You are on page 1of 12

3rdConference of Transportation Research Group of India (3rd CTRG)

Perception Based Level of Service for Uncontrolled


Midblock Crossing
Vivek R Dasa*, Satish Chandrab, Rajat Rastogic
a
Professor, Department of CTM & Highway Technology, Dayanandasagar College of Engineering, Bangalore, India, vivekdurgadath@gmail.com
(Corresponding author)
b
Professor, Department of Civil Enginnering, IIT Roorkee, Roorkee, India, satisfce@iitr.ac.in
c
Associate Professor, Department of Civil Engineering, IIT Roorkee, Roorkee, India, rajat.iitr@gmail.com

Abstract - Studies conducted in various Indian cities suggest that walking is a predominant mode of transport. Most
of the trips within the city areas are walking trips where walking either act as a feeder mode to other modes of
transportation like car, bus, rickshaw etc., or is used as main mode to work. This pedestrian manoeuvre in an urban
environment is mainly of two types: one interacting with traffic while crossing the road and other interacting with
traffic as on sidewalks. Out of these, crossing is more important as it increases the vulnerability of pedestrians to
accidents. It has been mainly observed that traffic conflicts and pedestrian delay have effect on pedestrians comfort
and safety as they cross. Pedestrian waiting time data were collected from 17 locations in five cities of India namely
Chandigrah and New Delhi in North India and Chennai, Coimbatore and Erode in South India. A K-means
clustering technique was used to define the waiting time ranges and cluster parameter Silhouettes, was used to fix
the number of clusters and to check whether the data structure lies within the clusters or not. The waiting time values
corresponding to these cluster centres were used to define the initial LOS service criteria. The developed level of
service was further modified using pedestrian perception. The variation observed during the perception survey is
used to revise the LOS criteria. From the study it was observed that average waiting time at all locations was below
23s. Pedestrian perception survey revealed that the acceptable value of waiting times varies from 2 to 10.5 s for
different category of pedestrians.

Key Words – Pedestrians, Level of Service, Midblock, Traffic

1. INTRODUCTION
Pedestrian environment is multi-dimensional. It has been mainly observed that traffic
conflicts and pedestrian delay have effect on pedestrians‟ comfort and safety as they cross the
crosswalks. HCM (2000) has used pedestrian delay as a criterion to define pedestrian level of
service (LOS) for crossing facilities. Delay is used as a basic criterion to define LOS for
pedestrian crossings. Measurement of pedestrian delay in field is relatively easier if the crossing
conditions are well defined and controlled, as may be available in developed countries. In
developing nations like India the task of measuring pedestrian delay at a crossing site is difficult
due to scattered crossing pattern and poor regulatory structure. The pedestrians after starting the
crossing maneuver at the road, at times, change their crossing pattern due to wrong judgment of
traffic conditions prevailing on the road. This makes the estimate of delay biased towards higher
values and poor LOS. Similarly, an approaching conflicting vehicle may cause the pedestrian to
almost run condition to cross the road. It may result in negative values of delay also. On the other
hand, a pedestrian waiting at the curb for an acceptable gap in the traffic stream will take a
decision to cross the road only when she/he feels that it is possible to complete the crossing in
the available gap. Once she/he enters the road, she will stick to her/his decision. Availability of
gap and decision of the pedestrian will depend on the traffic flow, the vehicle speed, the width of
the road and availability of the median. Therefore, waiting time at the curb is a more appropriate
measure to define the LOS under untidy traffic and pedestrian behaviour. It is easy to measure
and better perception of quality of facility from pedestrian‟s view point. Therefore use of waiting
time at curb is proposed in the present study to define level of service. Another approach used to
3rdConference of Transportation Research Group of India (3rd CTRG)

