Passenger arrival and waiting time distributions dependent on train service frequency and station characteristics: A smart card data analysis

Waiting time at public transport stops is perceived by passengers to be more onerous than in-vehicle time, hence it strongly influences the attractiveness and use of public transport. Transport models traditionally assume that average waiting times are half the service headway by assuming random passenger arrivals. However, research agree that two distinct passenger behaviour types exist: one group arrives randomly, whereas another group actively tries to minimise their waiting time by arriving in a timely manner at the scheduled departure time. This study proposes a general framework for estimating passenger waiting times which incorporates the arrival patterns of these two groups explicitly, namely by using a mixture distribution consisting of a uniform and a beta distribution. The framework is empirically validated using a large-scale automatic fare collection system from the Greater Copenhagen Area covering metro, suburban, and regional rail stations thereby giving a range of service headways from 2 to 60 min. It was shown that the proposed mixture distribution is superior to other distributions proposed in the literature. This can improve waiting time estimations in public transport models. The results show that even at 5-min headways 43% of passengers arrive in a timely manner to stations when timetables are available. The results bear important policy implications in terms of providing actual timetables, even at high service frequencies, in order for passengers to be able to minimise their waiting times.     

An application-oriented model of passenger waiting time based on bus departure time intervals

This paper develops an application-oriented model to estimate waiting times as a function of bus departure time intervals. Bus stops are classified into Type A and B depending on whether they are connected with urban rail transit systems. Distributions of passenger arrival rates are analyzed based on field data for Beijing. The results indicate that the best fits for the distribution of passenger arrival rates for Type A and B bus stops are the lognormal distribution and gamma distribution, respectively. By analyzing relationships between passenger arrival rates and bus departure time intervals, it is demonstrated that parameters of the passenger arrival rate distribution can be expressed by the average and coefficient of variation of bus departure time intervals in functional relationships. The validation shows that the model provides a reliable estimation of the average passenger waiting time based on readily available bus departure time intervals.     

Development of a probabilistic model to optimize disseminated real‐time bus arrival information for pre‐trip passengers

In the advent of Advanced Traveler Information Systems (ATIS), the total wait time of passengers for buses may be reduced by disseminating real‐time bus arrival times for the next or series of buses to pre‐trip passengers through various media (e.g., internet, mobile phones, and personal digital assistants). A probabilistic model is desirable and developed in this study, while realistic distributions of bus and passenger arrivals are considered. The disseminated bus arrival time is optimized by minimizing the total wait time incurred by pre‐trip passengers, and its impact to the total wait time under both late and early bus arrival conditions is studied. Relations between the optimal disseminated bus arrival time and major model parameters, such as the mean and standard deviation of arrival times for buses and pre‐trip passengers, are investigated. Analytical results are presented based on Normal and Lognormal distributions of bus arrivals and Gumbel distribution of pre‐trip passenger arrivals at a designated stop. The developed methodology can be practically applied to any arrival distributions of buses and passengers.     

A stochastic optimization model for transit network timetable design to mitigate the randomness of traveling time by adding slack time

Transit network timetabling aims at determining the departure time of each trip of all lines in order to facilitate passengers transferring either to or from a bus. In this paper, we consider a bus timetabling problem with stochastic travel times (BTP-STT). Slack time is added into timetable to mitigate the randomness in bus travel times. We then develop a stochastic integer programming model for the BTP-STT to minimize the total waiting time cost for three types of passengers (i.e., transferring passengers, boarding passengers and through passengers). The mathematical properties of the model are characterized. Due to its computational complexity, a genetic algorithm with local search (GALS) is designed to solve our proposed model (OPM). The numerical results based on a small bus network show that the timetable obtained from OPM reduces the total waiting time cost by an average of 9.5%, when it is tested in different scenarios. OPM is relatively effective if the ratio of the number of through passengers to the number of transferring passengers is not larger than a threshold (e.g., 10 in our case). In addition, we test different scale instances randomly generated in a practical setting to further verify the effectiveness of OPM and GALS. We also find that adding slack time into timetable greatly benefits transferring passengers by reducing the rate of transferring failure.     

