Master Degree in Transport Planning and Operations

2014/2015

Performance analysis of a public transport interchange from the pedestrian

circulation perspective using microsimulation tools: the Colégio Militar case

study

André Filipe Ferreira Ramos*

* Student nr. 76819 – IST/UL

A R T I C L E I N F O A B S T R A C T

Keywords:

Pedestrian circulation

Public transport interchanges

Level of service

Microsimulation

Efficient alternatives to private transport are becoming truly necessary at a time when road traffic

congestion increases and vehicle emissions are a constant problem.

Public transport is, therefore, an essential part of the sustainable development of cities: network’s

organization is a major factor, as long as it allows efficient and comfortable transfers in interchanges,

providing to passengers a sense of continuity on their trips. However, these network nodes are often

associated with high pedestrian flows with constraints on pedestrian movement, which discourages

their use.

The analysis methods for the performance of public transport interchanges are usually based on

aggregate values, which may result in highly optimistic results. However, since the development of

microsimulation tools allows different ways of measuring these infrastructures’ performance, it

becomes necessary to establish appropriate guidelines, similar to those already in place for

conventional analysis.

With this in mind, and based on different studies regarding the use of microsimulation, a set of

“best practices" were compiled, and a complete methodology suggested, so that it can applied to both

existing interchanges and to the planning of future ones. One major change (compared with

conventional methods) involves considering 30-second periods in the performance analysis, instead of

considering the average value of the simulated period.

In order to apply this methodology, a microsimulation model of the Colégio Militar/Luz subway

station (in Lisbon) was developed, after which a diagnosis of the current situation was done and

different alternative scenarios tested, to both take advantage and demonstrate the potential of these

tools.

© 2015 IST/UL All rights reserved.

1. Introduction

Currently, public transport interchanges are crucial

points of the transport system, since they allow

passengers to transfer between different lines or transport

modes. However, the walking distances that are created

add some undesired uncertainty and inconvenience to

their trips (Ceder et al., 2013; de Abreu e Silva and

Bazrafshan, 2013; Guo and Wilson, 2011; Hadas and

Ranjitkar, 2012; Kalakou and Moura, 2014; McCord et al.,

2006; Nesheli and Ceder, 2014).

Simultaneously, microsimulation tools have been

evolving in the past years, and it is possible to simulate

the individual behavior of each pedestrian and the

interaction between them, as well as a reliable

representation of the surrounding environment

(Fernandes et al., 2013).

Yet, since the interchange analysis didn’t follow, to

date, any set of “guidelines” or “best practices” (Galiza et

al., 2009), this work’s objective was to define a

methodology that compiles the different items to collect

and analyze, so that it can be applied to other public

transport interchanges. At the same time, a major

improvement to current practices was considering periods

of 30 seconds in the performance analysis of the different

areas.

This paper starts with a summarization of the main

research developments on the pedestrian circulation on

2

interchanges, as well as the main characteristics of the

microsimulation tools, particularly the chosen one for this

work, the PTV Vissim (Chapter 2). On Chapter 3, the

proposed methodology is described, and Chapter 4

presents the analysis of the case study, which is based on

the subway station of Colégio Militar/Luz, in Lisbon.

Finally, on Chapter 5 the main conclusions and

suggestions are summarized.

2. Literature review

2.1. Pedestrian circulation on interchanges

Most of the pedestrian transfers on a public transport

journey require a walking distance which adds an

additional uncertainty to the overall time, making the

quality of the connection between the two modes a main

aspect on the public transport passengers’ satisfaction.

For the same reason, passengers usually prefer direct

journeys, to avoid that uncertainty (Ceder et al., 2013;

Guo and Wilson, 2011; Hadas and Ranjitkar, 2012;

McCord et al., 2006).

Therefore, the interchange is a place that allows the

passengers to transfer between modes and aggregates

and redistributes the pedestrian flows in order to improve

the operational efficiency of the transport system

(Domingues, 2011; IMTT, 2011; Shah et al., 2013; Sun et

al., 2010; Terzis, 1998; Yang et al., 2010; Zhang et al.,

2009).

There are several critical points on an interchange,

and Transport for London (2009) identifies three main

zones: movement spaces, opportunity spaces and

decision spaces. One of those decision spaces are the

ticket barriers, which create an “intermediate step” on

passengers’ trip and increase their journey time. Ticket

barriers represent one of the most critical points of public

transport interchanges, not only for the presence of

crowds but also because for its complexity (Davidich et

al., 2013).

One way of predicting the capacity of an interchange

is using the “fundamental diagram”, which correlates the

three main characteristics of the pedestrian movements:

flow, density and speed (Daamen et al., 2005; Davidich

and Köster, 2012).

Indeed, the pedestrian speed is of vital importance

and it depends on several aspects, both individual

characteristics (age, sex, culture, body size, health

conditions, luggage, group size, trip motive, etc.) or from

the environment (temperature, location, type of floor,

presence of obstacles or bidirectional flows and

pedestrian density, for instance) (Chattaraj et al., 2009;

Cheng et al., 2014; Fruin, 1987; Galiza and Ferreira,

2013; Galiza et al., 2009; Hoogendoorn and Daamen,

2005; Johansson et al., 2008; Kholshevnikov et al., 2008;

Koh and Zhou, 2011; Lam et al., 2003; Löhner, 2010;

Miguel, 2013; Qu et al., 2014; Rotton et al., 1990; Shah et

al., 2015, 2013; Willis et al., 2004; Yuen et al., 2013).

According to Líbano Monteiro (1994), each stream of

pedestrians should be planned to avoid crossing with

other streams. However, since it is not always possible,

pedestrians usually organize themselves “efficiently”,

through the spontaneous formation of unidirectional

“stripes” (Helbing et al., 2005; Johansson et al., 2008;

Moussaïd et al., 2011; Singh et al., 2009). Alternatively,

several types of physical obstacles could be used as

dividers, like plants, barriers, columns or furniture

(Helbing et al., 2005; Seriani and Fernández, 2015).