define LOS for crossing condition is based on the questionnaire survey (Petrish et al. 2006, Chu,
et al. 2003, Landis et al. 2005). In such studies pedestrians‟ perception with respect to delay
while crossing a road is used as a measure of LOS. In the present study, waiting time is used as a
criterion to define the LOS for pedestrian crossing conditions. Cluster analysis was used to
identify the LOS ranges based on waiting time. These are further modified by the results of
pedestrians‟ perception survey.
2.LITERATURE REVIEW
Different methods are reported in literature for computing pedestrian delay and LOS at
pedestrian crossing. These are presented in the following paragraphs.
Roddin (1981) described a method for calculating moderate (less than 18 s) mean
pedestrian delay at unsignalized intersections. He found that delay increases linearly with traffic
volume. Griffiths et al. (1984) performed extensive simulation analysis on zebra crossings.
They found that pedestrian delay depends heavily on both pedestrian and vehicle flows and the
effect of increasing vehicle flow occurred primarily at low pedestrian volumes. Smith et al.
(1987) demonstrated the effect of crossing width and conflicting vehicle volume on pedestrian
delay. It was found that the delay increases with increase in the traffic volume. This is contrary
to the findings of Griffiths et al. (1984). The delay increases with an increase in road width as
well. The percentage increase in delay is more for higher road widths especially above 11 m and
for traffic volume higher than 800 veh/hr. Virkler (1996) developed a model for calculating
delay based on queuing theory. Delay was found to increase with traffic volume as a polynomial
function of power two. According to HCM (2000), the average delay per pedestrian for a
crosswalk depends on critical gap and traffic flow.
LOS is a measure-of-effectiveness by which traffic engineers determine the quality of
service of elements of transportation infrastructure. HCM (2000) describes LOS mainly based
on pedestrian delay. LOS is divided into six groups depending on two parameters, delay and
likelihood of risk taking behaviour in the case of unsignalized crossing. Chu and Baltes (2002)
developed a LOS methodology for pedestrian crossing streets at mid-block locations. Thirty-
three mid-block locations in Tampa and St. Petersburg, Russia were identified and a total of 96
people were hired to test the mid-block crossings. The participants at each site were asked to
observe mid-block crossings for 3-minute periods and then rate the difficulty of crossing on a
six-point scale (A to F). A total of 767 observations were made. Results showed that the level of
crossing difficulty tended to increase with the width of painted medians, signal spacing, and
turning movements, and that the presence of pedestrian signals lowered the perception of
crossing difficulty. The presence of pedestrian signals and cycle length were also shown to be
statistically significant. The final linear regression model had an R-square value of 0.34 and
contained 15 variables relating to traffic volumes, turning volumes, age of pedestrian, average
vehicle speed, crossing width, presence of pedestrian signal, cycle length, and signal spacing.
Chu et al. (2003) placed 86 people at 48 intersection and mid-block locations and asked them to
identify one of six routes they might take to cross the street. They obtained a total of 1,028
observations of 4,334 cases and fitted a 2-level nested logit model to the survey responses.
Muraleetharan et al. (2004) proposed a method to estimate the overall level-of-service of
sidewalks and crosswalks on the basis of total utility values, which come from a conjoint
analysis. Pedestrians were surveyed at four locations around Hokkaido University, Japan, and
their perceptions of the ease of walking on a sidewalk or crosswalk were scaled from 1 to 10. A
R Das, Chandra, Rastogi
2
3rdConference of Transportation Research Group of India (3rd CTRG)

linear relationship was assumed between the total utility of a sidewalk/crosswalk and the overall
level-of-service of that sidewalk/crosswalk. Kadali and Vedagiri (2015) studied pedestrian
perceived LOS with respect to data collected from different land-use type such as shopping,
residential and business areas. The ordered probit (OP) model was developed by using NLOGIT
software package, with number of vehicles encountered, road crossing difficulty as well as safety
considered as primary factors along with pedestrian individual factors (gender and age), land-use
type and roadway geometry. From the model results, they concluded that perceived safety,
crossing difficulty, land-use condition, number of vehicles encountered, median width and
number of lanes have significant effect on pedestrian perceived LOS at unprotected (un-
signalized) mid-block crosswalks in mixed traffic scenario. Pedestrian road crossing is also
incorporated in multi-modal LOS analyses (Winters et al., 2001). In several related studies,
measures of effectiveness for crossing at junctions are proposed (Crider et al., 2001; Sarkar,
1995). Baltes and Chu (2002) proposed difficulty to cross as a measure of effectiveness for mid-
block locations. Pedestrians‟ road crossing options were seen as measures of accessibility to
transit by Phillips et al. (2001).
3. STUDY LOCATIONS AND DATA COLLECTION
The selection of study locations is governed by two aspects; selection of a city for
conducting the pedestrian study and selection of study location within the city. The cities are
identified based on their population and possibility of high pedestrian movement. The cities
identified are Chandigarh and New Delhi in North India and Chennai, Coimbatore and Erode in
South India. The study locations within a city are chosen based on width of the road (number of
lanes), provision of medians and type of land use on adjacent land. The selected study locations
are categorized as having two lanes or three lanes with or without median. The details are given
in Table 1. Data was collected from videographic survey and pedestrian waiting time was
extracted from the videograph.
Table 1: Details of study location
Sl Width of Type of Traffic
Location(ID) Land use Road Type
No road, m control (PCU/hr)
Chandigarh
1 Rock Garden (1) Recreational 12.8 3 – lane two way No 1634
2 Sukna Lake (2) Recreational 13.6 3 – lane two way No 1569
3 Sector 17 (3) Commercial 10.6 2 – lane one way Zebra 892
Inter State Bus Terminus
4 Mixed 9.0 2 – lane one way No 1586
(ISBT) (4)
Post Graduate Institute of
5 Educational 9.5 2 – lane one way No 2191
Medicine (PGI) (5)
Chennai
6 TH Road (6) Shopping 11.6 2 – lane two way No 2834
7 Old Washemanpet (7) Mixed 7.25 2 – lane one way Zebra 1417
8 Kaladipet (8) Shopping 10.0 2 – lane two way No 2510
Commercial/
9 Near Marina Beach (9)
Recreational 7.0 2 – lane one way Zebra 1600
Coimbatore
10 Gandhipuram (10) Mixed 7.0 2 – lane one way Zebra 2395
R Das, Chandra, Rastogi
3
3rdConference of Transportation Research Group of India (3rd CTRG)

