- Email: [email protected]
On passenger saturation flow in public transport doors
Transportation Research Part A 78 (2015) 102–112
Contents lists available at ScienceDirect
Transportation Research Part A journal homepage: www.elsevier.com/locate/tra
On passenger saturation flow in public transport doors Rodrigo Fernández a,⇑, Alejandra Valencia b, Sebastian Seriani a a b
Universidad de los Andes, Chile Pontificia Universidad Católica de Valparaíso, Chile
a r t i c l e
i n f o
Article history: Received 9 January 2014 Received in revised form 17 April 2015 Accepted 1 May 2015 Available online 3 June 2015 Keywords: Public transport Passengers Saturation flow Door capacity
a b s t r a c t In previous studies the authors have shown passengers’ boarding and alighting times for the Transantiago system obtained at the Pedestrian Accessibility and Movement Environment Laboratory (PAMELA) of University College London. Following this line of research, the aim of this paper is to demonstrate the existence of pedestrian saturation flows in public transport doors and show some values of this variable under different conditions. The methodology to achieve this aim was real-scale experiments performed in both PAMELA and the Human Dynamics Laboratory at Universidad de los Andes in Santiago de Chile. Different groups of people getting off a mock-up of a public transport vehicle were recorded by means of video cameras. The videos were then visually processed to find values of passenger saturation flow according to door configurations. The variables studied were the vertical gap between the platform and the vehicle chassis and the width of the door. Results indicate that it is possible to define values of passenger saturation flows for different characteristics of public transport doors. These values proved to be statistically sensitive to both the vertical gap and the width of the door. In addition, results indicate that there seems to be both a vertical gap and a door width for which the flow of passenger reaches its optimum rate. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction A main challenge for public transport is to provide a good level of service. This is focused on quantitative and qualitative characteristics of the journey. One of the most important variables that affect the level of service is dwell time. This variable is defined as the time that a public transport vehicle remains stopped while transferring passengers, including acceleration and braking time (TRB, 2000). Dwell time depends on the number of boarding and alighting passengers and how quick they carry out these tasks. The speed of passengers is determined by the fare collection method, internal layout of vehicles and the density of passengers. However as it is said in the Transit Capacity and Quality of Service Manual the most important variable that affects dwell time is the number of doors, because ‘‘the greater the number of door channels, the less time required to serve a given passenger flow’’ (TRB, 2003: 23). Passenger flow for public transport doors is defined as the number of passengers that pass through a car doorway width. The data can be partitioned into boarding, alighting or mixed flow. When this flow reaches the highest value it can be called passenger saturation flow, similar to the concept of vehicle saturation at junctions. According to Akçelik (1995), saturation flow is used at junctions as a basic characteristic to calculate the capacity of a traffic signal approach during a typical signal cycle (C). For a given junction approach, saturation flow is defined as the ⇑ Corresponding author. http://dx.doi.org/10.1016/j.tra.2015.05.001 0965-8564/Ó 2015 Elsevier Ltd. All rights reserved.
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
103
maximum discharge rate of a queue of vehicles during the effective green time (g) of that approach, as shown in Fig. 1. At the start of the green period (G) there is a transient period (L1), called start loss, before the discharge rate reaches its maximum, which is the saturation flow (S) for that approach. If the queue remains until the start of the amber time (A), there is another transient period (L2), called end gain, until the start of the red time (R). Thus, the effective green time is defined as g = G L1 + L2. The value of the saturation flow and transient periods depend on both the traffic composition and geometry of the junction approach. The histogram shown in Fig. 1 shows the discharge of vehicles during 0.1 min if the queue remains until the end of the amber time. In this example, the green plus amber time is 0.7 min or 42 s. This sort of traffic behaviour is obtained if there is no blockage downstream of the stop line; e.g., the downstream street is not blocked by the backup of vehicles which may reduce the discharge rate. The relationship between saturation (S) and capacity (Q) of a traffic signal approach is Q = uS, where u = g/C. We tried to apply a similar concept to the alighting capacity of a public transport door. In the case of a public transport door, the cycle time ‘‘C’’ can be considered as the time between the stop and the start of the vehicle at the station, and the effective green time ‘‘g’’ would be the time during which doors are opened. Therefore, if a vehicle remains a different time period at the station and/or the time during which door are opened changes, the passenger capacity of the door will be different. Which do remain constant is the passenger saturation flow, as in the case of a signalized junction. The flow–time curve shown in Fig. 1 indicates that not all vehicles in a queue take the same time to cross the stop line at a traffic signal. The question posed by this research is: Could the same behaviour occur in public transport doors? Our hypothesis is that the same sort of discharge curve of passengers may be found at public transport doors. That is, an unobstructed alighting process of a bunch of passengers through a public transport door may be similar to the unobstructed discharge of a queue of vehicles at a traffic signal. Following this hypothesis, the aim of this paper is to study the passenger saturation flow in public transport doors by mean of real experiments. The results are part of a research project funded by FONDECYT – Chile (No. 1120219) which has the objective of generating design guidelines to public transport facilities in developing countries. This includes the following specific objectives: (a) Demonstrate the concept of passenger saturation flow in public transport doors; (b) to measure passenger saturation flows in laboratory; (c) to study the effects of different physical design variables on the passenger saturation flow such as door width and vertical gap. This article is made of five sections, including this introduction. In the second chapter the literature review on boarding and alighting times as well as door capacity is summarized. Next, in chapter three the methodology for obtained passenger saturation flows is shown. Results and analysis are presented in chapter four. Finally, the conclusions of this research are provided in chapter five. 2. Literature review During 2008 and 2009 at UCL PAMELA, a real-scale mock-up of a bus and people getting on and off the mock-up were used to obtain average boarding and alighting times per passenger (Fernandez, 2011). In the 2008 experiments, at PAMELA three variables were controlled: platform height, door width, and fare collection method. In 2009 the effect of the density of passengers inside the vehicle on the average boarding and alighting times was studied. In this case, different passenger densities were considered. More than 300 video records of boarding and alighting processes were obtained in those experiments. Tables 1 and 2 show a summary of the results of the experiments obtained from at least 30 samples for each density, where l is the mean boarding or alighting time per passenger, r is the corresponding standard deviation
Flow [veh/0.1min]
Effective flow histogram
S Actual flow histogram
0.1
0.2
0.3
L1
0.4 g
G
0.5
0.6
Time [min]
0.7
L2 A
R
Fig. 1. Typical flow discharge at a traffic signal.
104
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Table 1 Boarding times per passenger, PAMELA experiments (Fernandez, 2011). Density (pass/m2)
l (s/pass)
Maximum (s/pass)
Minimum (s/pass)
r (s/pass)
C (%)
0 1 2 4 6
0.89 1.13 1.37 1.67 2.03
1.04 1.45 1.75 2.66 3.57
0.63 0.75 0.95 1.20 1.36
0.10 0.14 0.21 0.34 0.56
11 12 15 20 28
Table 2 Alighting times per passenger, PAMELA experiments (Fernandez, 2011). Density (pass/m2)
l (s/pass)
Maximum (s/pass)
Minimum (s/pass)
r (s/pass)
C (%)
0 1 2 4 6
0.59 0.92 1.11 1.89 5.92
0.66 1.17 1.46 2.53 9.73
0.50 0.80 0.94 1.26 2.06
0.04 0.08 0.13 0.33 1.70
7 9 12 17 29
and c = r/l is the corresponding coefficient of variation of the results. The mean alighting or boarding time is calculated for each respective process, dividing the total alighting or boarding time by the number of passengers that get off or on the vehicle. As can be seen in the tables, for the same passenger density inside the bus there is a range of boarding and alighting times per passenger so that c ranges from approximately 10% to 30%. Table 1 also shows that as passenger density inside the bus increases, boarding times increase almost linearly from 0.9 to 2.0 s per passenger. However, in the case of alighting times shown in Table 2 these increase almost exponentially from 0.6 to 6.0 s per passenger. It is interesting to note that passengers take longer to get on than to get off the vehicle; up to a density of 4 passengers per square metre. Above that density the situation reverses because the congestion inside the vehicle makes it more difficult for a passenger to reach the exit door. In relation to door widths and their relationship to the dwell time in trains, the Transit Capacity and Quality of Service Manual (TRB, 2003) states that between 1.14 and 1.37 m-width there is no major change in dwell time, reaching average values of 1.77 s/pass during alighting processes, 1.99 s/pass during boarding processes and 2.68 s/pass for two-way flow; i.e., boarding and alighting processes. For that range of door widths TRB (2003) states that a single stream is formed and occasionally three passengers will move through the doorway simultaneously. The same reference states that a door with steps approximately doubles the time of boarding and alighting in relation to a door without steps. Recently, Schelenz et al. (2013) stated that the number of doors can affect directly the time of boarding and alighting. This can be proved by simulation. For example, a bus with 5 doors get an average dwell time of 19.22 s; 39% shorter than a bus with 3 doors. Similarly, Schelenz et al. (2012) compared the alighting and boarding times with respect to the number of passengers and doors. For example, for a 3-door bus, when the number of passengers increases from 30 to 60, the average alighting time goes up by nearly 4 s. Also, Rexfelt et al. (2014) used a mock-up of a public transport vehicle to prove that a bus with 4 doors will have a dwell time 17% lower than a bus with 3 doors. However those studies did not consider variations of door width. Other authors (Schadschneider et al., 2008) state that the capacity of bottlenecks (like doors in a corridor) can grow with increased width. When pedestrians are formed in lanes, the capacity will be increased only if a new lane is formed. This is shown in different models (Predtechenskii and Milinskii, 1978; Hoogendoorn and Daamen, 2005; Kretz et al., 2006; Nagai et al., 2006). All these models were performed in laboratory experiments. However, the closest work related with our research is Daamen et al. (2008). They studied the boarding and alighting times with respect to different vertical and horizontal gaps and door widths in rail vehicles by performing laboratory experiments. Some of their results are shown in Table 3.