2.2. Pedestrian circulation models

Explanatory models of human behavior during

circulation improved significantly during the second half of

the twentieth century. There are two main types of

models: the macroscopic models, that focus on the

analysis of the observable dimensions (like density and

flow), and the microscopic ones, focused on modeling the

individual behavior of each pedestrian (Brščić et al., 2014;

Zanlungo et al., 2011).

On the side of microscopic models, Helbing and

Molnár (1995) were responsible for the “social force

model”. Despite its relative simplicity, this model

reproduces quite accurately the dynamics of pedestrians

(Johansson et al., 2008) and corrected several

weaknesses of other models (Helbing and Molnár, 1998).

According to the “social force model”, the pedestrian

desired speed is similar to a “social force”, which

represents the effect of the surrounding environment

(other pedestrians or obstacles) on the pedestrian

behavior and will cause the motivation to act (Figure 1).

Figure 1 – Schematic representation of processes leading to

behavioral changes. Source: adapted from Helbing and Molnár (1995)

Much of the success of this model is due to its ability

to reproduce some of the “self-organization” phenomena

observed in a real context, as the formation of

unidirectional “stripes” (Davidich et al., 2013; Helbing et

al., 2007, 2005). Since its introduction, the “social force

model” has been adapted either by the authors

themselves or by other researchers, namely to include the

3

effect of group walking, collision prediction or different

speeds (Helbing et al., 2000; Johansson et al., 2008;

Kretz, 2015; Kretz et al., 2011; Moussaïd et al., 2010;

Zanlungo et al., 2011).

2.3. Quality assessment and performance

analysis of interchanges

The assessment of the quality of interchanges have

been studied in the past twenty years, not only by several

academic researchers but also by the European

Commission, funding several research projects like

PIRATE (Promoting Interchange Rationale, Accessibility

and Transfer Efficiency), GUIDE (Group for Urban

Interchanges Development and Evaluation) or MIMIC

(Mobility, InterModality and InterChanges).

Regarding a quantitative performance analysis, the

“level of service” is still the most used indicator (Galiza et

al., 2009). Fruin (1987) was responsible for the used

reference values for sidewalks, corridors, stairways and

waiting areas, that depend on pedestrian space, density,

flow per unit width and average speed (TRB, 2003).

According to De Gersigny et al. (2010), the level of

service should be at least “C” at corridors and waiting

areas, while it should reach “D” at stairways.

2.4. Pedestrian microsimulation

Pedestrian micro-simulation has become a very

powerful tool in the past few years, although it requires

higher computational resources (Kretz et al., 2011). Its

success is due to the ability to simulate the behavior of

each individual agent of the system, thus the simulation is

influenced by the interaction of all the agents and results

in a more accurate and realistic representation (Bandini et

al., 2014; Fernandes et al., 2013; Helbing et al., 2005).

There are several software products that allow

pedestrian microsimulation, but Vissim stands out from

the others, as it was the first professional tool to enable

the simulation of pedestrians and vehicles simultaneously,

as well as the interactions between them (Bönisch and

Kretz, 2009; Cortés et al., 2010). Since 2008, Vissim

(through its ‘Viswalk’ module) uses the “social force

model” from Helbing e Molnár (1995), which allows the

definition of different types of pedestrian and their own

characteristics (speed, body size, etc.) but requires high

amounts of data, time and highly accurate calibration and

validation (Fellendorf and Vortisch, 2010; Fernandes et

al., 2013; Galiza and Ferreira, 2013; Galiza et al., 2009;

PTV, 2013).

3. A performance analysis methodology

using microsimulation

Even though different different performance analysis

of interchanges have been conducted for decades using

the work of Fruin (1987), there is still a lack of guidelines

regarding the use of microsimulation for this purpose

(Galiza et al., 2009).

Therefore, a methodology that could be replicated in

different transport interchanges where comparable results

were needed was established. For that, the

recommendations of many researchers of the subject

were followed, but there were also consideration for the

specifics of the Portuguese reality and the available data.

Galiza et al. (2009) suggest four key steps on these

types of analysis (Figure 2): first, the correct definition of

the scope of the work; secondly, the collection of all the

required data for the correct characterization of the “base

scenario”; as a third step, modeling that “base scenario”;

and, finally, studying of the desired changes in

infrastructure, demand or supply.

When choosing the scope of the work and its main

objectives, the necessary inputs for the correct

characterization and modeling of the problem should be

defined, as well as the methodology and the type of

output needed (Galiza et al., 2009).

It may be necessary to consider an extensive

collection of data in order to “feed” the microsimulation

model. For instance, to use the correct demand of the

infrastructure, the actual ticketing systems (which allow

access to the information in real time), video surveillance

cameras and passenger counts are some of the

possibilities.

To measure the pedestrian walking speed, the

tracking technique (Zacharias, 1993) can be used.

However, other elements like the number of passengers

in each ticket booth (and comparing it to the total number

of passengers), the time spent to pass the ticket barrier,

number of people waiting in the line to buy a ticket, the

number of passengers with bags or children and the

group composition are also useful to build the

microsimulation model (Fernández et al., 2010).

The modeling of the public transport supply is also

essential to ensure that the correct flows are modeled,

considering the real patterns of arrivals and departures.

It is important to remember that any kind of data

collection must consider “typical days”, which means that

days with severe seasonal effects, special events or

incidents should be discarded (Fruin, 1987).

It is also necessary to ensure that the calibration of

the microsimulation model is extremely accurate so that it

can be used to obtain reliable estimates (Galiza et al.,

2009).

4

Figure 2 – Schematic representation of the performance analysis methodology using microsimulation.