11 Near Govt. Hospital (11) Mixed 7.8 2 – lane one way Zebra 2079
12 Near City Bus Stand (12) Mixed 7.8 2 – lane one way Zebra 1663
13 100th ft Road (13) Mixed 7.0 2 – lane one way No 2203
New Delhi
14 Connaught Place (14) Shopping 8.4 2 – lane one way Zebra 1658
15 Near Kalkaji Temple (15) Recreational 10.4 2 – lane one way Zebra 2482
16 Near Andra Bhawan (16) Residential 9.6 2 – lane one way No 2858
Erode
17 Near CSI School (17) Educational 8.6 2 – lane one way No 2243

4. WAITING TIME ANALYSIS


Pedestrians‟ time in waiting before crossing the road plays a key role in decision making
process. Long waiting time will make the pedestrians impatient and will encourage the illegal
crossing. This may also result in chaos on the road and finally results in accidents. Waiting time
indicates an excess time consumed on a transportation facility in comparison to a reference
value. It is a performance measure directly experienced by a road user. Waiting time has also
been widely accepted as the key performance index for determining LOS at crossing. The
following sections present the analysis of waiting time at different location selected for the
present study.

4.1 AVERAGE WAITING TIME


Table 2 gives the summary of waiting time (in seconds) at various locations. For 2 – lane
two way road system values were found to vary from 3.69 to 22.4 s. Three trends were observed
for this type of road system. One exceptionally low value at location 7 (3.69 s), moderate value
at locations 3, 4, 5, 9, 10, 12, 13, 15, 16 and 17 (6 to 14 s) and exceptionally high value of 22.4 s
at location 14. Low value of waiting time at location 7 is mainly due to high percentage of single
stage crossings (82 %) at this location. High value of waiting time at Location 14 is due to
commercial/ shopping nature of the area and pedestrians waited until they found a safe
acceptable gap. Skewness values at all the location are positive indicating that bulk of values lies
to the left of the mean value. This shows that the pedestrians waiting time is generally low in
mixed traffic conditions. The data collected from all location are clubbed to develop LOS
criteria based on clustering.

Table 2 Average pedestrian waiting time (s) at various locations

Std.
ID Location Mean Skewness Min Max
Deviation

2 – lane one way system


3 Sector 17 13.14 2.48 0.08 9.62 16.28
4 ISBT 11.14 5.02 0.10 2.12 19.95
5 PGI 11.29 6.31 0.87 1.26 28.71
7 Oldwashermanpet 3.69 2.62 1.11 0.71 13.54
R Das, Chandra, Rastogi
4
3rdConference of Transportation Research Group of India (3rd CTRG)