Table 3 Door capacities (pass/s) for a door width of 80 cm (Daamen et al., 2008). Vertical gap (mm)
Horizontal gap (mm) No luggage
50 200 400 600
Luggage
50
150
300
50
150
300
0.91 – 0.81 –
– 0.89 – 0.84
0.85 0.88 0.77 0.77
0.69 – 0.65 –
– 0.62 – 0.60
0.73 0.64 0.63 0.56
105
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
According to Daamen et al. (2008) the capacity of public transport doors decreases when gaps (vertical and horizontal) increase. However, for small vertical gaps, door capacity slightly increases as horizontal gap increases. They also found that capacity decreases up to 25% if passengers are carrying luggage. However, this is out of the scope of this article as we studied urban public transport conditions; i.e., passengers with no luggage. Following the work of Daamen et al. (2008), Daamen and Hoogendoorn (2010) identified the relation between door capacity and four experimental variables (door width, population, light intensity and the presence of an open door) in evacuation conditions. However this situation is out of the scope of this article. 3. Methodology We used the methods described in Road Note 34 – RN34 – (RRL, 1963) for measuring passenger saturation flows in both PAMELA and the Human Dynamics Laboratory (HDL) of Universidad de los Andes. The RN34 describes a method for measuring vehicle saturation flow at traffic signals. It consists of counting the number of vehicles crossing the stop-line in successive intervals during a saturated green period, as shown in Fig. 1. ‘‘The basis of this method is to divide the saturated portion of each green period into short intervals of time and to average the flows in those saturated intervals which are free from ‘lost time’ effects, to give a measure of the saturation flow. . .’’ (RRL, 1963:2). The RN34 also states that ‘‘. . .the method consists of recording the number of vehicles discharging from the queue in successive 0.1-min intervals... When the flow is no longer at the saturation level because the queue has disappeared on one or more lanes the recording of the flow in 0.1-min interval should be discontinued’’ (RRL, 1963: 4). As an example for understanding the RN 34 method, Table 4 shows the application at a traffic signal in Santiago de Chile. Nine green plus amber fully-saturated periods of 36 s are shown. According to the RN 34 method, each green plus amber period must be divided into 0.1-min or 6-s intervals, plus a last interval of variable length. In this example six 6-s intervals give 36-s green plus amber time. The saturation flow is calculated as the average ratio between the number of vehicles and the number of saturated green plus amber periods, excluding the first and the last interval. In this case, the saturation flow is S = (2.67 + 2.67 + 3.22 + 2.78 + 2.44)/5 = 2.76 veh/0.1-min; i.e., 1653 veh/h. This means 1653 veh/h per lane which is a typical value when there are only light vehicles but various types of movements at the junction (Akçelik, 1995: Table 5.1). Fig. 2 shows the discharge histogram at this junction, where the dotted line is the average height of the saturated intervals (intervals 2–6); in other words, the saturation flow. As can be seen in the figure, there are differences in the height of the saturated intervals and the difference between two adjacent saturated intervals is about 20%. Therefore, in each saturated period the outflow is not the same because in this particular site all movements take place (through, right, left), then due to the traffic composition or traffic movements, not all intervals contain the same number of vehicles. In order to achieve the same condition for public transport doors, unrestrained discharges of passengers must be studied. As a result, alighting processes are needed from public transport vehicles in which passengers are not obstructed by the congestion caused by them. This condition is difficult to obtain in real systems with enough of alighting demand because of the width of the platform, the size of the stop shelter or pavement width. Therefore, a controlled environment could be necessary such as a laboratory and a real-scale mock-up. This environment was provided for this study by both PAMELA and HDL. Saturation flow was measured according to the RN34 but applied to the passenger alighting process. We divided the alighting period into 2 or 3-s intervals and counted the number of passengers in each interval from the time the door is opened until all the passengers have gotten off the vehicle. Then a histogram that represents the alighting period is constructed. The width of the intervals was chosen in order to have at least 4 saturated intervals, according to the minimum subjects for experimentation (20 people). In the application of RN34 to passengers, the first interval starts when the doors start to open, because passengers jump off the vehicle as soon as they see a gap in the door. Then the average height of the saturated intervals gives the saturation
Table 4 RN34 method applied in a traffic signal in Santiago de Chile. Green plus Amber periods
1 2 3 4 5 6 7 8 9 Total Sample Average
0.1-min intervals
Last interval
1
2
3
4
5
6
Vehicles
Time (s)
1 1 1 1 1 2 1 1 1 10 9 1.11
2 3 2 4 3 2 3 2 3 24 9 2.67
3 2 3 2 2 3 3 3 3 24 9 2.67
3 2 3 3 3 4 4 4 3 29 9 3.22
2 3 2 3 3 4 3 2 3 25 9 2.78
3 3 3 3 2 1 3 2 2 22 9 2.44
2 2 1 2 3 1 1 2 2 16 9 1.78
2.00 2.83 3.33 2.00 1.00 2.33 2.50 2.00 2.25 20.24 9 2.25
106
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Saturation flow [veh/6-s]
3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00
1
2
3
4
5
6
7
6-sec intervals [s] Fig. 2. Example of a flow discharge at a traffic signal in Santiago de Chile.
flow. Intervals belong to the start loss period; and those in which there is no more saturation are not considered for the calculation. To decide when saturation ends, we observe that the height of an interval is clearly below the average height of the saturated intervals. We applied our engineering criterion and experience to take this decision. Obviously, there is no final transient because there is nothing similar to the amber period in the case of passengers. On the contrary, when everybody gets off the train, the measurement stops.
3.1. Experiments at PAMELA in University College London In our experiments at PAMELA in 2008 and 2009, we recorded both boarding and alighting processes in a constructed bus mock-up. As shown in Fig. 3, the mock-up is 10-m long by 2.5-m wide with a 1.6-m door in the middle, which can be arranged to be opened halfway. The mock-up can be raised to different heights to represent different levels of platform. In order to test our hypothesis we took only the alighting processes in which all passengers get off the vehicle. A total of 120 alighting processes were observed. For each alighting process a number of 50 passengers get off the vehicle. Passengers were male and female adult Londoners, of different ages. The instruction given to passengers was ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. The control variables we considered for PAMELA experiments were door widths and platform height. Two door widths were tested: a wide door (1600 mm) and a narrow door (800 mm). Somehow these values try to represent extreme values of door widths of public transport vehicles in Santiago de Chile; i.e., 800 mm for small buses and 1600 mm for metro cars. Similarly, three platform heights were tested: a high platform (300 mm height), a normal platform (150 mm height) and no platform (0 mm height). The first case represents a specially designed stop in which the platform has the same height than the vehicle; so the is no vertical gap for alighting passengers. The normal platform represents the case in which passengers alight into the pavement; therefore the vertical gap has 150 mm. Finally, in the no-platform case passengers have to alight to the road surface; hence the vertical gap is 300 mm, which is common in developing countries or for countryside services in any setting.