Fonte: adapted from Galiza et al. (2009)

The validation should ensure a good match between

the model and the modeled system (Teknomo et al.,

2006). To evaluate the similarity between the modeled

flows and the real flows, Galiza et al. (2009) recommend

a difference no higher than 5% regarding the sum of all

flows of the modeled interchange. The authors also

recommend the use of the GEH indicator, considering that

at least 85% of the values should be less than 5 and there

is no GEH higher than 10 (DMRB, 1991). According to the

rules of the Federal Highway Administration for road

microsimulation models (FHWA, 2004), GEH should also

be lower than 4 for the sum of all flows.

To validate the model, other aspects can be used:

time between two points in the interchange, percentage of

pedestrians in each ticket barrier, number of passengers

in queue, for instance.

Modeling an interchange with microsimulation tools

allows for a closer view of the pedestrians’ movement and

the perception of the constraints that really affect them.

Additionally, it is possible to obtain from the model the

pedestrian density (in pedestrians/m²), and, consequently,

the level of service. However, the usual analysis often

focus on the hourly average level of service, which will

lead to more favorable results and eliminates many of the

advantages of using a microsimulation model.

For that reason, it is suggested that the assessment

should be based on two different criteria:

Hourly average level of service;

Number of 30-second periods during which the

level of service is worse than the acceptable

minimum.

The second point is based on the one used to

calculate the design hourly volume (DHV) of a designed

road. The volume used is not the “busiest” hour, but the

30th most loaded of the year (the volume that is exceeded

29 times during the year). Above this value, the volume of

traffic rises exponentially (the slope of the curve changes

noticeably), so the DHV represents a kind of compromise

between economic and operational aspects (AASHTO,

2001; CCDR-N, 2008).

Depending on the place on the station and on the

station itself, the number of periods for which it is

“affordable” to have worse levels of service should be

different. For instance, within the zone of the ticket

barriers, when the level of service is worse than “C”

during more than 30 seconds, it is likely to exist an

inefficient situation.

An alternative way of assessing the efficiency of the

interchange should be the measurement of the delay

experienced by the passengers, based on the

measurement of the unconstrained travel time and the

real travel time inside the interchange; this should be a

more useful comparison between alternative scenarios.

Finally, alternatives to the “base scenario” should be

tested further, particularly when its performance is not

good. However, even with a satisfactory result, the

response of the infrastructure to demand growths should

be assessed.

The alternative scenarios can also be related to

changes on the supply side or changes in the

infrastructure, like a change in the location or the number

of ticket barriers, kiosks or shops, vending machines or

customer service centers. There also might be the

opportunity to test physical obstacles to guide passenger

flows more efficiently (Nai et al., 2012) or the creation of

escalators or moving walkways.

4. Case study analysis: the Colégio Militar

interchange

To apply the suggested methodology and test different

solutions, the interchange of Colégio Militar, in Lisbon,

was chosen as case study.

This is an intermodal interchange where the blue line

of Metropolitano de Lisboa (the local subway system)

connects to Santa Apolónia rail station and Amadora Este

(outside Lisbon), as seen on Figure 3, and it has also

several urban and suburban bus operators services

(Carris, Rodoviária de Lisboa and Lisboa Transportes).

Figure 3 – The location of Colégio Militar interchange.

Source: Metropolitano de Lisboa

The subway operates approximately between

6:30 a.m. and 1:00 a.m. the next day, with intervals of 5

minutes (during peak periods) to 12 minutes (at the

5

evening). The three bus operators have a total number of

21 lines at this station.

It has nine access points, four of them from the bus

station and one with internal connection to Colombo

shopping center. On the inside, the station has a central

lobby that allows the access to the boarding platforms

through 22 ticket barriers (Figure 4): the “base schema”

consists in 8 barriers dedicated to incoming passengers

and 14 dedicated to exiting passengers (which include the

two “special” barriers, dedicated to passengers with

reduced mobility or with large volumes, which are

automatically reversible).

Figure 4 – Ticket barriers in Colégio Militar/Luz subway station.

4.1. Data collection

To help modeling the infrastructure, the subway

company (Metropolitano de Lisboa) provided the detailed

blueprints of the station. Additionally, a few other

dimensions that were not available were collected that

were not available yet (e.g., stairs’ steps and ticket

barriers dimensions). Additionally, different elements that

could become obstacles – like ticket booths, shops,

kiosks, vending machines and hawkers – were identified.

The ticketing data of a normal week of 2014 was also

requested to the Metropolitan Transport Authority of

Lisbon (AMTL, Autoridade Metropolitana de Transportes

de Lisboa). In other words, it corresponds to a week of

ticket validations on the subway station of Colégio

Militar/Luz, as well as on the other stations of the subway

network and on the buses that pass by the station.

In order to represent the passengers’ movement

inside the station and attend to the demand oscillations

inside the rush hour, the peak hours were refined to 15-

minutes periods, which allows to more clearly identify the

busiest periods. Thus, according to the ticketing data, the

morning peak hour takes place between 8:30 a.m. and

9:30 a.m., while the afternoon peak hours occurs between

6:00 p.m. and 7:00 p.m. (considering both incoming and

exiting passengers).

Some additional corrections were made to the

ticketing volumes, namely an increase of 1.2% on the

demand between 2014 and 2015 and an additional

volume of 8.1% due to the estimated fraud (both values

taken from Metropolitano de Lisboa official documents)

(ML, 2015). Some assumptions have also been made,

namely a maximum transfer time (time interval between

ticket validations) of 60 minutes and some compatibility

regarding a group of tickets that were not included in the

data received.

The total number of passengers considering their

previous or following journey leg is shown on Table 1.

Table 1 – Estimated demand at the station on morning and afternoon peak hours.

Movement 8:30 – 9:30 a.m. 6:00 – 7:00 p.m.

