K.G. Zografos et al.

    A real-time decision support system for roadway network incident response logistics

    Transportation Research Part C: Emerging Technologies

    (2002)
  • R. Balakrishna et al.

    Simulation-based evaluation of advanced traveler information systems

    Transportation Research Record: Journal of the Transportation Research Board

    (2005)
  • M. Ben-Akiva et al.

    Traffic simulation with dynamit

  • M.E. Ben-Akiva et al.

    Simulation laboratory for evaluating dynamic traffic management systems

    Journal of Transportation Engineering

    (1997)
  • O. Cats et al.

    Mesoscopic modeling of bus public transportation

    Transportation Research Record: Journal of the Transportation Research Board

    (2010)
  • Cited by (15)

    • Modeling the effect of real-time crowding information (RTCI) on passenger distribution in trains

      2022, Transportation Research Part A: Policy and Practice
      Citation Excerpt :

      The application of this model to the urban public transport network in Krakow, Poland shows improvements in passengers’ travel experience. Using a fixed-point solution, Noursalehi et al. (2021) presented an on-line simulation-based decision support platform that provides predictive information on crowding on platforms and trains. The information is provided through a three-level descriptive scale based on the train boarding likelihood.

    View all citing articles on Scopus
    View full text