Transfer demand prediction for timed transfer coordination in public transport operational control

Timed transfer coordination in public transit reduces passenger transfer time by providing seamless interconnected transfers. The problem arises when a Receiving Vehicle (RV) arrives to the transfer stop before a Feeding Vehicle (FV) carrying transferring passengers. Timed transfer coordination in operational control dynamically decides whether a RV is held at the transfer stop to allow transfers, or departs as scheduled. While transfer demand is essential for implementing timed transfer coordination, this variable is generally not available in public transit because of the lack of passenger transfer plans. The problem of acquiring this variable in real‐time has also received limited attention in the related literature. This paper proposes a new method to dynamically predict the transfer demand. We anticipate the transferring probability from each individual passenger by examining historical travel itineraries. Three different types of models (simple analytical, statistical, and computation intelligence model) are developed to forecast the number of transferring passengers. Numerical experiments using observed Automatic Vehicle Location and Automatic Fare Collection data from South East Queensland, Australia show the accuracy and applicability of the proposed models in timed transfer coordination. Copyright © 2017 John Wiley & Sons, Ltd.     

A bi-level model for single-line rail timetable design with consideration of demand and capacity

This paper proposes a bi-level model to solve the timetable design problem for an urban rail line. The upper level model aims at determining the headways between trains to minimize total passenger cost, which includes not only the usual perceived travel time cost, but also penalties during travel. With the headways given by the upper level model, passengers’ arrival times at their origin stops are determined by the lower level model, in which the cost-minimizing behavior of each passenger is taken into account. To make the model more realistic, explicit capacity constraints of individual trains are considered. With these constraints, passengers cannot board a full train, but wait in queues for the next coming train. A two-stage genetic algorithm incorporating the method of successive averages is introduced to solve the bi-level model. Two hypothetical examples and a real world case are employed to evaluate the effectiveness of the proposed bi-level model and algorithm. Results show that the bi-level model performs well in reducing total passenger cost, especially in reducing waiting time cost and penalties. And the section loading-rates of trains in the optimized timetable are more balanced than the even-headway timetable. The sensitivity analyses show that passenger’s desired arrival time interval at destination and crowding penalty factor have a high influence on the optimal solution. And with the dispersing of passengers' desired arrival time intervals or the increase of crowding penalty factor, the section loading-rates of trains become more balanced.     

Passenger-demands-oriented train scheduling for an urban rail transit network

This paper considers the train scheduling problem for an urban rail transit network. We propose an event-driven model that involves three types of events, i.e., departure events, arrival events, and passenger arrival rates change events. The routing of the arriving passengers at transfer stations is also included in the train scheduling model. Moreover, the passenger transfer behavior (i.e., walking times and transfer times of passengers) is also taken into account in the model formulation. The resulting optimization problem is a real-valued nonlinear nonconvex problem. Nonlinear programming approaches (e.g., sequential quadratic programming) and evolutionary algorithms (e.g., genetic algorithms) can be used to solve this train scheduling problem. The effectiveness of the event-driven model is evaluated through a case study.     

Assessing the financial and social costs of public transport in differing operating environments and with endogenous demand

This paper uses a previously developed spreadsheet cost model which simulates public transport modes operated on a 12-km route to analyse the total costs of different passenger demand levels. The previous cost model was a very powerful tool to estimate the social and operator costs for different public transport technologies. However, as the model is strategic, some basic assumptions were made which are relaxed in this paper. First, the speed-flow equation in the original spreadsheet model assumes that speed decreases according to the ratio of the current frequency and the lane capacity which is based on the safety headway without taking into account passenger boardings. However, this may vary in different operating environments. Therefore, the speed-flow equation is improved by moving from a linear equation to a piecewise equation that considers the features of different operating environments. Second, the model assumes that supply is sufficient to meet demand. However, when the level of demand is high for the lower-capacity public transport technologies, passengers may find the incoming vehicle full and therefore, they have to wait more than one service interval. This paper applies queuing theory to investigate the probability of having to wait longer than the expected service headways which will affect the average passenger waiting time. The extra waiting time for each passenger is calculated and applied in the spreadsheet cost model. Third, the original model assumed that demand was externally fixed (exogenous). To evaluate the differences after applying these equations, endogenous demand rather than exogenous demand will be investigated by using the elasticities for passenger waiting time and journey time.     