Fig. 3. Mock-up in PAMELA of University College London.
107
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
3.2. Experiments at the Human Dynamics Laboratory in Universidad de los Andes The HDL at the Universidad de los Andes was designed to study human dynamics; from parts of the body to groups of people. In these experiments, a mock-up of the hall of a public transport vehicle was built for our study. Fig. 4 shows the concept and the layout of the mock-up, and Fig. 5 shows a picture during the experiments at the HDL. In order cover any type of vehicle – from normal buses to metro cars – the mock-up of the hall is 3-m long by 2.5-m wide with a door in the middle which can be opened to 2-m wide. At both sides of the hall there is a 0.8-m doorway to represent the aisle of the vehicle. The size of the platform is variable according to the transit system to be investigated. In the HDL experiments we only tested door width as a control variable (Fernandez et al., 2013); therefore, a zero vertical gap was assumed. Seven door widths were tested: 600, 800, 1100, 1300, 1650, 1850, and 2000 mm. These numbers were chosen because they represent the different doors of public transport vehicles in Santiago de Chile: small buses (600 and 800 mm), normal and articulated buses (1100 mm), and metro cars (1300 and 1650 mm). As far as we know, there do not exist 1850 and 2000-mm doors in urban transit systems. We chose those values to see the passenger behaviour under a reality which can only be tested in a laboratory. A number of 10 alighting processes were recorded for each width, making a total of 50 observations. In each alighting process, 26 passengers get off the vehicle. Passengers were students of Universidad de los Andes, 18–25 years old, 10% female and 90% male. As in the case of PAMELA, the instruction given to passengers was ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. 4. Results and analysis 4.1. Results from PAMELA experiments Fig. 6 shows a typical histogram of an alighting process. The figure shows 21 intervals of three seconds each. We chose 3-s intervals because of the acceleration abilities of passengers are different for vehicles. If a longer interval was chosen, we did not have enough number of intervals to calculate saturation flows.
Exit hall
Entry hall
Ticket sensor
2.30 0.78
Ticket sensor
1.40
1.40
1.50
1.10
1.10
(a) Entry and exit halls of a public transport vehicle
“Vehicle” “Aisle”
“Aisle” Sliding
“Platform
Sensor
Camera
(b) Layout of the mock-up Fig. 4. Mock-up in the HDL at the Universidad de los Andes.
108
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Fig. 5. Experiments at the HDL of Universidad de los Andes.
The saturation flow obtained is 2.559 passengers in a 3-s interval – i.e., 0.853 passengers per second – for a narrow door and no vertical gap. The first interval represents the start loss and the last three intervals correspond to the end of the saturation. To decide when saturation flow ends, we observe the difference between two adjacent saturated intervals. In this case, the difference between the flow in the intervals 18 and 19 is 31% which indicates that saturation lasts until the 18th interval. In Fig. 7 the same is shown for a wide door and 150-mm vertical gap. As can be seen in the figure, saturation ends in the 9th interval, because the difference between the flow in the intervals 9 and 10 is 32%. As can be seen in the figures, the number of intervals decreases as the door widens, because more passengers can pass thought the door simultaneously. Thus, 4.405 pass/3-s interval can pass through a wide door if there is no vertical gap between the vehicle and the platform; i.e., 1.468 pass/s. Note that this is not twice the flow through a narrow door, as discussed below. Table 5 shows the summary of results of the experiments. As expected, it can be observed from the table that a wide door can reach higher values of saturation flow than a narrow door. However, if the door width doubles, the saturation flow does not increase in the same proportion. For example, for the no-gap case the increase of saturation flow between the two widths is only 72%. The same occurs for the other gaps for which doubling the door width increases the saturation flow in about 65%. This suggests the existence of an optimum width for which the passenger flow per metre reaches its maximum rate. In relation to the saturation flow expressed in passengers per second per metre, it can be observed from Table 5 that the average value obtained was a bit more than 1.0 pass/s-m (1.047). A statistical hypothesis test between the different values of saturation flow in pass/s-m was performed to decide if they are statistically equal at 5% significance level in order to recommend that average value. Thus, for each vertical gap we tested if saturation flows for an 800 and a 1600-mm door are statistically equal (the H0 hypothesis). Similarly, for each door width we tested if saturation flows for 0, 150 and 300-mm vertical gap are statistically equal. As a result, the H0 hypothesis was rejected in all cases at the 5% level. This means that passenger saturation flows in pass/s-m are all statistically different and a single average value cannot be recommended.
Door 800 mm -Gap 0 mm
Number of passengers
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
3-s intervals Frequency (pass/interval)
Saturation Flow (pass/3-s)
Fig. 6. Histogram of an alighting process for a narrow door and no vertical gap.
109
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Door 1600 mm - Gap 150 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
8
9
10
11
12
13
3-s Intervals Frequency (pass/interval)
Saturation Flow (pass/3-s)
Fig. 7. Histogram of an alighting process for a wide door and 150-mm vertical gap.
If we analyse the behaviour of the saturation flow with respect to the door width and the vertical gap, in Table 5 it can be observed that the maximum saturation flow is obtained for a wide door and a medium vertical gap. In relation to the vertical gap, results suggest the existence of an optimum vertical gap for which the passenger’s saturation flow reaches its maximum value. After watching the videos, a possible explanation is that for 150 mm or less, people do a little ‘‘jump’’ to get off the vehicle, which may speed up the alighting process. On this issue, Tyler (2002: 149) states that ‘‘a vertical gap of 75 mm could be tolerated if the horizontal gap were 25 mm. However, if the horizontal gap is more than 50 mm, the vertical gap can be no more than 50 mm. In no case should either gap be more than 100 mm.’’Our results show that a 150-mm as the ‘‘best’’ vertical gap, because we did not test gaps between 150 and 0 mm. Further experiments at the HDL of Universidad de los Andes may provide a deeper understanding on this issue. Comparing the above results with Daamen et al. (2008) for an 80-cm door width in Table 3, it can be seen that there are similarities. For example, Daamen et al. (2008) arrived at a door capacity between 0.85 and 0.91 pass/s for a 50-mm vertical gap, which is similar to our result of 0.853 pass/s for no vertical gap. Table 3 also shows that for a 200-mm vertical gap, door capacity reaches a value between 0.88 and 0.89 pass/s, which is similar to our result of 0.90 pass/s for a 150-mm vertical gap. What is relevant is that in both studies capacity slightly increased for an intermediate vertical gap: 200 mm in the case of Daamen et al. (2008) and 150 mm in our case.
4.2. Results from the HDL experiments Experiments at the HDL of Universidad de los Andes show similar histograms of the alighting processes to those found in PAMELA. For example, Fig. 8 shows 4 intervals of three second each for which the saturation flow is 4.389 passengers per 3-s interval or 1.467 passengers per second for an 800-mm door. The first interval represents the start loss and the last two ones are the end of the saturation, because the flow difference between the 5th and 6th intervals is 24%. In Fig. 9 the same is shown for a 1300-mm door. In this case, the saturation flow is 5.2 passengers per 2-s interval; i.e., 2.6 passengers per second. In this case, the flow difference between the 5th and 6th intervals is 66%. It was observed that if 1-s intervals are used, the discharge histogram is constructed of more samples (8 intervals). However, the value of the passenger saturation would be the same, because the start loss last for 2 intervals and the saturation period ends in the 7th interval. Therefore, the saturation period remains for 4 intervals, same as for 3-s intervals or 2-s intervals. It should be noted that the saturation flows found at the HDL are greater than those found at PAMELA for the same vertical gap. For example, for the 800-mm door we measured a saturation flow of 1.467 pass/s in the HDL versus 0.853 pass/s in PAMELA; that is, a 72% higher value. We do not have an ultimate explanation for this difference. However, possible
Table 5 Values of passenger saturation flow for the study cases in PAMELA. Door width (mm)
Vertical gap (mm)
Saturation flow (pass/s)
Saturation flow (pass/s-m)
800 800 800 1600 1600 1600
0 150 300 0 150 300
0.853 0.991 0.900 1.468 1.610 1.481
1.066 1.239 1.125 0.918 1.007 0.926
110
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Door 800 mm - Gap 0 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
3-sec intervals Frequency (pass/interval) Fig. 8. Histogram of an alighting process for an 800-mm door and no vertical gap.
