Empirical Studies on Impacts of Obstacle inside Corridor on Pedestrian Flow

Available online at www.sciencedirect.com

ScienceDirect Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

2nd Conference of Transportation Research Group of India (2nd CTRG)

Empirical Studies on Impacts of Obstacle inside Corridor on Pedestrian Flow Ujjal Chattaraj a1, Partha Chakrobortyb, Arumuga Subhashinia a

Department of Civil Engineering, National Institute of Technology Rourkela, Rourkela 769 008, India b Department of Civil Engineering, Indian Institute of Technology Kanpur, Kanpur 208 016, India

Abstract

Empirically, pedestrian motion can be studied at various levels. At the macroscopic level one may study the basic flow parameters (like speed and density) of pedestrian motion, at the microscopic level one may track the paths followed by individual pedestrians while moving. Whereas, at a mesoscopic level one can study pedestrian motion by concentrating on how the flow parameters change spatially and temporally. Such studies help in understanding pedestrian flow at a reasonably fundamental level and also aid in understanding how various geometric features impact pedestrian motion. In this study, macroscopic and mesoscopic level data are collected from various experiments aimed at understanding impact of corridor geometry on uni–directional and bi– directional pedestrian flow. First one is a “basic uniform width” or a “reference” corridor. To impose geometric variations in the corridor two different procedures are adapted, i) narrowing (symmetric as well as asymmetric) of the corridor, and ii) partial bifurcation of the corridor (by placing obstacle for a certain distance inside the corridor). Experimental details and results on narrowed corridors and their comparisons with the “reference” corridor are presented in another earlier paper. This study concentrates on experimental details and results on partially bifurcated corridor and their comparisons with the “reference” corridor. These experiments are intended to observe the lateral and longitudinal variations in density and longitudinal variations in speed due to the impact of various corridor geometry when pedestrians can walk side–by–side as well as overtake one another. To understand the lateral and longitudinal variations in density some crucial parameters are introduced in this study, which prove to be good quantitative measure in this regard. Results both on density and speed accentuate the fact that partial bifurcation in the corridor adversely impacts pedestrian flow inside the corridor. © Published by Elsevier Ltd. Ltd. © 2013 2013The TheAuthors. Authors. Published by Elsevier Selection under responsibility of International Scientific Committee. Selectionand andpeer-review peer-review under responsibility of International Scientific Committee. Keywords: Indian Pedestrians; mesoscopic level data; partially bifurcated corridor

1

* Ujjal Chattaraj. Tel.: +0-661-246-2327; fax : +0-661-246-2031 E-mail address:[email protected]

1877-0428 © 2013 The Authors. Published by Elsevier Ltd. Selection and peer-review under responsibility of International Scientific Committee. doi:10.1016/j.sbspro.2013.11.161

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

669

1. Introduction Walking is the basic means of transportation. Though various studies have been made on other means of transportation, walking occupies a prominent part which can provide very high levels of accessibility to different facilities. In this study an attempt is made to study Indian pedestrians experimentally by analysing the unidirectional and bi-directional pedestrian flow in a partially bifurcated corridor in comparison with a basic corridor. Generally the pedestrians aim to reach the destination earlier and without bumping onto other pedestrians or obstacles. The space where pedestrians move contains both static and dynamic objects. The static objects are either static obstacles in the flow space which typically repel pedestrians from them or static goals (like egress points) which attract pedestrians. The dynamic objects are other pedestrians who, it is assumed, have a dual impact. On one hand they attract other pedestrians as they can act as pathfinders while on the other hand they repel as they are also physical objects not to be collided with. Over the years various studies on relationships of pedestrian streams have been reported (for example, Oeding (1963), Chattaraj, Seyfried, and Chakroborty (2009), Helbing, Farkas, Molnar and Vicsek (2002) Helbing., Bunza, Johansson and Werner (2005)). Studies on pedestrian crowd by Hoogendoorn and Daamen (2005) have observed zipper effect, Henderson(1974) on Fluid mechanics of Crowd Motion. This paper is divided into three sections of which this is the first to introduce the matter the next section discusses about the experiments conducted and the results obtained on pedestrian flow. Finally, in the last section the conclusions on the experiment of impact of obstacles on pedestrian flow are presented. 2. Experiments and Results The principal factors that affect pedestrian movement are the other pedestrians, the geometry of the facility in which the pedestrian is moving and the choices the pedestrian may have to make when faced with multiple competing goals.