Comparing static and dynamic threshold based control strategies

This research extends a static threshold based control strategy used to control headway variation to a dynamic threshold based control strategy. In the static strategy, buses are controlled by setting a threshold value that holds buses at a control point for a certain amount of time before allowing the bus to continue along the route. The threshold remains constant each time the bus stops at the control point. The dynamic strategy involves the same principle of holding buses at a bus stop; however, a different threshold value is chosen each time the bus holds at a control point. The results indicate that in cases where the static threshold is set equal to the scheduled headway, very low headway variation and passenger system times result; however, passengers on board the bus are penalized by extra delay on the bus while waiting at the control point. The dynamic strategy reduces the penalty to passengers delayed on-board the bus at a control point at the expense of a slight increase in overall passenger system time.The results indicate that in most cases, the tradeoff of the slight increase in waiting time for the significant decrease in on-board delay penalty makes the dynamic strategy an acceptable choice.     

User perspectives in public transport timetable optimisation

The present paper deals with timetable optimisation from the perspective of minimising the waiting time experienced by passengers when transferring either to or from a bus. Due to its inherent complexity, this bi-level minimisation problem is extremely difficult to solve mathematically, since timetable optimisation is a non-linear non-convex mixed integer problem, with passenger flows defined by the route choice model, whereas the route choice model is a non-linear non-continuous mapping of the timetable. Therefore, a heuristic solution approach is developed in this paper, based on the idea of varying and optimising the offset of the bus lines. Varying the offset for a bus line impacts the waiting time passengers experience at any transfer stop on the bus line.In the bi-level timetable optimisation problem, the lower level is a transit assignment calculation yielding passengers’ route choice. This is used as weight when minimising waiting time by applying a Tabu Search algorithm to adapt the offset values for bus lines. The updated timetable then serves as input in the following transit assignment calculation. The process continues until convergence.The heuristic solution approach was applied on the large-scale public transport network in Denmark. The timetable optimisation approach yielded a yearly reduction in weighted waiting time equivalent to approximately 45 million Danish kroner (9 million USD).     

A case study on the coordination of last trains for the Beijing subway network

Passengers may make several transfers between different lines to reach their destinations in urban railway transit networks. Coordination of last trains in feeding lines and connecting lines at transfer stations is especially important because it is the last chance for many travellers to transfer. In this paper, a mathematical method is used to reveal the relationships between passenger transfer connection time (PTCT) and passenger transfer waiting time (PTWT). A last-train network transfer model (LNTM) is established to maximize passenger transfer connection headways (PTCH), which reflect last-train connections and transfer waiting time. Additionally, a genetic algorithm (GA) is developed based upon this LNTM model and used to test a numerical example to verify its effectiveness. Finally, the Beijing subway network is taken as a case study. The results of the numerical example show that the model improves five connections and reduces to zero the number of cases when a feeder train arrives within one headway’s time after the connecting train departed.     

Train scheduling for minimizing passenger waiting time with time-dependent demand and skip-stop patterns: Nonlinear integer programming models with linear constraints

This paper focuses on how to minimize the total passenger waiting time at stations by computing and adjusting train timetables for a rail corridor with given time-varying origin-to-destination passenger demand matrices. Given predetermined train skip-stop patterns, a unified quadratic integer programming model with linear constraints is developed to jointly synchronize effective passenger loading time windows and train arrival and departure times at each station. A set of quadratic and quasi-quadratic objective functions are proposed to precisely formulate the total waiting time under both minute-dependent demand and hour-dependent demand volumes from different origin–destination pairs. We construct mathematically rigorous and algorithmically tractable nonlinear mixed integer programming models for both real-time scheduling and medium-term planning applications. The proposed models are implemented using general purpose high-level optimization solvers, and the model effectiveness is further examined through numerical experiments of real-world rail train timetabling test cases.     