Door 1300 mm -Gap 0 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
2-sec intervals Frequency (pass/2-s interval) Saturation Flow (pass/2-s)
Fig. 9. Histogram of an alighting process for a 1300-mm door and no vertical gap.
Table 6 Values of passenger saturation flow for the study cases in HDL-UANDES. Door width (mm)
Saturation flow (pass/s)
Saturation flow (pass/s-m)
600 800 1100 1300 1650 1850 2000
1.093 1.467 2.067 2.600 3.400 3.680 3.840
1.822 1.833 1.879 2.000 2.061 1.989 1.920
explanations could be: (a) the passenger density was 75% greater in the HDL than in PAMELA (3.5 pass/m2 versus 2.0 pass/m2); (b) watching the videos of the experiments on both laboratories it was found that the discharge process at PAMELA was more relaxed than in the HDL, which may be explained because of the conditions of the experiments (e.g., the characteristics of the mock-up); and (c) The response of passengers to the experimental instruction ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. These are new variables that seem to affect the passenger saturation flows, which should be further studied. In summary, results obtained at the HDL of Universidad de los Andes, in passenger per second and passenger per second per metre, are shown in Table 6 and Fig. 10. As can be seen in Fig. 10, the passenger saturation flows in passenger per second grows with the door width, as expected. However, there seems to be a decreasing return as the doors become wider. In fact, from 1650 mm, the rate of increase in the
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
111
4.00
3.50
Saturation Flow
3.00
2.50
2.00
1.50
1.00
0.50
0.00 600
800
1100
1300
1650
1850
2000
Door Width (mm)
Saturation Flow (pass/s)
Saturation Flow (pass/s-m)
Fig. 10. Behaviour of passenger saturation flow for the study cases in HDL-UANDES.
saturation flow decreases. This is more apparent if we observe the saturation flow in pass/s-m. Clearly, the highest saturation flow is reached for a 1650-mm door. Therefore, according to our data, this width seems to be optimum. 5. Conclusions The first conclusion of this work is that our hypothesis was validated; i.e., the same sort of discharge curve of vehicles at signalized junctions can be found in doors of public transport for passengers. Obviously, the discharge curves cannot be identical, because of the different nature of the discharge regulation. In the case of vehicles, regulation is carried out by a traffic signal with red, green and amber indications. For passengers, a door opens and closes, an alarm sounds, or all passengers alight. The last case is similar to a traffic signal in which the queue of vehicles clears before the amber period: during some intervals at the beginning of the green period, the saturation flow can be measured. The theoretical consequence is that the capacity of a public transport door cannot be considered just as the inverse of the average boarding or alighting time per passenger. This is because of the existence of an initial transient period and a final decay in the flow rate as the bunch of passengers inside the vehicle decreases. Therefore, the capacity of a public transport door is the inverse of the headways during the saturation period. As far as we know, this concept has not been reported in the literature, despite the important contributions of Hoogendoorn and Daamen (2005), Daamen et al. (2008), and Daamen and Hoogendoorn (2010). It was found in this work that passenger saturation flows depends on the width of the door, the height of the step, and the passenger density inside the vehicle at the start of the alighting process. Thus, the values found in this study are in the range between 0.9 and 2.0 passengers per second per metre, depending on the aforementioned variables. In relation to the methodology, it is important to mention the successful application of the Road Note 34 (RN 34) approach for pedestrians. We can say that the RN 34 can also be used for the estimation of pedestrian saturation flows. In order to do that, it is only necessary to change the proposed 6-s intervals for cars into 2 or 3-s intervals for people. Despite the reduction in the time width of the intervals, the number of saturated intervals in the HDL experiments was no more than 4 intervals. Are these enough to get a reliable saturation flow? We postulate that they are sufficient – looking at the examples of vehicle saturation flows, with no more than 5 saturated intervals we get a ground truth value. Passenger saturation flows as those found in this research can be used by traffic engineers to estimate delays of public transport vehicles at bus stops and metro stations. This in turn can help transport planners with the calculation of commercial speed, fleet size, type of vehicles, and operational costs of public transport systems. Therefore, our findings may have influence in the formulation of transport policies that pay more attention to the flow of passengers. The use of real-scale laboratory experimentation as PAMELA at University College London and the HDL of Universidad de los Andes will help researchers working in public transport modelling at different levels, since operations at stations and bus stops are being considered increasingly more in transport modelling. The work done so far indicates that some advances on transport modelling are only possible through real-scale laboratory experimentation.
112
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Further research on passenger saturation flow in public transport doors at the HDL of Universidad de los Andes may involve the following studies. Explore more values of vertical gaps to see if an optimum can be found; our hypothesis is that the optimum may be between 0 and 150 mm, according to Tyler (2002). Study the effect and magnitude of the passenger density and types of passengers inside the vehicle at the start of the alighting process on passenger saturation flows. Test the effect and magnitude of the location of horizontal and vertical handrails inside the vehicle on passenger saturation flows. Explore the effect and magnitude of the physical and operational design of the platform on passenger saturation flows. Finally, further studies can be performed to obtain pedestrian saturation flows in other pedestrian devices, such as turnstiles, pedestrian crossings, and revolving doors. Acknowledgements This research is being financed by FONDECYT – Chile, Grant No. 1120219. The authors would like to thank the staff of PAMELA for their help in conducting the experiments as well as Joaquin Fernandez for his help with the experiments and analysis of results at the Human Dynamics Laboratory of Universidad de los Andes. We also thank the comments of two anonymous reviewers, which have significantly helped to improve this article. References Akçelik, R., 1995. Traffic Signals – Capacity and Timing Analysis. Australian Road Research Board Ltd., Research Report 123, Sixth Reprint, Vermont South, Victoria. Daamen, W., Hoogendoorn, S., 2010. Capacity of doors during evacuation conditions. Proc. Eng. 3, 53–66. Daamen, W., Lee, Y., Wiggenraad, P., 2008. Boarding and alighting experiments: an overview of the set up and performance and some preliminary results on the gap effects. Transp. Res. Rec. 2042, 71–81. Fernandez, R., 2011. Experimental study of bus boarding and alighting times. In: European Transport Conference 2011, 10–12 October 2011, Glasgow. Fernandez, R., Valencia, A., Seriani, S., Fernandez, J., Zegers, P., 2013. Passenger saturation flows through public transport doors. In: European Transport Conference 2013, 30 September–2 October 2013, Frankfurt. Hoogendoorn, S.P., Daamen, W., 2005. Pedestrian behavior at bottlenecks. Transport. Sci. 39 (2), 147–159. Kretz, T., Grünebohm, A., Schreckenberg, M., 2006. Experimental study of pedestrian flow through a bottleneck. J. Stat. Mech.: Theory Exp., P10014 Nagai, R., Fukamachi, M., Nagatan, T., 2006. Evacuation of crawlers and walkers from corridor through an exit. Physica A 367, 449–460. Predtechenskii, V.M, Milinskii, A.I., 1978. Planing for Foot Traffic Flow in Buildings. Amerind Publishing, New Delhi. Rexfelt, O., Schelenz, T., Karlsson, M., Suescun, A., 2014. Evaluation the effects of bus design on passenger flow: is agent-based simulation a feasible approach? Transp. Res. Part C 38, 16–27. RRL, 1963. A Method for Measuring Saturation Flow at Traffic Signals. Road Note 34, Road Research Laboratory, Crowthorne. Schadschneider, A., Klingsch, W., Kluepfel, H., Kretz, T., Rogsch, C., Seyfried, A., 2008. Evacuation dynamics: empirical results, modeling and applications. Encyclopedia of Complexity and Systems Science. Springer, Berlin. Schelenz, T., Suescun, A., Wikström, L., Karlsson, M., 2012. Passenger-centered design of future buses using agent-based simulation. Proc. – Soc. Behav. Sci. 48, 1662–1671. Schelenz, T., Suescun, A., Karlsson, M., Wikström, L., 2013. Decision making algorithm for bus passenger simulation during the vehicle design process. Transp. Policy 25, 178–185. TRB, 2000. Highway Capacity Manual 2000, Special Report 209, Transportation Research Board, Washington, D.C. TRB, 2003. Transit Capacity and Quality of Service Manual, Part 5 Rail Transit Capacity, second ed., Transportation Research Board, Washington, D.C. Tyler, N., 2002. Accessibility and the Bus System. Thomas Telford, London.