288 180

16 25

23 75

572 1.480

169 329

25 107

45 150

960 1.338

Source: Autoridade Metropolitana de Transportes de Lisboa

The passengers’ transfers were distributed through

the entries and exits nearest to the bus stops. However, in

addition to the ticketing data received, a plan of data

collection in the station was necessary, since most of the

passengers did not have a previous or subsequent leg in

their trips. The data collection was carried out in May

2015 and included:

a) Passenger counts on the “decision points”

(junctions) of the station;

b) Pedestrian speeds measurement;

c) Pedestrian passing times through the ticket

barriers;

d) Pedestrian waiting times at the ticket booths and

ATM.

The sampling counts were carried out for 10 to 20

minutes at each point during the two peak hours, and

proved the strong relationship of the station with the

Colombo shopping center, both in the morning peak hour

and the afternoon peak hour. The calculated distributions

were later applied to the volumes for which the entry or

exit was unknown.

The walking speed was measured from

96 passengers in free flow, using the tracking technique in

a straight path (with about 32 meters long), recording the

sex and apparent age of the observed passenger. The

result of this measurement revealed an average speed of

1.38 m/s (Figure 5), not much far from Fruin (1987), Willis

et al. (2004) or Costa (2010) observations. The time spent

to cross the ticket barriers was measured for 224

passengers, which resulted in an average time 8% lower

6

on exit movements comparing to entry movements

(Figure 6).

Additionally, the time spent on the ticket booths and

ATM machines and the number of passengers in the

ticket booth line were also measured, as well as all the

“intermediate movements” inside the station (e.g., going

to the coffee or to the ticket booth) and the dwell time of

the trains.

Figure 5 – Measured walking speed of the passengers.

Figure 6 – Measured time spent to cross the ticket barriers.

4.2. “Base scenario” model

The “base scenario” model consists of 134 areas

(corridors or waiting areas), 208 obstacles (walls, columns

or barriers, as seen on Figure 7) and 21 stairways

(between the lobby and the platform, and between the

station and the surface). The modeling of the station was

limited to the entrance/exit of the subway station, but it

included all the “intermediate movements” of the

passengers in an attempt to ensure the highest possible

realism.

Based on the ticketing data, the flows for each

movement in the 15-minutes periods of the peak hours

were modeled, while the subway services were created

with the frequency of arrivals described by Metropolitano

de Lisboa (with a random variation to ensure the

existence of subways arriving to the station in both

directions). The measured walking speeds were

introduced in the software for the circulation in corridors,

while for the stairs the values obtained by Kretz et al.

(2008) were used.

Regarding some specific aspects of modeling, the

flows at the station were represented in two different

ways: the distributions obtained from the data collection

and the ticketing data were introduced as “pedestrian

static routes”, while the choice of the ticket barrier to

enter/exit the station was modeled as “pedestrian partial

routes”, letting the model decide which was the path (or

barrier) with lower delay in each moment. In each barrier

it was included also the time spent by the passengers to

cross it.

For the calibration and validation tasks, several values

for the main parameters of the “social force model” were

tested, and the final solution consists on smaller values of

ReactToN, Asocial, Bsocial, Asocial,Iso and Bsocial,Iso (Table

2), the ones that directly influence the distance between

pedestrians, which seemed to be high with the original

parameters. All the modeled routes have their “dynamic

potential” activated, with the default Vissim parameters.

Additionally, the time step of the model was switched to

0.2 s.

To validate the model, four key indicators have been

chosen:

Passengers flows in each entrance and exit of

the station;

Walking speed in the measured section;

Passenger “share” in each group of ticket

barriers;

Visual comparison between the observed

situation and the microsimulation model (namely

the reproduction of some pedestrian behaviors).

Since the model requires a warm up period (during

which the station is not yet with the real flows), the

analysis period was chosen to be the intermediate 30-

minute period of the simulation hour (between the

seconds 900 and 2,700 of the simulation).

Table 2 – Parameters of the social force model used.

Parameter Value

𝛕 0,4 s

𝛌 0,176

𝐑𝐞𝐚𝐜𝐭𝐓𝐨𝐍 5 pedestrians

𝐀𝐬𝐨𝐜𝐢𝐚𝐥 0,2 m/s²

𝐁𝐬𝐨𝐜𝐢𝐚𝐥 2,0 m

𝐀𝐬𝐨𝐜𝐢𝐚𝐥,𝐈𝐬𝐨 2,0 m/s²

𝐁𝐬𝐨𝐜𝐢𝐚𝐥,𝐈𝐬𝐨 0,1 m

𝐕𝐃 3 s

The comparison between the pedestrian flows

modeled and the real ones revealed a GEH lower than 5

in all the paths, as well as a GEH lower than 4 in the total

volume. With regard to the walking speed, the individual

values get from each pedestrian showed that

approximately 98% of the pedestrians in the morning

peak hour and 89% in the afternoon peak hour revealed a

walking speed 5% smaller than the “desired speed”.

Passengers’ “choices” in the ticket barriers zones

were also compared with the ticketing data, with good

results for several groups of barriers in the morning peak

hour. In the afternoon peak hour, where the “dynamic

potential” is more “requested”, passengers do greater

detours, due to a more congested situation.

7

Finally, some visual comparisons were also taken out,

for both modeled periods. Some of the intended behaviors

were, in fact, observed in the model, including the queue

formation around the ticket barriers and the number of

passengers in ticket booths.

For the performance analysis, the level of service was

calculated for a set of areas (circulation areas or waiting

areas), presented in Figure 8. The circulation areas never

reach a level of service lower than “C” in any moment of

the two peak periods, even though there are some

congested moments due to the “platooned’ exit of the

passengers; the only occurrence of a level of service “C”

is during the afternoon peak hour, and only in the

corridors that lead to the Colombo shopping center.

In the waiting areas, and during the morning peak

hour, only one point reaches an average level of service

of “D”, even if during some moments the level of service

reaches the levels “E” or “F”. In the afternoon peak hour,

several waiting areas near the ticket barriers present an

average level of service of “D”, which reveals a higher

probability of longer queues.