(a)

(b) Figure 1: (a) Schematic showing the experimental set–up for the “basic” corridor (b) Schematic showing the experimental set–up for the partially bifurcated corridor

670

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

These experiments are intended to observe the lateral and longitudinal variations in density and longitudinal variations in speed due to the impact of obstacle in corridor geometry. Figure 1(a) shows the basic experimental set–up. A corridor of 40 m × 2 m with reasonable circulation area on all sides of the corridor was created by putting metal barricades (height ~ 1.6 m) on a paved surface. Experiments were conducted outdoors during daytime in pleasant weather. Two high resolution video cameras (640× 480 pixels) are fitted at a height of approximately 10 m and a distance of 10 m from the near edge of the corridor and along the extended centre line of the corridor. The use of two cameras allows one to zoom in on about half the corridor length thereby allowing a good resolution during video recording. In order to determine the position of pedestrians on this corridor a grid (0.4 m × 0.4 m) is constructed using thin but highly visible wires at a height of 1.65 m, this is approximately equal to the average height of Indian people (Brennan, Mcdonald and Shlomowitz ,1995) from the ground. Once the grid is constructed the fixed cameras record this grid without altering the camera angle or position. For every experiment, at every instant of time ‘t’ the cells (i, j) that are occupied in the grid are noted. As can be seen from Figure 1(a) the experiment starts with randomly distributed pedestrians inside the corridor in such a way that there is no major variation in pedestrian density over space. Data collection begins after some time from the start of the experiment. This time is decided subjectively by the experimenter and depends on when the experimenter feels that the boundary impacts of starting time have vanished. Data is collected for approximately 5 minutes for each experiment. One of the primary thrusts of this category of experiments is to study the impact of corridor geometry on pedestrian flow parameters. It is felt that, if the pedestrian density in the corridor is high then the observed behaviour would not only have the impact of the geometry but will also have the impact of congestion. On the other hand, if the crowd levels are kept very low then one may not get to observe the role that corridor geometry plays on pedestrian movement. Hence, in this study a density of approximately 0.6 persons/m2 is used; with this density typically at any given instant of time there are approximately 50 persons inside the corridor. Figure 1(a) shows the basic 40 m × 2 m corridor. Figure 1 (b) is a partially bifurcated corridor with the width of each of the narrow channels being approximately 40% of the basic corridor. Figure 1 (a) also shows the grid, the i,j axes, and the i, j values on which pedestrian locations at every instant of time is observed. Two spatial units as lane and zone, at an aggregate level are introduced. In order to study the spatial variations in flow parameters from observation the following quantities are used, Relative overall (or relative overall directional) lane density is given by,

OLDl

ROLDl

L

OLDl

l 1

Where, T

J

OLDl

Olt, j

t 1 j 1

T is the total time for which the experiment is conducted and J is the maximum number of cells along the “j” direction and L is the maximum number of lanes. In case of bi-directional flow the relative overall directional lane density for people moving from a to b, RODLDl(a b) is given by T

ODLDl ( a

J

b) t 1 j 1

Olt, j ( a

b)

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

RODLDl ( a

ODLDl ( a b)

L

671

b)

ODLDl ( a

b)

l 1

Relative zonal (or relative zonal directional) lane density RZLDl,z can be obtained by modifying the definitions of ROLDl JU , z

T

ZLDl , z

Olt, j

t 1 j J L ,z

ZLDl , z

RZLDl , z

L

ZLDl , z

l 1

where, JL,z and JU,z are the lower and upper limits of j for a given zone z In case of bi–directional flow the expressions for zonal directional lane density and relative zonal directional lane density are JU , z

T

ZDLDl , z ( a

b)

Olt, j ( a

b)

t 1 j JL ,z

RZDLDl , z ( a

ZDLDl , z b)

L

ZDLDl , z ( a

b)

l 1

Speed is calculated, by dividing the entire length of the corridor into 13 small subzones. The space mean speed is K

Ss

( JU ,s

J L,s )

tUk , s t Lk , s

k 1

K

where, JU,s and JL,s are the upper and lower “j” values, respectively defining a subzone, s, tkU,s and tkL,s are the times at which pedestrian k is at JU,s and JL,s, respectively, and K is the number of pedestrians considered.