Contents lists available at ScienceDirect
Transportation Research Part A journal homepage: www.elsevier.com/locate/tra
On passenger saturation flow in public transport doors Rodrigo Fernández a,⇑, Alejandra Valencia b, Sebastian Seriani a a b
Universidad de los Andes, Chile Pontificia Universidad Católica de Valparaíso, Chile
a r t i c l e
i n f o
Article history: Received 9 January 2014 Received in revised form 17 April 2015 Accepted 1 May 2015 Available online 3 June 2015 Keywords: Public transport Passengers Saturation flow Door capacity
a b s t r a c t In previous studies the authors have shown passengers’ boarding and alighting times for the Transantiago system obtained at the Pedestrian Accessibility and Movement Environment Laboratory (PAMELA) of University College London. Following this line of research, the aim of this paper is to demonstrate the existence of pedestrian saturation flows in public transport doors and show some values of this variable under different conditions. The methodology to achieve this aim was real-scale experiments performed in both PAMELA and the Human Dynamics Laboratory at Universidad de los Andes in Santiago de Chile. Different groups of people getting off a mock-up of a public transport vehicle were recorded by means of video cameras. The videos were then visually processed to find values of passenger saturation flow according to door configurations. The variables studied were the vertical gap between the platform and the vehicle chassis and the width of the door. Results indicate that it is possible to define values of passenger saturation flows for different characteristics of public transport doors. These values proved to be statistically sensitive to both the vertical gap and the width of the door. In addition, results indicate that there seems to be both a vertical gap and a door width for which the flow of passenger reaches its optimum rate. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction A main challenge for public transport is to provide a good level of service. This is focused on quantitative and qualitative characteristics of the journey. One of the most important variables that affect the level of service is dwell time. This variable is defined as the time that a public transport vehicle remains stopped while transferring passengers, including acceleration and braking time (TRB, 2000). Dwell time depends on the number of boarding and alighting passengers and how quick they carry out these tasks. The speed of passengers is determined by the fare collection method, internal layout of vehicles and the density of passengers. However as it is said in the Transit Capacity and Quality of Service Manual the most important variable that affects dwell time is the number of doors, because ‘‘the greater the number of door channels, the less time required to serve a given passenger flow’’ (TRB, 2003: 23). Passenger flow for public transport doors is defined as the number of passengers that pass through a car doorway width. The data can be partitioned into boarding, alighting or mixed flow. When this flow reaches the highest value it can be called passenger saturation flow, similar to the concept of vehicle saturation at junctions. According to Akçelik (1995), saturation flow is used at junctions as a basic characteristic to calculate the capacity of a traffic signal approach during a typical signal cycle (C). For a given junction approach, saturation flow is defined as the ⇑ Corresponding author. http://dx.doi.org/10.1016/j.tra.2015.05.001 0965-8564/Ó 2015 Elsevier Ltd. All rights reserved.
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
103
maximum discharge rate of a queue of vehicles during the effective green time (g) of that approach, as shown in Fig. 1. At the start of the green period (G) there is a transient period (L1), called start loss, before the discharge rate reaches its maximum, which is the saturation flow (S) for that approach. If the queue remains until the start of the amber time (A), there is another transient period (L2), called end gain, until the start of the red time (R). Thus, the effective green time is defined as g = G L1 + L2. The value of the saturation flow and transient periods depend on both the traffic composition and geometry of the junction approach. The histogram shown in Fig. 1 shows the discharge of vehicles during 0.1 min if the queue remains until the end of the amber time. In this example, the green plus amber time is 0.7 min or 42 s. This sort of traffic behaviour is obtained if there is no blockage downstream of the stop line; e.g., the downstream street is not blocked by the backup of vehicles which may reduce the discharge rate. The relationship between saturation (S) and capacity (Q) of a traffic signal approach is Q = uS, where u = g/C. We tried to apply a similar concept to the alighting capacity of a public transport door. In the case of a public transport door, the cycle time ‘‘C’’ can be considered as the time between the stop and the start of the vehicle at the station, and the effective green time ‘‘g’’ would be the time during which doors are opened. Therefore, if a vehicle remains a different time period at the station and/or the time during which door are opened changes, the passenger capacity of the door will be different. Which do remain constant is the passenger saturation flow, as in the case of a signalized junction. The flow–time curve shown in Fig. 1 indicates that not all vehicles in a queue take the same time to cross the stop line at a traffic signal. The question posed by this research is: Could the same behaviour occur in public transport doors? Our hypothesis is that the same sort of discharge curve of passengers may be found at public transport doors. That is, an unobstructed alighting process of a bunch of passengers through a public transport door may be similar to the unobstructed discharge of a queue of vehicles at a traffic signal. Following this hypothesis, the aim of this paper is to study the passenger saturation flow in public transport doors by mean of real experiments. The results are part of a research project funded by FONDECYT – Chile (No. 1120219) which has the objective of generating design guidelines to public transport facilities in developing countries. This includes the following specific objectives: (a) Demonstrate the concept of passenger saturation flow in public transport doors; (b) to measure passenger saturation flows in laboratory; (c) to study the effects of different physical design variables on the passenger saturation flow such as door width and vertical gap. This article is made of five sections, including this introduction. In the second chapter the literature review on boarding and alighting times as well as door capacity is summarized. Next, in chapter three the methodology for obtained passenger saturation flows is shown. Results and analysis are presented in chapter four. Finally, the conclusions of this research are provided in chapter five. 2. Literature review During 2008 and 2009 at UCL PAMELA, a real-scale mock-up of a bus and people getting on and off the mock-up were used to obtain average boarding and alighting times per passenger (Fernandez, 2011). In the 2008 experiments, at PAMELA three variables were controlled: platform height, door width, and fare collection method. In 2009 the effect of the density of passengers inside the vehicle on the average boarding and alighting times was studied. In this case, different passenger densities were considered. More than 300 video records of boarding and alighting processes were obtained in those experiments. Tables 1 and 2 show a summary of the results of the experiments obtained from at least 30 samples for each density, where l is the mean boarding or alighting time per passenger, r is the corresponding standard deviation
Flow [veh/0.1min]
Effective flow histogram
S Actual flow histogram
0.1
0.2
0.3
L1
0.4 g
G
0.5
0.6
Time [min]
0.7
L2 A
R
Fig. 1. Typical flow discharge at a traffic signal.
104
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Table 1 Boarding times per passenger, PAMELA experiments (Fernandez, 2011). Density (pass/m2)
l (s/pass)
Maximum (s/pass)
Minimum (s/pass)
r (s/pass)
C (%)
0 1 2 4 6
0.89 1.13 1.37 1.67 2.03
1.04 1.45 1.75 2.66 3.57
0.63 0.75 0.95 1.20 1.36
0.10 0.14 0.21 0.34 0.56
11 12 15 20 28
Table 2 Alighting times per passenger, PAMELA experiments (Fernandez, 2011). Density (pass/m2)
l (s/pass)
Maximum (s/pass)
Minimum (s/pass)
r (s/pass)
C (%)
0 1 2 4 6
0.59 0.92 1.11 1.89 5.92
0.66 1.17 1.46 2.53 9.73
0.50 0.80 0.94 1.26 2.06
0.04 0.08 0.13 0.33 1.70
7 9 12 17 29
and c = r/l is the corresponding coefficient of variation of the results. The mean alighting or boarding time is calculated for each respective process, dividing the total alighting or boarding time by the number of passengers that get off or on the vehicle. As can be seen in the tables, for the same passenger density inside the bus there is a range of boarding and alighting times per passenger so that c ranges from approximately 10% to 30%. Table 1 also shows that as passenger density inside the bus increases, boarding times increase almost linearly from 0.9 to 2.0 s per passenger. However, in the case of alighting times shown in Table 2 these increase almost exponentially from 0.6 to 6.0 s per passenger. It is interesting to note that passengers take longer to get on than to get off the vehicle; up to a density of 4 passengers per square metre. Above that density the situation reverses because the congestion inside the vehicle makes it more difficult for a passenger to reach the exit door. In relation to door widths and their relationship to the dwell time in trains, the Transit Capacity and Quality of Service Manual (TRB, 2003) states that between 1.14 and 1.37 m-width there is no major change in dwell time, reaching average values of 1.77 s/pass during alighting processes, 1.99 s/pass during boarding processes and 2.68 s/pass for two-way flow; i.e., boarding and alighting processes. For that range of door widths TRB (2003) states that a single stream is formed and occasionally three passengers will move through the doorway simultaneously. The same reference states that a door with steps approximately doubles the time of boarding and alighting in relation to a door without steps. Recently, Schelenz et al. (2013) stated that the number of doors can affect directly the time of boarding and alighting. This can be proved by simulation. For example, a bus with 5 doors get an average dwell time of 19.22 s; 39% shorter than a bus with 3 doors. Similarly, Schelenz et al. (2012) compared the alighting and boarding times with respect to the number of passengers and doors. For example, for a 3-door bus, when the number of passengers increases from 30 to 60, the average alighting time goes up by nearly 4 s. Also, Rexfelt et al. (2014) used a mock-up of a public transport vehicle to prove that a bus with 4 doors will have a dwell time 17% lower than a bus with 3 doors. However those studies did not consider variations of door width. Other authors (Schadschneider et al., 2008) state that the capacity of bottlenecks (like doors in a corridor) can grow with increased width. When pedestrians are formed in lanes, the capacity will be increased only if a new lane is formed. This is shown in different models (Predtechenskii and Milinskii, 1978; Hoogendoorn and Daamen, 2005; Kretz et al., 2006; Nagai et al., 2006). All these models were performed in laboratory experiments. However, the closest work related with our research is Daamen et al. (2008). They studied the boarding and alighting times with respect to different vertical and horizontal gaps and door widths in rail vehicles by performing laboratory experiments. Some of their results are shown in Table 3.