Optimal slack time for timed transfers at a transit terminal

At transit terminals where two routes interchange passengers, total system costs may be reduced by allowing some “slack” time in the vehicle schedules to decrease the probability of missed connections. Transfer cost functions are formulated and used to determine optimal slack time for simple systems with transfers between one bus route and one rail line. Some analytic results are derived for empirical discrete and Gumbel distributions of bus arrival times. Relations between the optimal slack times and headways, transfer volumes, passenger time values, bus operating costs, and standard deviations of bus and train arrivals are also developed numerically using normally distributed arrivals. However, the proposed numerical approach can optimize slack times for any observed arrival distributions. The results provide some guidelines on desirable slack times and show that schedule coordination between the two routes is not worth attempting when standard deviations of arrivals exceed certain levels. Possible extensions of this work are suggested in the last section.     

A model of bus bunching under reliability-based passenger arrival patterns

If bus service departure times are not completely unknown to the passengers, non-uniform passenger arrival patterns can be expected. We propose that passengers decide their arrival time at stops based on a continuous logit model that considers the risk of missing services. Expected passenger waiting times are derived in a bus system that allows also for overtaking between bus services. We then propose an algorithm to derive the dwell time of subsequent buses serving a stop in order to illustrate when bus bunching might occur. We show that non-uniform arrival patterns can significantly influence the bus bunching process. With case studies we find that, even without exogenous delay, bunching can arise when the boarding rate is insufficient given the level of overall demand. Further, in case of exogenous delay, non-uniform arrivals can either worsen or improve the bunching conditions, depending on the level of delay. We conclude that therefore such effects should be considered when service control measures are discussed.     

An equity‐based passenger flow control model with application to Hong Kong‐Shenzhen border‐crossing

Cross‐border passengers from Hong Kong to Shenzhen by the east Kowloon‐Canton Railway (KCR) through the Lo Wu customs exceed nearly 200 thousand on a special day such as a day during the Chinese Spring Festival. Such heavy passenger demand often exceeds the processing and holding capacity of the Lo Wu customs for many hours a day. Thus, passengers must be metered off at all entrance stations along the KCR line through ticket rationing to restrain the number of passengers waiting at Lo Wu within its safe holding capacity. This paper proposes an optimal control strategy and model to deal with this passenger crowding and control problem. Because the maximum passenger checkout rate at Lo Wu is fixed, total passenger waiting time is not affected by the control strategy for given time‐dependent arriving rates at each station. An equity‐based control strategy is thus proposed to equalize the waiting times of passengers arriving at all stations at the same time. This equity is achieved through optimal allocation of the total quota of tickets to all entrance stations for each train service. The total ticket quota for each train service is determined such that the capacity constraint of the passenger queue at Lo Wu is satisfied. The control problem is formulated as a successive linear programming problem and demonstrated for the KCR system with partially simulated data.     

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

In this study, we focus on improving system-wide equity performance in an oversaturated urban rail transit network based on multi-commodity flow formulation. From the system perspective, an urban rail transit network is a distributed system, where a set of resources (i.e., train capacity) is shared by a number of users (i.e., passengers), and equitable individuals and groups should receive equal shares of resources. However, when oversaturation occurs in an urban rail transit network during peak hours, passengers waiting at different stations may receive varying shares of train capacity leading to the inequity problem under train all-stopping pattern. Train skip-stopping pattern is an effective operational approach, which holds back some passengers at stations and re-routes their journeys in the time dimension based on the available capacity of each train. In this study, the inequity problem in an oversaturated urban rail transit network is analyzed using a multi-commodity flow modeling framework. In detail, first, discretized states, corresponding to the number of missed trains for passengers, are constructed in a space-time-state three-dimensional network, so that the system-wide equity performance can be viewed as a distribution of all passengers in different states. Different from existing flow-based optimization models, we formulate individual passenger and train stopping pattern as commodity and network structure in the multi-commodity flow-modeling framework, respectively. Then, we aim to find an optimal commodity flow and well-designed network structure through the proposed multi-commodity flow model and simultaneously achieve the equitable distribution of all passengers and the optimal train skip-stopping pattern. To quickly solve the proposed model and find an optimal train skip-stopping pattern with preferable system-wide equity performance, the proposed linear programming model can be effectively decomposed to a least-cost sub-problem with positive arc costs for each individual passenger and a least-cost sub-problem with negative arc costs for each individual train under a Lagrangian relaxation framework. For application and implementation, the proposed train skip-stopping optimization model is applied to a simple case and a real-world case based on Batong Line in the Beijing Subway Network. The simple case demonstrates that our proposed Lagrangian relaxation framework can obtain the approximate optimal solution with a small-gap lower bound and a lot of computing time saved compared with CPLEX solver. The real-world case based on Batong Line in the Beijing Subway Network compares the equity and efficiency indices under the operational approach of train skip-stopping pattern with those under the train all-stopping pattern to state the advantage of the train skip-stopping operational approach.     