(a) (b) Figure 2: Snapshots of experimental setup for the (a) “basic” corridor (b) partially bifurcated corridor

672

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

(a)

0 .3

0 .3

0 .2

0 .2

0 .1

(d )

0 .4

ROLDll

ROLDl

0 .4

0

0 .1 0

1

2

3 l (a)

4

5

1

Legend: ROLDl

2

3 l

4

5

(b)

BLDV Figure 3: ROLDl variations for the (a) basic (b) partially bifurcated corridors Figure 3 (a) is for the geometry shown in Figure 1(a) and shows the impact of corridor edges on the lateral distribution of relative density. It can be clearly seen that the two extreme “lanes” have the least density while the inside “lanes” have almost twice the density as the extreme “lanes”. Given the “keep left” policy followed in India it is not surprising that more people tend towards the left. From the figure it is clear that the boundaries of the corridor have significant impact on the efficient utilization of the corridor. One can say that the effective width of a corridor is less than the actual width of the corridor. In this case, approximately 73% of the people use 60% of the width. This implies, although in a somewhat naive way, that the effective width is only about 82% (= 60 × 100/73) of the total width; i.e., out of 2 m of width available, only 1.65 m is used effectively. Strictly speaking these numbers are valid only for densities around 0.6 persons/m2 and for the kind of barricades used to simulate the boundaries. The fact is that boundaries impact the effective use of the corridor width in much the same way road edges (road width) impact the road usage by vehicles. Figure 3 (b) shows the ROLDl for the case where there is bifurcated narrowing. As can be seen from Figure 1 (b), 1.6 m length of Lane 3 was in effect made inaccessible in this corridor. Figure 3(b) clearly shows that pedestrians have moved away from Lane 3 either to the left of it or to the right of it. In fact the total ROLD of lanes to left of Lane 3 is marginally higher than the total ROLD of lanes to the right (the values are 0.42 and 0.39, respectively). This may indicate that even though while overtaking (or going around some obstacle) pedestrians prefer to do so from right (in cultures like India where “keep left” policy is practiced); the “keep left” urge amongst pedestrians seems to be stronger. Another important point to note is that ROLD2 has gone down when compared to Figure 3 (a), whereas ROLD1 has gone up. The reduction in ROLD of Lane 2 is primarily due to the boundary effect of the inaccessible zone present in Lane 3; possibly for the same reason the ROLD1 has increased as more people from Lane 2 has shifted to Lane 1.

673

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

fu s

0 .3

0 .2

0

0 .3

0 .2 0 .1 0

1

2

3

4

5

c

0 .4

RZLDl,z

0 .3

0 .1

0 .2 0 .1 0

1

2

3

4

5

1

2

3

l

l

l

(a)

(b)

(c)

ds

0 .4 0 .3

0 .3

0 .2

0 .2

0 .1 0

fd s

0 .4

RZLDl,z

RZLDl,z

us

0 .4

RZLDl,z

RZLDl,z

0 .4

4

5

Legend RZLDl,z BLDV

0 .1 0

1

2

3

4

5

1

2

3

l

l

(d)

(e)

4

5

Figure 4: RZLDl,z variations for partially bifurcated corridor at (a) far upstream (b) upstream (c) constricted (d)downstream and (e) far downstream zone Figure 4 shows the variations of RZLDl,z in five different zones of the partially bifurcated corridor. The dotted line in each of the five sub–figures represent the RZLDl,z versus l plot for the basic corridor (BLDV) There is a shift away from the lanes blocked due to narrowing even in the far upstream region. Although, apparently, it seems that as one nears the block in lane 3 people move to the right lanes; a closer look reveals a slightly different story. The RZLDl,us of lanes to the left of Lane 3 (i.e., RZLD1,us + RZLD2,us = 0.47) are slightly higher than the RZLDl,us of lanes to the right of lane 3 (i.e., RZLD4,us + RZLD5,us = 0.43). This tendency of people to gravitate more towards the left lane is accentuated as one nears the obstacle. In the constricted zone two things emerge. One of them is that more people shift to the left opening than to the right opening (RZLD1,c + RZLD2,c = 0.56 and RZLD4,c + RZLD5,c = 0.45). Another interesting aspect is that the outer lanes (Lanes 1 and 5) have the higher RZLDl,c values. This indicates that pedestrians somehow view the edges of obstacles creating the constriction as a bigger threat than the edges of the basic corridor. This could be because pedestrians are “used to” the edges of the corridor by the time they “encounter” the edges of the obstacle. The RZLDl,ds distribution does tend to show a movement towards BLDV ; however still the left two lanes have greater total density than the right two lanes, with people moving from the left two and right two crowded lanes to the more vacant Lane 3 (recall in the constriction zone the left two lanes had the more total RZLDl,ds than the right two lanes).