Table 3 Door capacities (pass/s) for a door width of 80 cm (Daamen et al., 2008). Vertical gap (mm)
Horizontal gap (mm) No luggage
50 200 400 600
Luggage
50
150
300
50
150
300
0.91 – 0.81 –
– 0.89 – 0.84
0.85 0.88 0.77 0.77
0.69 – 0.65 –
– 0.62 – 0.60
0.73 0.64 0.63 0.56
105
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
According to Daamen et al. (2008) the capacity of public transport doors decreases when gaps (vertical and horizontal) increase. However, for small vertical gaps, door capacity slightly increases as horizontal gap increases. They also found that capacity decreases up to 25% if passengers are carrying luggage. However, this is out of the scope of this article as we studied urban public transport conditions; i.e., passengers with no luggage. Following the work of Daamen et al. (2008), Daamen and Hoogendoorn (2010) identified the relation between door capacity and four experimental variables (door width, population, light intensity and the presence of an open door) in evacuation conditions. However this situation is out of the scope of this article. 3. Methodology We used the methods described in Road Note 34 – RN34 – (RRL, 1963) for measuring passenger saturation flows in both PAMELA and the Human Dynamics Laboratory (HDL) of Universidad de los Andes. The RN34 describes a method for measuring vehicle saturation flow at traffic signals. It consists of counting the number of vehicles crossing the stop-line in successive intervals during a saturated green period, as shown in Fig. 1. ‘‘The basis of this method is to divide the saturated portion of each green period into short intervals of time and to average the flows in those saturated intervals which are free from ‘lost time’ effects, to give a measure of the saturation flow. . .’’ (RRL, 1963:2). The RN34 also states that ‘‘. . .the method consists of recording the number of vehicles discharging from the queue in successive 0.1-min intervals... When the flow is no longer at the saturation level because the queue has disappeared on one or more lanes the recording of the flow in 0.1-min interval should be discontinued’’ (RRL, 1963: 4). As an example for understanding the RN 34 method, Table 4 shows the application at a traffic signal in Santiago de Chile. Nine green plus amber fully-saturated periods of 36 s are shown. According to the RN 34 method, each green plus amber period must be divided into 0.1-min or 6-s intervals, plus a last interval of variable length. In this example six 6-s intervals give 36-s green plus amber time. The saturation flow is calculated as the average ratio between the number of vehicles and the number of saturated green plus amber periods, excluding the first and the last interval. In this case, the saturation flow is S = (2.67 + 2.67 + 3.22 + 2.78 + 2.44)/5 = 2.76 veh/0.1-min; i.e., 1653 veh/h. This means 1653 veh/h per lane which is a typical value when there are only light vehicles but various types of movements at the junction (Akçelik, 1995: Table 5.1). Fig. 2 shows the discharge histogram at this junction, where the dotted line is the average height of the saturated intervals (intervals 2–6); in other words, the saturation flow. As can be seen in the figure, there are differences in the height of the saturated intervals and the difference between two adjacent saturated intervals is about 20%. Therefore, in each saturated period the outflow is not the same because in this particular site all movements take place (through, right, left), then due to the traffic composition or traffic movements, not all intervals contain the same number of vehicles. In order to achieve the same condition for public transport doors, unrestrained discharges of passengers must be studied. As a result, alighting processes are needed from public transport vehicles in which passengers are not obstructed by the congestion caused by them. This condition is difficult to obtain in real systems with enough of alighting demand because of the width of the platform, the size of the stop shelter or pavement width. Therefore, a controlled environment could be necessary such as a laboratory and a real-scale mock-up. This environment was provided for this study by both PAMELA and HDL. Saturation flow was measured according to the RN34 but applied to the passenger alighting process. We divided the alighting period into 2 or 3-s intervals and counted the number of passengers in each interval from the time the door is opened until all the passengers have gotten off the vehicle. Then a histogram that represents the alighting period is constructed. The width of the intervals was chosen in order to have at least 4 saturated intervals, according to the minimum subjects for experimentation (20 people). In the application of RN34 to passengers, the first interval starts when the doors start to open, because passengers jump off the vehicle as soon as they see a gap in the door. Then the average height of the saturated intervals gives the saturation
Table 4 RN34 method applied in a traffic signal in Santiago de Chile. Green plus Amber periods
1 2 3 4 5 6 7 8 9 Total Sample Average
0.1-min intervals
Last interval
1
2
3
4
5
6
Vehicles
Time (s)
1 1 1 1 1 2 1 1 1 10 9 1.11
2 3 2 4 3 2 3 2 3 24 9 2.67
3 2 3 2 2 3 3 3 3 24 9 2.67
3 2 3 3 3 4 4 4 3 29 9 3.22
2 3 2 3 3 4 3 2 3 25 9 2.78
3 3 3 3 2 1 3 2 2 22 9 2.44
2 2 1 2 3 1 1 2 2 16 9 1.78
2.00 2.83 3.33 2.00 1.00 2.33 2.50 2.00 2.25 20.24 9 2.25
106
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Saturation flow [veh/6-s]
3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00
1
2
3
4
5
6
7
6-sec intervals [s] Fig. 2. Example of a flow discharge at a traffic signal in Santiago de Chile.
flow. Intervals belong to the start loss period; and those in which there is no more saturation are not considered for the calculation. To decide when saturation ends, we observe that the height of an interval is clearly below the average height of the saturated intervals. We applied our engineering criterion and experience to take this decision. Obviously, there is no final transient because there is nothing similar to the amber period in the case of passengers. On the contrary, when everybody gets off the train, the measurement stops.
3.1. Experiments at PAMELA in University College London In our experiments at PAMELA in 2008 and 2009, we recorded both boarding and alighting processes in a constructed bus mock-up. As shown in Fig. 3, the mock-up is 10-m long by 2.5-m wide with a 1.6-m door in the middle, which can be arranged to be opened halfway. The mock-up can be raised to different heights to represent different levels of platform. In order to test our hypothesis we took only the alighting processes in which all passengers get off the vehicle. A total of 120 alighting processes were observed. For each alighting process a number of 50 passengers get off the vehicle. Passengers were male and female adult Londoners, of different ages. The instruction given to passengers was ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. The control variables we considered for PAMELA experiments were door widths and platform height. Two door widths were tested: a wide door (1600 mm) and a narrow door (800 mm). Somehow these values try to represent extreme values of door widths of public transport vehicles in Santiago de Chile; i.e., 800 mm for small buses and 1600 mm for metro cars. Similarly, three platform heights were tested: a high platform (300 mm height), a normal platform (150 mm height) and no platform (0 mm height). The first case represents a specially designed stop in which the platform has the same height than the vehicle; so the is no vertical gap for alighting passengers. The normal platform represents the case in which passengers alight into the pavement; therefore the vertical gap has 150 mm. Finally, in the no-platform case passengers have to alight to the road surface; hence the vertical gap is 300 mm, which is common in developing countries or for countryside services in any setting.
Fig. 3. Mock-up in PAMELA of University College London.