Crowding in public transport systems: Effects on users,operation and implications for the estimation of demand

The effects of high passenger density at bus stops, at rail stations, inside buses and trains are diverse. This paper examines the multiple dimensions of passenger crowding related to public transport demand, supply and operations, including effects on operating speed, waiting time, travel time reliability, passengers’ wellbeing, valuation of waiting and in-vehicle time savings, route and bus choice, and optimal levels of frequency, vehicle size and fare. Secondly, crowding externalities are estimated for rail and bus services in Sydney, in order to show the impact of crowding on the estimated value of in-vehicle time savings and demand prediction. Using Multinomial Logit (MNL) and Error Components (EC) models, we show that alternative assumptions concerning the threshold load factor that triggers a crowding externality effect do have an influence on the value of travel time (VTTS) for low occupancy levels (all passengers sitting); however, for high occupancy levels, alternative crowding models estimate similar VTTS. Importantly, if demand for a public transport service is estimated without explicit consideration of crowding as a source of disutility for passengers, demand will be overestimated if the service is designed to have a number of standees beyond a threshold, as analytically shown using a MNL choice model. More research is needed to explore if these findings hold with more complex choice models and in other contexts.     

Optimizing coordinated transfer with probabilistic vehicle arrivals and passengers' walking time

Supporting efficient connections by synchronizing vehicle arrival time and passengers' walking time at a transfer hub may significantly improve service quality, stimulate demand, and increase productivity. However, vehicle travel times and walking times in urban settings often varies spatially and temporally due to a variety of factors. Nevertheless, the reservation of slack time and/or the justification of vehicle arrival time at the hub may substantially increase the success of transfer coordination. To this end, this paper develops a model that considers probabilistic vehicle arrivals and passengers walking speeds so that the slack time and the scheduled bus arrival time can be optimized by minimizing the total system cost. A case study is conducted in which the developed model is applied to optimize the coordination of multiple bus routes connecting at a transfer station in Xi'an, China. The relationship between decision variables and model parameters, including the mean and the standard deviation of walking time, is explored. It was found that the joint impact of probabilistic vehicle arrivals and passengers' walking time significantly affects the efficiency of coordinated transfer. The established methodology can essentially be applied to any distribution of bus arrival and passenger walking time. Copyright © 2017 John Wiley & Sons, Ltd.     

Multiple objective programming for bus operation: A case study for Taipei city

This paper, based on the viewpoints of both the passengers and the operator, uses a multi-objective utility function to formulate an effective bus transportation system. The utility function is composed of six factors of passenger concern including riding time, waiting time, degree of crowdedness, transferring frequencies, standing probability and walking time; and also the two major factors of concern to the bus operator: cost and profit. By using a multi-objective programming technique, a series of noninferior solutions are generated. Also, an application to the Taipei city bus system is presented.     

Dispatching control at transfer stations in multi‐hub transit networks

Among dispatching control approaches, the holding option has attracted the most attention in bus control. However, holding a vehicle at a transfer station may exacerbate the delays because more passengers might accumulate at downstream stations and may also affect other connecting routes at other transfer stations. Our problem is to minimize the total costs of dispatching ready vehicles at each transfer station along coordinated routes in a multi‐hub transit network. The total costs include the waiting cost for on‐board passengers, the missed connection costs for late arrival passengers at the subject transfer station and possible transfer costs at downstream transfer stations. We develop a heuristic algorithm to optimize the holding times based on real time information about late vehicles. The results show that ready vehicles should be held longer when the arrival variances of late vehicles are small or when many late connecting passengers are expected.