674

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

1 .5 Speed (m/s)

Speed (m/s)

1 .5 1 0 .5

1 0 .5 0

0 0

10

20

30

0

10

20

30

Distance from entry (m) (a)

Distance from entry (m) (b) Figure 5: Longitudinal speed profile in the case of uni–directional flow for the (a) basic corridor (b) partially bifurcated corridor Figure 5 presents the space mean speeds obtained from a sample of pedestrians over the entire length of the corridor. The following inferences can be made from the speed data. When there is no geometric variation in the corridor the speeds are more or less same throughout the corridor (Figure 5 (a)); the average is about 1.4 m/s and the standard deviation is about 0.06 m/s. When constrictions are introduced the average speeds fall; in all the three cases the average speed is around 1.3 m/s, i.e., a drop of about 7%. A t–test on the difference of means show that this difference is statistically significant. As is expected the speeds are the lowest around the constricted zone; the lowest speed invariably occurs a couple of metres upstream of the constricted zone (Figure 5 (b)); this is in consonance with the general understanding that congestion is severest just upstream of a bottleneck; the lowest speeds in all the three cases are about 20% lower than the basic corridor average speed. Statistical tests on the difference of means show that this difference is statistically significant. Further, on an average, the speeds in the constricted zones are 14% to 18% lower than 1.4 m/s (basic corridor average speed).

(a)

(b)

Figure 6: Snapshots showing the experimental set–up and experiments for the (a) “basic” corridor (b) partially bifurcated corridor

675

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

The same basic corridor (using the four corridor geometries shown before) was used to study pedestrian behavior when flow is bi–directional. The overall density is still kept at around 0.6 persons/m2, the only difference is that approximately half of them travel in a direction opposite to the other half. The ROLDl for bi– directional flow is much the same way as in uni–directional case.

( )

( )

0.4

0 .4

0.3 RODLDl

0 .3

0.2

RODLDl

0 .2 0 .1

0.1

0

0 1

2

3

4

5

1

2

3

Legend: L R

l (a)

4

5

l (b)

Figure 7: RODLDl variations for the (a) basic (b) partially bifurcated corridors

0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0

(a ) RZDLDl,us

RZDLDl,fus

In order to understand the variation of the relative directional density among the five lanes the RODLDl versus l is plotted in Figure 8. Recall that l denotes the lane number with 1 as the leftmost lane. Hence, in the bi– directional case, since leftmost is direction dependent, l = 1 for (L R) is physically the same lane as l = 5 for the (R L) case. This is different from the uni–directional case where a specific value of always indicates a specific physical lane. From Figure 8 (a), (b) the following can be observed Pedestrians avoid the lanes to their right a bit more than in the uni–directional case possibly because in the bi–directional case pedestrians in a given direction feel that they only have width equivalent to two to three lanes from the left (as opposed to width equivalent to five lanes in the uni–directional case) when they move. That is, given that pedestrians feel they have very little choice to move around laterally, the impact of variation in geometry on the lateral rearrangement of pedestrians is minimal. Because of the previous fact the first and second lanes from the left (` = 1 and 2) have higher occupancy and this goes down very rapidly as one moves to the right. The impact of the edge is also seen through the fact that RODLD1(a b) is less than RODLD2(a b).

1

2

3 l

4

5

0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0

(b )

1

2

3 l

4

5

676

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

(c ) RZDLDl,ds

RZDLDl,c

0 .7 0 .6 0 .5 0 .4 0 .3 0 .2 0 .1 0 1

2

3

4

5

(d )

1

2

RZDLDl,fds

l 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

3 l

4

5

(e ) Legend: L R BDLDV

1

2

3 l

4

5

Figure 8: RZDLDl,z variations for partially bifurcated corridor at (a) far upstream (b) upstream (c) constricted (d) downstream and (e) far downstream

1 .5

1 .5

1

1

Speed (m/s)

Speed (m/s)

It presents the variations in RZDLDl,z for the five different zones of the partially bifurcated corridor. Results from the speed data (Figures 10 (a), (b)) show that the average speeds for the basic corridor are about 7% lower in the bi–directional case than the uni–directional case; this is primarily due to greater friction between pedestrians moving in opposing directions. In the case of symmetric and partially bifurcated constrictions the speed again falls by about 7% from that of the uni–directional case.