The model data was grouped in periods of 30

consecutive seconds and ordered by the growth of

pedestrian density from one period to the other (60

periods of 30 seconds). According to Figure 9, after the

55th most loaded period the growth of densities is

exponentially higher; for this reason, it was set as limit the

existence of more than 5 periods of (at least) 30 seconds

in the analyzed 30 minutes in which the level of service

was “D” or worst.

With this criteria, and within the afternoon peak hour,

the existence of at least 5 periods of 30 seconds with

level of service of “D” or worse were identified, located in

the entry ticket barriers nearer to the Colombo shopping

center. In the exit direction, there were also several

barriers that presented a level of service lower than “C”. In

the morning peak hour, the clearance time is always

smaller and no serious problems were found.

Figure 7 – Ticket barriers in the microsimulation model.

Figure 8 – Circulation areas and waiting areas analyzed in the “base scenario”.

8

Figure 9 – 30-second periods ordered by the growth of the pedestrian density.

4.3. Scenario analysis

Since this is an interchange with more than 25 years, the

ability to change the infrastructure is very small. For that

reason, 4 alternative scenarios regarding other type of

changes were chosen, based on the “base scenario”:

Scenarios A1/A2: changes in the layout of the ticket

barriers zones;

Scenario B: creation of new kiosks in the corridors;

Scenario C: creation of routing systems that ensure

unidirectional flows.

Additionally, it was also tested an increase in the

demand in the “base scenario” and in the other four

alternative scenarios.

Changes in the layout of the ticket barriers

zones

The areas close to the ticket barriers are where more

congested periods occur. For that reason, alternative

solutions should be studied, in order to improve the overall

performance of the station.

A feasible solution could be to simply increase the

number of barriers; however, as it currently happens, the

number would be highly oversized for the off-peak periods.

For that reason, different solutions with lowers levels of

investment were decided to be tested first.

In a first scenario (scenario A1), a solution was tested

where only the direction of the ticket barriers was changed –

all the barriers from the same group should be only entry or

exit barriers. In a second scenario, the ticketing system was

changed to an “open system”, which means that the ticket

barriers could be totally removed, with the passenger

registration done with video surveillance and frequent

inspection operations or with new upcoming technologies

(scenario A2).

Scenario A1 – Changes in the direction of the ticket

barriers

This solution demands, in some cases, greater routes to

some passengers but largely eliminates some of the major

constraints caused by the bidirectional flows in the ticket

barriers areas.

The analysis carried out for both peak periods showed

that the average level of service was never worse than “B”,

although some levels of service of “D”/”E”/”F” were

registered during some moments. The average delay time

per passenger on their exit from the station decreases about

10% in the morning peak hour and 14% in the afternoon

peak hour with this new solution.

Scenario A2 – Elimination of the ticket barriers

(“open ticketing system”)

Although is not common to have “open systems” in

heavy transport modes (like subway or rail), the existing

technologies nowadays make this possible even for large

passenger volumes.

So, in the tested scenario, it was decided to maintain the

existing delimitation of the central lobby (where the

information about the subway service is placed), removing

just the ticket barriers. However, in an “open system”, the

elements that exist in the central lobby (ticket booths or

information about the service) could be rearranged.

The results corroborate the advantages of this type of

system: the average delay time of the passengers is

reduced by more than 40% in both peak periods.

Creation of new kiosks in the corridors

Given the inexistence of major problems in the corridors

of the station, there is the possibility of creating new kiosks

in order to generate additional revenues to the infrastructure

manager (in this case, Metropolitano de Lisboa).

Therefore, scenario B consists of the placement of two

kiosks of different sizes in two different corridors where the

level of service is less favorable in the “base scenario”

(corridors near the Colombo shopping center).

Both in the morning peak hour and in the afternoon peak

hour, only slights impacts were seen in these areas. In no

case the minimum level of service is worse than “C”, so the

placement of this new commercial equipment would be

possible. It should be noted, however, that the possible

reductions of space due to the queues in these places were

not considered.

9

Figure 10 – Corridor connecting the subway station to the bus station in scenario C.

Creation of routing systems that ensure

unidirectional flows

Although the corridors did not present greater constraints

in any of the peak periods of the “base scenario”, the

unidirectional volumes encourage the use of physical

elements to separate the different flows: in the morning peak

hour, some corridors present a distribution of 83%/17%,

while in the afternoon peak hour the distribution is 56%/44%

(but the flow is three times higher).

So, in scenario C, the impact of physical separators that

force the passengers to follow on a one-way corridor (Figure

10) was evaluated. The comparison between the average

delay times of the passengers revealed that no substantial

gains were obtained from the placement of those elements

(mostly due to the additional detours that some passengers

have to realize).

Demand increase

As a complement to the previous scenarios, the impact

of an increase in the demand was also studied, since it is

most likely to happen in the next years: the blue line is set to

be expanded in 2020, and this station will work as a

bifurcation, so its importance is going to increase. On the

other hand, two new office towers above the Colombo

shopping center are planned for construction, which should

also increase the demand of this station.

Therefore, the “base scenario” and the four alternative

scenarios were tested to an overall increase of 20% in the

demand of the afternoon peak hour (since it is clearly the

most congested one) and the results are summarized below:

In the “base scenario”, several areas present an

average level of service of “E”, and this indicator

reaches “D”, “E” or “F” in almost 50% of the 30-

second periods;

In scenario A1, some areas present a higher

number of levels of service of “D” comparing to the

current demand, and within longer periods;

In scenario A2, the station “responds” well to the

demand increase, with only slight increases on the

average delay of each passenger;

In scenario B, the levels of service remain almost

unchanged, proving that the new commercial kiosks

could be put even with higher demands;

Finally, in scenario C, the average delay increases

12% when compared to the current demand and

the new physical elements remain unjustified.

