Quality of Service and Congestion in Urban Public Transport: a Micro-simulation Approach

Quality of Service and Congestion in Urban Public Transport: a Micro-simulation Approach


Vincent Benezech, ITF, International Transport forum of the OECD, Francois Combes, CEREMA and LVMT, Ecole de Ponts, ParisTech, Julien Ponton, Ecole Des Ponts ParisTech


This paper introduces a micro-simulation approach to build a link function relating the passenger flow on a public transport line to its quality of service. Numerical illustrations are also provided.


A public transport is often characterized by its “capacity”, i.e. the maximum number of passengers it is able to transport over a given time period, or the maximum passenger flow rate it is able to accommodate. However, defining the capacity of a transport line is not as easy as it may seem. Indeed, a public transport line is similar to a road in that the level of service it provides decreases with the number of those users. This has two consequences: first, a capacity can only be defined for a given level of service, and not in general; second, public transport lines are congestion prone services, with all the consequences.
The relationship between passenger flows and quality of service is not well understood yet. While transport engineering has investigated quite in detail the relationship of car flow and quality of service for the case of road transport (this relationship is captured by the classic “speed-flow diagram” that is inherent to each road infrastructure), similar study for public transport are limited to the estimation of very basic relationship between the number of transferring passengers and the dwell times of vehicles.
The objective of this paper is to provide a methodology to characterize a public transport line by a relationship between passenger flow rate and quality of service. This raises three difficulties. First, it is observable, for a daily user of urban public transport, that when the demand reaches a certain level on a given public transport line, travel time reliability decreases at least as fast as speed. This means that the relationship studied in this paper should account not only for expected travel time, but also for travel time variability. Second, the reaction of a public transport line to its passenger demand heavily depends on the way it is operated: attention should be paid to the operation protocol of the line, and its parameters. Third, many effects take place, and their roles are intertwined: let us quote the influence of the demand on dwell times, and then on vehicle movements, and on headways, etc.
The methodology for this paper is based on a model representing a transport line with a given fleet size and schedule, transporting a given transport demand. This model is a multi-agent model, incorporating two categories of agents. First, travelling agents (which represent either a unique passenger or a very small group of travelers), entering the line at their origin station and exiting it at their destination station; second, vehicle agents, which can carry travelers, and move along the line according to the operating protocol and to the constraints they are confronted to. The model allows to examine the average travel time and travel time dispersion perceived by passengers, and to define a function, akin to a microeconomic production function, relating the passenger flow to the line’s quality of service, i.e. its speed and reliability. Note that while the model could be easily extended to take into account comfort, this dimension is left aside for simplicity.
The paper is organized in four parts. In the first part, the relevant literature related to quality of service and public transport simulation is examined. It shows that, while micro-simulation models are somehow common to examine line operations, the interaction with quality of service as experienced by passengers is never studied. The second part presents a basic microeconomic discussion of the characterization of the line in terms of quality of service that is used as a theoretical basis for the results of the micro-simulation. The third part presents the architecture of the multi-agent model, the parameters that are included and the way the agents interact. In the final part, some simulation results are discussed. The simulation focuses on one suburban rail line from the Paris region. It is carried out for several days of the year having differing levels of ridership. Simulation results are close to observations, showing the capacity of the model to capture the link between ridership and quality of service.


Association for European Transport