0 .5 0 0

10

20

30

0 .5 0 0

10

20

30

Distance from entry (m) Distance from entry (m) (a) (b) Figure 9: Longitudinal speed profile in the case of bi–directional flow for the (a) basic (b) partially bifurcated corridors

Ujjal Chattaraj et al. / Procedia - Social and Behavioral Sciences 104 (2013) 668 – 677

677

3. Conclusion In this study experiments are conducted to understand impacts of obstacle on pedestrian movement on corridors. Existing literature on pedestrian dynamics is quite abundant in studies at macroscopic scale where observations are recorded at (i) studies on speed, density and their interrelationship, and (ii) studies on the different phenomena that can be observed when studying pedestrian dynamics. The experiments are intended to study the relationship between distance headway (or linear density) and speed for pedestrian motion and the experiments on pedestrian motion in corridors are designed to understand the lateral and longitudinal variations in density and speed of pedestrian streams which allow overtaking. These experiments are aimed at studying the effect of geometric variations in corridors on pedestrian motion. The experiments from wide corridors with varying geometry yield information on effect of corridor edges on pedestrian behaviour, effect of narrowing on pedestrian behaviour both upstream and downstream of constriction, and effect of opposing flow on pedestrian behaviour. The main contributions of this study are determination of the fundamental qualities of pedestrian behaviour for Indians, large scale empirical observations on pedestrian behaviour in a variety of different situations at the mesoscopic scale. References Chattaraj, U., Seyfried, A. and Chakroborty, P. (2009). Comparison of Pedestrian Fundamental Diagram Across Cultures. Advances in Complex systems, 12(3), pp. 393–405. Chattaraj, U., Chakroborty, P. and Seyfried, A. (2010a). Empirical Studies on Pedestrian Motion Through Corridors of Different Geometries. In Proceedings of the 89th Annual Meeting of the Transportation Research Board, Washington, D.C, USA. Chattaraj, U., Seyfried, A. and Chakroborty, P. (2010b). Understanding Pedestrian Motion Across Cultures: Experiments and Modelling. In Proceedings of the 8th Conference on Traffic and Granular Flow (to appear), Shanghai, China. Hankin, B.D. and Wright, R.A. (1958). Passenger Flow in Subways. Operational Research Quarterly, 9(2), pp. 81–88. Helbing, D., Farkas, I.J., Molnar, P. and Vicsek, T. (2002). Simulation of Pedestrian Crowds in Normal and Evacuation Situation. In Pedestrian and Evacuation Dynamics, (Eds.: Schreckenberg, M. and Sharma, S.D.), Springer, Berlin, Heidelberg, Germany, pp. 21–35. Helbing, D., Bunza, L., Johansson, A. and Werner, T. (2005). Self-Organized Pedestrian Crowd Dynamics: Experiments, Simulations and Design Solutions. Transportation Science, 39(1), pp. 1–24. Helbing, D., Johansson, A. and Al-Abideen, H.Z. (2007). Dynamics of Crowd Disasters: An Empirical Study. Physical Review E, 75(4), pp. 046109 (1–7). Helbing, D. and Molnar, P. (1995). Social Force Model for Pedestrian Dynamics. Physical Review E, 51(5), pp. 4282–4286. Helbing, D., Farkas, I.J., Molnar, P. and Vicsek, T. (2002). Simulation of Pedestrian Crowds in Normal and Evacuation Situation. In Pedestrian and Evacuation Dynamics, (Eds.: Schreckenberg, M. and Sharma, S.D.), Springer, Berlin, Heidelberg, Germany, pp. 21–35. Henderson, L.F. (1971). The Statistics of Crowd Fluids. Nature, 229(5284), pp. 381– 383. Henderson, L.F. (1974). On the Fluid Mechanics of Human Crowd Motion. Transportation Research, 8(6), pp. 509–515. Henderson, L.F. and Lyons, D.J. (1972). Sexual Differences in Human Crowd Motion. Nature, 240(5380), pp. 353–355. Hoogendoorn, S.P. and Daamen, W. (2004). Self–Organization in Walker Experiments. In Proceedings of the 5th Symposium on Traffic and Granular Flow, (Eds.: Hoogendoorn, S.P., Luding, S., Bovy, P.H.L., Schreckenberg, M. and Wolf, D.E.), Springer, Delft, The Netherlands, pp. 121–132. Hoogendoorn, S.P. and Daamen, W. (2005). Pedestrian Behavior at Bottlenecks. Transportation Science, 39(2), pp. 147–159. Brennan, L., Mcdonald, J. and Shlomowitz, R. (1995). The Variation in Indian Height .Man in India, 75(4), pp. 327–337.