5. Conclusions

The analyses carried out for the Colégio Militar/Luz

subway station allow for a quick perception of the potential

of microsimulation tools applied to the performance analysis

of public transport interchanges.

The different analyses showed the existence of moments

where the level of service gets worse than “C”; however, in

average terms it would not be possible to detect those

moments or to figure out their extension. For that reason, the

consideration of the 30-second periods as suggested in the

methodology allows a better knowledge of the real

performance of an analyzed infrastructure.

The tested changes in the different scenarios provide

significant reductions in the average delay times of the

passengers and an improvement on their satisfaction

regarding the walking conditions through the station. The

tested demand increase also proves that, even if the present

situation is not perfect, it could get worse if no intervention is

done in this interchange.

The observed constraints are similar to many others that

occur in other subway stations in Lisbon and in the rest of

the world. Therefore, the suggested methodology aims to

become a contribution to the discussion on the use of

microsimulation models on interchanges analysis and the

benefits of these models.

However, deeper research should be done in order to

develop a more consistent and systematic approach

regarding to the consideration of the 30-second periods in

the performance analysis of the level of service. The criteria

used in this case was obtained for the current demand and

the specific number of arrivals observed in Colégio

Militar/Luz, so it is important to build a richer database in

order to apply this methodology to other interchanges.

Also, it is worth keeping in mind that Lam and Cheung

(2000) recommend caution in the application of international

10

examples, since different interventions or reference values

should be adjusted to each reality, which means that results

can't be directly transferred. For instance, the analyzed case

study does not seem to recommend the use of routing

systems to ensure unidirectional flows (maybe due to its

demand volumes); however, this technique is applied all

over the world and recognized as a contributor to

performance improvements (Helbing et al., 2005; Seriani

and Fernández, 2015).

There are some limitations that these tools will not be

able to avoid (the erratic behavior of tourists or unfamiliar

passengers, for instance), but those particular cases

generally represent a very small number of situations and

they are not evaluated with current methodologies. The

phenomena of walking in groups is also one type of behavior

that is not considered in current microsimulation tools, but

may be represented in upcoming tools, increasing the quality

of this kind of analysis.

Finally, a full cooperation between operators and

transport authorities is essential to an effective improvement

of the interchanges performance.

References

AASHTO, 2001. A Policy on Geometric Design of Highways and Streets.

American Association of State Highway and Transportation Officials,

Washington, D.C. doi:10.1029/2004JC002361

Bandini, S., Mondini, M., Vizzari, G., 2014. Modelling negative interactions

among pedestrians in high density situations. Transp. Res. Part C Emerg.

Technol. 40, 251–270. doi:10.1016/j.trc.2013.12.007

Bönisch, C., Kretz, T., 2009. Simulation of Pedestrians Crossing a Street 1–9.

Brščić, D., Zanlungo, F., Kanda, T., 2014. Density and Velocity Patterns

during One Year of Pedestrian Tracking. Transp. Res. Procedia 2, 77–86.

doi:10.1016/j.trpro.2014.09.011

CCDR-N, 2008. Correntes de Tráfego, in: Manual de Planeamento Das

Acessibilidades E Da Gestão Viária. Comissão de Coordenação e

Desenvolvimento Regional do Norte.

Ceder, A. (Avi), Chowdhury, S., Taghipouran, N., Olsen, J., 2013. Modelling

public-transport users’ behaviour at connection point. Transp. Policy 27,

112–122. doi:10.1016/j.tranpol.2013.01.002

Chattaraj, U., Seyfried, A., Chakroborty, P., 2009. Comparison of Pedestrian

Fundamental Diagram Across Cultures 12.

doi:10.1142/S0219525909002209

Cheng, L., Yarlagadda, R., Fookes, C., Yarlagadda, P.K.D. V, 2014. A review

of methodologies behaviours in modelling pedestrian group. World J.

Mech. Eng. 1, 2–13.

Cortés, C.E., Burgos, V., Fernández, R., 2010. Modelling passengers, buses

and stops in traffic microsimulation: review and extensions. J. Adv.

Transp. 44, 72–88. doi:10.1002/atr.110

Costa, M., 2010. Interpersonal distances in group walking. J. Nonverbal

Behav. 34, 15–26. doi:10.1007/s10919-009-0077-y

Daamen, W., Hoogendoorn, S.P., Bovy, P.H.L., 2005. First-Order Pedestrian

Traffic Flow Theory. Transp. Res. Rec. 1934, 43–52. doi:10.3141/1934-

05

Davidich, M., Geiss, F., Mayer, H.G., Pfaffinger, A., Royer, C., 2013. Waiting

zones for realistic modelling of pedestrian dynamics: A case study using

two major German railway stations as examples. Transp. Res. Part C

Emerg. Technol. 37, 210–222. doi:10.1016/j.trc.2013.02.016

Davidich, M., Köster, G., 2012. Towards automatic and robust adjustment of

human behavioral parameters in a pedestrian stream model to measured

data. Saf. Sci. 50, 1253–1260. doi:10.1016/j.ssci.2011.12.024

De Abreu e Silva, J., Bazrafshan, H., 2013. User Satisfaction of Intermodal

Transfer Facilities in Lisbon, Portugal. Transp. Res. Rec. 2350, 102–110.

doi:10.3141/2350-12

De Gersigny, M.R., Hermant, L.F.L., Hermann, R., Ahuja, R., 2010. Applying

Microscopic Pedestrian Simulation To the Design Assessment of Various

Railway Stations in South Africa, in: 29th Annual Southern African

Transport Conference. Pretoria, South Africa, pp. 334–344.

DMRB, 1991. Design Manual for Roads and Bridges (Volume 12). Highways

England, London.