9 Near Marina Beach 12.61 10.83 1.03 1.31 37.5


10 Gandhipuram 9.33 6.82 1.98 0.63 53.17
11 Near Govt. Hospital 6.31 3.28 0.58 0.95 15.72
12 Near City Busstand 9.42 5.96 1.45 1.93 31.42
13 100 ft road 10.51 6.11 0.59 1.17 29.8
14 Connaught 22.4 19.43 1.30 1.59 70.24
15 Near Kalkaji Temple 9.99 6.43 0.81 1.79 27.45
16 Near Andra Bhawan 9.57 8.87 1.01 0.54 29.32

17 Near CSI School 10.82 4.49 0.70 3.61 21.09


Average 10.79 7.10 0.89 2.09 30.32
2 – lane two way system
6 TH road 5.99 2.24 -0.25 1.89 9.42
8 Kaladipet 4.90 3.36 1.04 0.45 17.59
Average 5.45 2.80 0.40 1.17 13.51
3 – lane two way system
1 Rock Garden * * * * *
2 Sukna Lake 3.09 1.86 1.19 0.84 9.82
Average 3.09 1.86 1.19 0.84 9.82

5. CLUSTER ANALYSIS
Clustering algorithm is used to arrive at the limits of waiting time for the different Levels
of service and this is further modified by pedestrian perception. Clustering is a statistical method
of partitioning a sample into homogeneous groups to produce an operational classification. The
greater is the similarity (or homogeneity) within a group, and the greater is the difference
between the groups, the „better‟ or more distinct would be the clustering. K-means clustering
technique was used to define the waiting time ranges and cluster validation parameter
(silhouettes) was used to fix the number of clusters and to check whether the data structure lies
within the clusters or not. Pedestrian waiting time at 17 crossing locations is considered. K-
means is a type of partition clustering algorithm which uses Euclidean distance to measure the
dissimilarity in the data sets. It is used to identify the data clusters of waiting time, which are
further defined as LOS ranges. It is explained below.
5.1K-MEANS CLUSTER ALGORITHM
K-means is one of the simplest algorithms that can solve the well known clustering
problem. The algorithm for K-means is explained below.
Step 1: Choose the number of clusters 1 < c < N and initializing random cluster centers from the
data set. Here, „c‟ is the number of clusters and „N‟ is the number of data points.
Step 2: Calculate distance matrix (Dik) from each data point (i) to kth cluster centre ( v nk ). v nk is
defined as mean for the data points over cluster „k‟ for nth iteration.
Step 3: Allocate data point to those clusters whose square of distance (Dik2 ) is minimum using
equation 1.

R Das, Chandra, Rastogi


5
3rdConference of Transportation Research Group of India (3rd CTRG)

Dik2  (xi  vkn )2 , for1  i  Nand1  k  c 1


where,
th
(Dik2 ) = distance matrix between data points (i) and the k cluster center.
xi = „ith‟ data point in cluster „k‟
Step 4: Calculate new cluster centers using equation 2:
Ni
 xi
v nk  i 1 2
Nk
Where Nk is the number of objects in the cluster „k‟. The above is iterated till vnk  vnk 1
Clustering is mostly unsupervised process, thus evaluating the result of the clustering
algorithms is very important. In the clustering process there are no predefined classes. Therefore,
it is difficult to find an appropriate value for measuring whether the cluster configuration found
during the process is acceptable or not. Several clustering validity approaches have been
developed like c-index, Dunn‟s validity index, silhouette method etc. (Legany et al.; 2006). In
the present problem, silhouette method was used to estimate the cluster validity. This is
explained in the next section.
5.2 SILHOUETTE VALIDATION METHOD
The silhouette validation technique (Rousseeuw, 1987) calculates the silhouette width for
each sample, average silhouette width for each cluster and overall average silhouette width for a
total data set. This compares the tightness and separation of each cluster. The silhouettes S(i) is
estimated using equation 3.
{b(i)  a(i)}
S(i)  3
max{a(i), b(i)}
Where,
a(i) = average dissimilarity of ith-object to all other objects in the same cluster
b(i)= minimum of average dissimilarity of ith-object to all objects in other cluster (in the closest
cluster).