107
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
3.2. Experiments at the Human Dynamics Laboratory in Universidad de los Andes The HDL at the Universidad de los Andes was designed to study human dynamics; from parts of the body to groups of people. In these experiments, a mock-up of the hall of a public transport vehicle was built for our study. Fig. 4 shows the concept and the layout of the mock-up, and Fig. 5 shows a picture during the experiments at the HDL. In order cover any type of vehicle – from normal buses to metro cars – the mock-up of the hall is 3-m long by 2.5-m wide with a door in the middle which can be opened to 2-m wide. At both sides of the hall there is a 0.8-m doorway to represent the aisle of the vehicle. The size of the platform is variable according to the transit system to be investigated. In the HDL experiments we only tested door width as a control variable (Fernandez et al., 2013); therefore, a zero vertical gap was assumed. Seven door widths were tested: 600, 800, 1100, 1300, 1650, 1850, and 2000 mm. These numbers were chosen because they represent the different doors of public transport vehicles in Santiago de Chile: small buses (600 and 800 mm), normal and articulated buses (1100 mm), and metro cars (1300 and 1650 mm). As far as we know, there do not exist 1850 and 2000-mm doors in urban transit systems. We chose those values to see the passenger behaviour under a reality which can only be tested in a laboratory. A number of 10 alighting processes were recorded for each width, making a total of 50 observations. In each alighting process, 26 passengers get off the vehicle. Passengers were students of Universidad de los Andes, 18–25 years old, 10% female and 90% male. As in the case of PAMELA, the instruction given to passengers was ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. 4. Results and analysis 4.1. Results from PAMELA experiments Fig. 6 shows a typical histogram of an alighting process. The figure shows 21 intervals of three seconds each. We chose 3-s intervals because of the acceleration abilities of passengers are different for vehicles. If a longer interval was chosen, we did not have enough number of intervals to calculate saturation flows.
Exit hall
Entry hall
Ticket sensor
2.30 0.78
Ticket sensor
1.40
1.40
1.50
1.10
1.10
(a) Entry and exit halls of a public transport vehicle
“Vehicle” “Aisle”
“Aisle” Sliding
“Platform
Sensor
Camera
(b) Layout of the mock-up Fig. 4. Mock-up in the HDL at the Universidad de los Andes.
108
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Fig. 5. Experiments at the HDL of Universidad de los Andes.
The saturation flow obtained is 2.559 passengers in a 3-s interval – i.e., 0.853 passengers per second – for a narrow door and no vertical gap. The first interval represents the start loss and the last three intervals correspond to the end of the saturation. To decide when saturation flow ends, we observe the difference between two adjacent saturated intervals. In this case, the difference between the flow in the intervals 18 and 19 is 31% which indicates that saturation lasts until the 18th interval. In Fig. 7 the same is shown for a wide door and 150-mm vertical gap. As can be seen in the figure, saturation ends in the 9th interval, because the difference between the flow in the intervals 9 and 10 is 32%. As can be seen in the figures, the number of intervals decreases as the door widens, because more passengers can pass thought the door simultaneously. Thus, 4.405 pass/3-s interval can pass through a wide door if there is no vertical gap between the vehicle and the platform; i.e., 1.468 pass/s. Note that this is not twice the flow through a narrow door, as discussed below. Table 5 shows the summary of results of the experiments. As expected, it can be observed from the table that a wide door can reach higher values of saturation flow than a narrow door. However, if the door width doubles, the saturation flow does not increase in the same proportion. For example, for the no-gap case the increase of saturation flow between the two widths is only 72%. The same occurs for the other gaps for which doubling the door width increases the saturation flow in about 65%. This suggests the existence of an optimum width for which the passenger flow per metre reaches its maximum rate. In relation to the saturation flow expressed in passengers per second per metre, it can be observed from Table 5 that the average value obtained was a bit more than 1.0 pass/s-m (1.047). A statistical hypothesis test between the different values of saturation flow in pass/s-m was performed to decide if they are statistically equal at 5% significance level in order to recommend that average value. Thus, for each vertical gap we tested if saturation flows for an 800 and a 1600-mm door are statistically equal (the H0 hypothesis). Similarly, for each door width we tested if saturation flows for 0, 150 and 300-mm vertical gap are statistically equal. As a result, the H0 hypothesis was rejected in all cases at the 5% level. This means that passenger saturation flows in pass/s-m are all statistically different and a single average value cannot be recommended.
Door 800 mm -Gap 0 mm
Number of passengers
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
3-s intervals Frequency (pass/interval)
Saturation Flow (pass/3-s)
Fig. 6. Histogram of an alighting process for a narrow door and no vertical gap.
109
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Door 1600 mm - Gap 150 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
8
9
10
11
12
13
3-s Intervals Frequency (pass/interval)
Saturation Flow (pass/3-s)
Fig. 7. Histogram of an alighting process for a wide door and 150-mm vertical gap.
If we analyse the behaviour of the saturation flow with respect to the door width and the vertical gap, in Table 5 it can be observed that the maximum saturation flow is obtained for a wide door and a medium vertical gap. In relation to the vertical gap, results suggest the existence of an optimum vertical gap for which the passenger’s saturation flow reaches its maximum value. After watching the videos, a possible explanation is that for 150 mm or less, people do a little ‘‘jump’’ to get off the vehicle, which may speed up the alighting process. On this issue, Tyler (2002: 149) states that ‘‘a vertical gap of 75 mm could be tolerated if the horizontal gap were 25 mm. However, if the horizontal gap is more than 50 mm, the vertical gap can be no more than 50 mm. In no case should either gap be more than 100 mm.’’Our results show that a 150-mm as the ‘‘best’’ vertical gap, because we did not test gaps between 150 and 0 mm. Further experiments at the HDL of Universidad de los Andes may provide a deeper understanding on this issue. Comparing the above results with Daamen et al. (2008) for an 80-cm door width in Table 3, it can be seen that there are similarities. For example, Daamen et al. (2008) arrived at a door capacity between 0.85 and 0.91 pass/s for a 50-mm vertical gap, which is similar to our result of 0.853 pass/s for no vertical gap. Table 3 also shows that for a 200-mm vertical gap, door capacity reaches a value between 0.88 and 0.89 pass/s, which is similar to our result of 0.90 pass/s for a 150-mm vertical gap. What is relevant is that in both studies capacity slightly increased for an intermediate vertical gap: 200 mm in the case of Daamen et al. (2008) and 150 mm in our case.
4.2. Results from the HDL experiments Experiments at the HDL of Universidad de los Andes show similar histograms of the alighting processes to those found in PAMELA. For example, Fig. 8 shows 4 intervals of three second each for which the saturation flow is 4.389 passengers per 3-s interval or 1.467 passengers per second for an 800-mm door. The first interval represents the start loss and the last two ones are the end of the saturation, because the flow difference between the 5th and 6th intervals is 24%. In Fig. 9 the same is shown for a 1300-mm door. In this case, the saturation flow is 5.2 passengers per 2-s interval; i.e., 2.6 passengers per second. In this case, the flow difference between the 5th and 6th intervals is 66%. It was observed that if 1-s intervals are used, the discharge histogram is constructed of more samples (8 intervals). However, the value of the passenger saturation would be the same, because the start loss last for 2 intervals and the saturation period ends in the 7th interval. Therefore, the saturation period remains for 4 intervals, same as for 3-s intervals or 2-s intervals. It should be noted that the saturation flows found at the HDL are greater than those found at PAMELA for the same vertical gap. For example, for the 800-mm door we measured a saturation flow of 1.467 pass/s in the HDL versus 0.853 pass/s in PAMELA; that is, a 72% higher value. We do not have an ultimate explanation for this difference. However, possible
Table 5 Values of passenger saturation flow for the study cases in PAMELA. Door width (mm)
Vertical gap (mm)
Saturation flow (pass/s)
Saturation flow (pass/s-m)
800 800 800 1600 1600 1600
0 150 300 0 150 300
0.853 0.991 0.900 1.468 1.610 1.481
1.066 1.239 1.125 0.918 1.007 0.926
110
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Door 800 mm - Gap 0 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
3-sec intervals Frequency (pass/interval) Fig. 8. Histogram of an alighting process for an 800-mm door and no vertical gap.
Door 1300 mm -Gap 0 mm
Number of passengers
6.0 5.0 4.0 3.0 2.0 1.0 0.0 1
2
3
4
5
6
7
2-sec intervals Frequency (pass/2-s interval) Saturation Flow (pass/2-s)
Fig. 9. Histogram of an alighting process for a 1300-mm door and no vertical gap.