Domingues, M.T.S. de A., 2011. Século XXI - A Mobilidade passa pelo

Interface. Transp. em Rev. n.o 102.

Fellendorf, M., Vortisch, P., 2010. Microscopic Traffic Flow Simulator VISSIM,

in: Fundamentals of Traffic Simulation. Jaume Barceló, pp. 63–93.

doi:10.1007/978-1-4419-6142-6

Fernandes, T., Remédio, A., Correia, G., 2013. Micro-simulação de Veículos

e Peões na Avaliação do Impacte da Ocorrência de Eventos de Grande

Procura, in: 7o Congresso Rodoviário Português - “Novos Desafios Para

a Atividade Rodoviária.”

Fernández, R., Seriani, S., Allard, P., 2010. Pedestrian microsimulation of

Metro-Bus interchanges. A case study in Santiago de Chile, in: European

Transport Conference. Glasgow, Scotland, UK.

FHWA, 2004. Traffic Analysis Toolbox Volume III: Guidelines for Applying

Traffic Microsimulation Modeling Software. US Department of

Transportation - Federal Highway Administration, Oakland, CA.

Fruin, J.J., 1987. Pedestrian Planning and Design, Revised Edition. Elevator

World.

Galiza, R., Ferreira, L., 2013. A methodology for determining equivalent

factors in heterogeneous pedestrian flows. Comput. Environ. Urban Syst.

39, 162–171. doi:10.1016/j.compenvurbsys.2012.08.003

Galiza, R., Kim, I., Ferreira, L., Laufer, J., 2009. Modelling Pedestrian

Circulation in Rail Transit Stations Using Micro-Simulation. pp. 1–24.

Guo, Z., Wilson, N., 2011. Assessing the cost of transfer inconvenience in

public transport systems: A case study of the London Underground.

Transp. Res. Part A Policy Pract. 45, 91–104.

doi:10.1016/j.tra.2010.11.002

Hadas, Y., Ranjitkar, P., 2012. Modeling public-transit connectivity with spatial

quality-of-transfer measurements. J. Transp. Geogr. 22, 137–147.

doi:10.1016/j.jtrangeo.2011.12.003

Helbing, D., Buzna, L., Johansson, A., Werner, T., 2005. Self-organized

pedestrian crowd dynamics: Experiments, simulations, and design

solutions. Transp. Sci. 39, 1–24. doi:10.1287/trsc.1040.0108

Helbing, D., Farkas, I., Vicsek, T., 2000. Simulating dynamical features of

escape panic. Nature 407, 487–490.

Helbing, D., Johansson, A., Al-Abideen, H.Z., 2007. Dynamics of crowd

disasters: An empirical study. Phys. Rev. E - Stat. Nonlinear, Soft Matter

Phys. 75, 1–7. doi:10.1103/PhysRevE.75.046109

Helbing, D., Molnár, P., 1995. Social force model for pedestrian dynamics.

Phys. Rev. E. doi:10.1103/PhysRevE.51.4282

Helbing, D., Molnár, P., 1998. Self-Organization Phenomena in Pedestrian

Crowds 569–577. doi:10.1068/b2697

Hoogendoorn, S.P., Daamen, W., 2005. Pedestrian Behavior at Bottlenecks.

Transp. Sci. 39, 147–159. doi:10.1287/trsc.1040.0102

IMTT, 2011. Guia para a Elaboração de Planos de Mobilidade e Transportes.

Instituto de Mobilidade e dos Transportes Terrestres, I.P.

11

Johansson, A., Helbing, D., Shukla, P.K., 2008. Specification of a Microscopic

Pedestrian Model by Evolutionary Adjustment to Video Tracking Data.

Adv. Complex Syst.

Kalakou, S., Moura, F., 2014. Bridging the Gap in Planning Indoor Pedestrian

Facilities. Transp. Rev. 34, 474–500.

doi:10.1080/01441647.2014.915441

Kholshevnikov, V. V., Shields, T.J., Boyce, K.E., Samoshin, D. a., 2008.

Recent developments in pedestrian flow theory and research in Russia.

Fire Saf. J. 43, 108–118. doi:10.1016/j.firesaf.2007.05.005

Koh, W.L., Zhou, S., 2011. Modeling and simulation of pedestrian behaviors in

crowded places. ACM Trans. Model. Comput. Simul. 21, 1–23.

doi:10.1145/1921598.1921604

Kretz, T., 2015. On Oscillations in the Social Force Model On Oscillations in

the Social Force Model. Physica A 438, 272–285.

doi:10.1016/j.physa.2015.07.002

Kretz, T., Große, A., Hengst, S., Kautzsch, L., Pohlmann, A., Vortisch, P.,

2011. Quickest Paths in Simulations of Pedestrians. Adv. Complex Syst.

14(5), 733–759. doi:10.1142/S0219525911003281

Kretz, T., Grünebohm, A., Kessel, A., Klüpfel, H., Meyer-König, T.,

Schreckenberg, M., 2008. Upstairs walking speed distributions on a long

stairway. Saf. Sci. 46, 72–78. doi:10.1016/j.ssci.2006.10.001

Lam, W.H.K., Cheung, C., 2000. Pedestrian Speed/Flow Relationships for

Walking Facilities in Hong Kong. J. Transp. Eng. 126, 343–349.

doi:10.1061/(ASCE)0733-947X(2000)126:4(343)

Lam, W.H.K., Lee, J.Y.S., Chan, K.S., Goh, P.K., 2003. A generalised

function for modeling bi-directional flow effects on indoor walkways in

Hong Kong. Transp. Res. Part A Policy Pract. 37, 789–810.

doi:10.1016/S0965-8564(03)00058-2

Líbano Monteiro, J.P.R.S., 1994. A Qualidade nos Percursos Pedonais em

Interfaces de Transportes (Tese de Mestrado). Instituto Superior Técnico,

Lisboa.