Table 6 Values of passenger saturation flow for the study cases in HDL-UANDES. Door width (mm)
Saturation flow (pass/s)
Saturation flow (pass/s-m)
600 800 1100 1300 1650 1850 2000
1.093 1.467 2.067 2.600 3.400 3.680 3.840
1.822 1.833 1.879 2.000 2.061 1.989 1.920
explanations could be: (a) the passenger density was 75% greater in the HDL than in PAMELA (3.5 pass/m2 versus 2.0 pass/m2); (b) watching the videos of the experiments on both laboratories it was found that the discharge process at PAMELA was more relaxed than in the HDL, which may be explained because of the conditions of the experiments (e.g., the characteristics of the mock-up); and (c) The response of passengers to the experimental instruction ‘‘please get-off the vehicle as if you were to continue your journey on foot during the rush hour’’. These are new variables that seem to affect the passenger saturation flows, which should be further studied. In summary, results obtained at the HDL of Universidad de los Andes, in passenger per second and passenger per second per metre, are shown in Table 6 and Fig. 10. As can be seen in Fig. 10, the passenger saturation flows in passenger per second grows with the door width, as expected. However, there seems to be a decreasing return as the doors become wider. In fact, from 1650 mm, the rate of increase in the
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
111
4.00
3.50
Saturation Flow
3.00
2.50
2.00
1.50
1.00
0.50
0.00 600
800
1100
1300
1650
1850
2000
Door Width (mm)
Saturation Flow (pass/s)
Saturation Flow (pass/s-m)
Fig. 10. Behaviour of passenger saturation flow for the study cases in HDL-UANDES.
saturation flow decreases. This is more apparent if we observe the saturation flow in pass/s-m. Clearly, the highest saturation flow is reached for a 1650-mm door. Therefore, according to our data, this width seems to be optimum. 5. Conclusions The first conclusion of this work is that our hypothesis was validated; i.e., the same sort of discharge curve of vehicles at signalized junctions can be found in doors of public transport for passengers. Obviously, the discharge curves cannot be identical, because of the different nature of the discharge regulation. In the case of vehicles, regulation is carried out by a traffic signal with red, green and amber indications. For passengers, a door opens and closes, an alarm sounds, or all passengers alight. The last case is similar to a traffic signal in which the queue of vehicles clears before the amber period: during some intervals at the beginning of the green period, the saturation flow can be measured. The theoretical consequence is that the capacity of a public transport door cannot be considered just as the inverse of the average boarding or alighting time per passenger. This is because of the existence of an initial transient period and a final decay in the flow rate as the bunch of passengers inside the vehicle decreases. Therefore, the capacity of a public transport door is the inverse of the headways during the saturation period. As far as we know, this concept has not been reported in the literature, despite the important contributions of Hoogendoorn and Daamen (2005), Daamen et al. (2008), and Daamen and Hoogendoorn (2010). It was found in this work that passenger saturation flows depends on the width of the door, the height of the step, and the passenger density inside the vehicle at the start of the alighting process. Thus, the values found in this study are in the range between 0.9 and 2.0 passengers per second per metre, depending on the aforementioned variables. In relation to the methodology, it is important to mention the successful application of the Road Note 34 (RN 34) approach for pedestrians. We can say that the RN 34 can also be used for the estimation of pedestrian saturation flows. In order to do that, it is only necessary to change the proposed 6-s intervals for cars into 2 or 3-s intervals for people. Despite the reduction in the time width of the intervals, the number of saturated intervals in the HDL experiments was no more than 4 intervals. Are these enough to get a reliable saturation flow? We postulate that they are sufficient – looking at the examples of vehicle saturation flows, with no more than 5 saturated intervals we get a ground truth value. Passenger saturation flows as those found in this research can be used by traffic engineers to estimate delays of public transport vehicles at bus stops and metro stations. This in turn can help transport planners with the calculation of commercial speed, fleet size, type of vehicles, and operational costs of public transport systems. Therefore, our findings may have influence in the formulation of transport policies that pay more attention to the flow of passengers. The use of real-scale laboratory experimentation as PAMELA at University College London and the HDL of Universidad de los Andes will help researchers working in public transport modelling at different levels, since operations at stations and bus stops are being considered increasingly more in transport modelling. The work done so far indicates that some advances on transport modelling are only possible through real-scale laboratory experimentation.
112
R. Fernández et al. / Transportation Research Part A 78 (2015) 102–112
Further research on passenger saturation flow in public transport doors at the HDL of Universidad de los Andes may involve the following studies. Explore more values of vertical gaps to see if an optimum can be found; our hypothesis is that the optimum may be between 0 and 150 mm, according to Tyler (2002). Study the effect and magnitude of the passenger density and types of passengers inside the vehicle at the start of the alighting process on passenger saturation flows. Test the effect and magnitude of the location of horizontal and vertical handrails inside the vehicle on passenger saturation flows. Explore the effect and magnitude of the physical and operational design of the platform on passenger saturation flows. Finally, further studies can be performed to obtain pedestrian saturation flows in other pedestrian devices, such as turnstiles, pedestrian crossings, and revolving doors. Acknowledgements This research is being financed by FONDECYT – Chile, Grant No. 1120219. The authors would like to thank the staff of PAMELA for their help in conducting the experiments as well as Joaquin Fernandez for his help with the experiments and analysis of results at the Human Dynamics Laboratory of Universidad de los Andes. We also thank the comments of two anonymous reviewers, which have significantly helped to improve this article. References Akçelik, R., 1995. Traffic Signals – Capacity and Timing Analysis. Australian Road Research Board Ltd., Research Report 123, Sixth Reprint, Vermont South, Victoria. Daamen, W., Hoogendoorn, S., 2010. Capacity of doors during evacuation conditions. Proc. Eng. 3, 53–66. Daamen, W., Lee, Y., Wiggenraad, P., 2008. Boarding and alighting experiments: an overview of the set up and performance and some preliminary results on the gap effects. Transp. Res. Rec. 2042, 71–81. Fernandez, R., 2011. Experimental study of bus boarding and alighting times. In: European Transport Conference 2011, 10–12 October 2011, Glasgow. Fernandez, R., Valencia, A., Seriani, S., Fernandez, J., Zegers, P., 2013. Passenger saturation flows through public transport doors. In: European Transport Conference 2013, 30 September–2 October 2013, Frankfurt. Hoogendoorn, S.P., Daamen, W., 2005. Pedestrian behavior at bottlenecks. Transport. Sci. 39 (2), 147–159. Kretz, T., Grünebohm, A., Schreckenberg, M., 2006. Experimental study of pedestrian flow through a bottleneck. J. Stat. Mech.: Theory Exp., P10014 Nagai, R., Fukamachi, M., Nagatan, T., 2006. Evacuation of crawlers and walkers from corridor through an exit. Physica A 367, 449–460. Predtechenskii, V.M, Milinskii, A.I., 1978. Planing for Foot Traffic Flow in Buildings. Amerind Publishing, New Delhi. Rexfelt, O., Schelenz, T., Karlsson, M., Suescun, A., 2014. Evaluation the effects of bus design on passenger flow: is agent-based simulation a feasible approach? Transp. Res. Part C 38, 16–27. RRL, 1963. A Method for Measuring Saturation Flow at Traffic Signals. Road Note 34, Road Research Laboratory, Crowthorne. Schadschneider, A., Klingsch, W., Kluepfel, H., Kretz, T., Rogsch, C., Seyfried, A., 2008. Evacuation dynamics: empirical results, modeling and applications. Encyclopedia of Complexity and Systems Science. Springer, Berlin. Schelenz, T., Suescun, A., Wikström, L., Karlsson, M., 2012. Passenger-centered design of future buses using agent-based simulation. Proc. – Soc. Behav. Sci. 48, 1662–1671. Schelenz, T., Suescun, A., Karlsson, M., Wikström, L., 2013. Decision making algorithm for bus passenger simulation during the vehicle design process. Transp. Policy 25, 178–185. TRB, 2000. Highway Capacity Manual 2000, Special Report 209, Transportation Research Board, Washington, D.C. TRB, 2003. Transit Capacity and Quality of Service Manual, Part 5 Rail Transit Capacity, second ed., Transportation Research Board, Washington, D.C. Tyler, N., 2002. Accessibility and the Bus System. Thomas Telford, London.