Löhner, R., 2010. On the modeling of pedestrian motion. Appl. Math. Model.

34, 366–382. doi:10.1016/j.apm.2009.04.017

McCord, M.M., Mishalani, R.G., Wirtz, J., 2006. Passenger Wait Time

Perceptions at Bus Stops. J. Public Transp. 9, 89–106.

Miguel, A.F., 2013. The emergence of design in pedestrian dynamics:

Locomotion, self-organization, walking paths and constructal law. Phys.

Life Rev. 10, 168–190. doi:10.1016/j.plrev.2013.03.007

ML, 2015. Plano de Atividades e Orçamento 2015.

Moussaïd, M., Helbing, D., Theraulaz, G., 2011. How simple rules determine

pedestrian behavior and crowd disasters. Proc. Natl. Acad. Sci. U. S. A.

108, 6884–6888. doi:10.1073/pnas.1016507108

Moussaïd, M., Perozo, N., Garnier, S., Helbing, D., Theraulaz, G., 2010. The

walking behaviour of pedestrian social groups and its impact on crowd

dynamics. PLoS One 5, 1–7. doi:10.1371/journal.pone.0010047

Nai, W., Dong, D., Chen, S., Zheng, W., Yang, W., 2012. Optimizing the

Usage of Walking Facilities between Platform and Concourse Layer in L-

Shaped Interchange Metro Station. Procedia - Soc. Behav. Sci. 43, 748–

757. doi:10.1016/j.sbspro.2012.04.148

Nesheli, M.M., Ceder, A. (Avi), 2014. Optimal combinations of selected tactics

for public-transport transfer synchronization. Transp. Res. Part C Emerg.

Technol. 48, 491–504. doi:10.1016/j.trc.2014.09.013

PTV, 2013. PTV VISSIM 7 User Manual. PTV Planug Trasport Verker AG.

Qu, Y., Gao, Z., Xiao, Y., Li, X., 2014. Modeling the pedestrian’s movement

and simulating evacuation dynamics on stairs. Saf. Sci. 70, 189–201.

doi:10.1016/j.ssci.2014.05.016

Rotton, J., Shats, M., Standers, R., 1990. Temperature and Pedestrian

Tempo: Walking Without Awareness. Environ. Behav. 22, 650–674.

doi:10.1177/0013916590225005

Seriani, S., Fernández, R., 2015. Pedestrian traffic management of boarding

and alighting in metro stations. Transp. Res. Part C Emerg. Technol. 53,

76–92. doi:10.1016/j.trc.2015.02.003

Shah, J., Joshi, G., Parida, P., 2013. Behavioral Characteristics of Pedestrian

Flow on Stairway at Railway Station. Procedia - Soc. Behav. Sci. 104,

688–697. doi:10.1016/j.sbspro.2013.11.163

Shah, J., Joshi, G., Parida, P., Arkatkar, S., 2015. Impact of Train Schedule

on Pedestrian Movement on Stairway at Suburban Rail Transit Station in

Mumbai, India. Adv. Civ. Eng. 2015, 1–9. doi:10.1155/2015/297807

Singh, H., Arter, R., Dodd, L., Langston, P., Lester, E., Drury, J., 2009.

Modelling subgroup behaviour in crowd dynamics DEM simulation. Appl.

Math. Model. 33, 4408–4423. doi:10.1016/j.apm.2009.03.020

Sun, L., Rong, J., Yao, L., 2010. Measuring transfer efficiency of urban public

transportation terminals by data envelopment analysis. J. Urban Plan.

Dev. 136, 314–319.

Teknomo, K., Takeyama, Y., Inamura, H., 2006. Application of Microscopic

Pedestrian Simulation Model. Transp. Res. Part F Traffic Psychol. Behav.

9, 15–27.

Terzis, G., 1998. GUIDE: Group for Urban Interchanges Development and

Evaluation, in: AET European Transport Conference. pp. 117–128.

Transport for London (TfL), 2009. Interchange Best Practice Guidelines 2009

- Quick Reference Guide.

TRB, 2003. Transit Capacity and Quality of Service Manual. Transport

Research Board, Washington, D.C.

Willis, A., Gjersoe, N., Havard, C., Kerridge, J., Kukla, R., 2004. Human

movement behaviour in urban spaces: Implications for the design and

modelling of effective pedestrian environments. Environ. Plan. B Plan.

Des. 31, 805–828. doi:10.1068/b3060

Yang, L.., Jia, H.., Juan, Z.., Zhang, J.., 2010. Service level classification of

facilities in passenger terminals based on pedestrian flow characteristics

analysis, in: ICCTP 2010: Integrated Transportation Systems: Green,

Intelligent, Reliable - Proceedings of the 10th International Conference of

Chinese Transportation Professionals. pp. 2581–2589.

doi:10.1061/41127(382)275

Yuen, J.K.K., Lee, E.W.M., Lo, S.M., Yuen, R.K.K., 2013. An intelligence-

based optimization model of passenger flow in a transportation station.

IEEE Trans. Intell. Transp. Syst. 14, 1290–1300.

doi:10.1109/TITS.2013.2259482

Zacharias, J., 1993. Reconsidering the impacts of enclosed shopping centres:

A study of pedestrian behaviour and within a festival market in Montreal.

Landsc. Urban Plan. 26, 149–160. doi:http://dx.doi.org/10.1016/0169-

2046(93)90013-4

Zanlungo, F., Ikeda, T., Kanda, T., 2011. Social force model with explicit

collision prediction. EPL (Europhysics Lett. 93, 68005.

Zhang, R., Li, Z., Hong, J., Han, D., Zhao, L., 2009. Research on

characteristics of pedestrian traffic and simulation in the underground

transfer hub in Beijing, in: ICCIT 2009 - 4th International Conference on

Computer Sciences and Convergence Information Technology. pp.

1352–1357. doi:10.1109/ICCIT.2009.222