Figure 1 Numbers of Clusters versus Silhouettes


The largest overall average silhouette indicates the best clustering (number of cluster).
Therefore, the numbers of clusters with maximum overall average silhouette width are taken as
the optimal numbers of the clusters. The clustering is done with SPSS package (version 17) and
cluster analysis is performed with number of clusters varying from 3 to 6. Silhouette value for
these cluster centres was found to vary from 0.55 to 0.66. Clustering with 5 clusters gave
R Das, Chandra, Rastogi
6
3rdConference of Transportation Research Group of India (3rd CTRG)

maximum overall average silhouette value of 0.66 (Refer Figure 1). The values corresponding to
five cluster centers and corresponding silhouettes for each cluster are given in Table 3. Cluster
center was found to have a lower value of 4.27 s and higher value of 101.67 s. For lower and
higher values, it was found that the values clustered very close to their centre, which is also clear
from their higher silhouette values of 0.7 and 1 respectively. In other cases it is around 0.55
which indicates that some values might have fallen in two successive ranges. The waiting time
values corresponding to these cluster centers are used to define LOS criteria in the present study.
The upper and lower limit of cluster center is taken as LOS criteria for a particular condition. It
is explained in next section.
Table 3 Details of 5 clusters

Waiting time,
Cluster centers Silhouettes
(Range)

1 4.27 (0 - 9.44) 0.70


2 14.61 (9.44 - 23.37) 0.51
3 32.13 (23.37 - 45.12) 0.54
4 58.1 (45.12 -79.89) 0.57
5 101.67 (79.89 -101.67) 1.00
Average 0.66
6. PROPOSED LOS CRITERIA
The five cluster centers are divided into six LOS categories representing very low to very
high waiting time condition at the curb side. The decimal values are rounded off to the nearest
higher integer. The LOS was split into six categories from A to F. The developed LOS criteria
based on cluster analysis are given in Table 4. The LOS criteria suggested above is based on
field data and using the technique of cluster analysis. This approach does not consider the
perception of pedestrians about these limits of waiting time. To include pedestrian perception,
the proposed LOS criterion was modified with pedestrian perception survey.
Table 4 LOS based on waiting time

LOS Waiting time (s)


A ≤9
B > 9 and ≤ 23
C > 23 and ≤ 45
D > 45 and ≤ 80
E > 80 and ≤ 102
F > 102

R Das, Chandra, Rastogi


7
3rdConference of Transportation Research Group of India (3rd CTRG)

6.1 PEDESTRIAN PERCEPTION SURVEY


The pedestrian interview survey was conducted at two locations; one in Roorkee city. and
another in Connaught place, New Delhi. Both locations are at uncontrolled midblock sections. A
total of 212 sample data were collected. The sample characteristics are given in Table 5.
Table 5 Sample characteristics
Pedestrian Characteristics Percent Sample
Male 80
Gender
Female 20
Young (Upto 20 years) 50
Age Middle Aged (20 – 50 Years) 40
Old (Above 50 Years) 10
Post Graduate 18
Graduate 25
Education th th
10 – 12 Standard 50
< 10th 7

A questionnaire was given to a pedestrian with eleven photographs depicting different


crossing conditions with varied waiting time. The questionnaire and the photographs of crossing
conditions were used for the perception survey. Data related to pedestrian characteristics like
age, gender and their maximum acceptable waiting time were collected. The traffic observed by
pedestrians during these waiting times is shown through these photographs. Among the
11photographs, four are corresponding to the waiting time of cluster centers (4s, 15s, 32s and
58s) based on clustering, five are intermediate to these (9s, 25s, 45s, 63s and 76s) and two are of
lower (no waiting time) and upper extreme cases (102s). They were asked to sort the shown
conditions from easy to very difficult to cross. A check whether the responding pedestrian has
understood the given crossing condition amply or not is made by using lower and upper extreme
cases. If this is not correct then the response was rejected. Pedestrians were asked to arrange the
photograph in the ascending order from 1 to 11 and demarcate them based on five conditions as
very low, low, satisfactory, high and very high waiting time conditions. The variation observed
from this survey is used to revise the LOS criteria developed based on cluster analysis.

1) Waiting Time 0s 2) Waiting Time 4 s

R Das, Chandra, Rastogi


8
3rdConference of Transportation Research Group of India (3rd CTRG)

3) Waiting Time 9 s 4) Waiting Time 14 s

5) Waiting Time 25 s 6) Waiting Time 32 s

7) Waiting Time 45 s 8) Waiting Time 58 s

R Das, Chandra, Rastogi


9
3rdConference of Transportation Research Group of India (3rd CTRG)

9)Waiting Time 63 s 10) Waiting Time 76 s

11) Waiting Time 102 s

Figure 3 Photos for pedestrian perception survey

6.2ACCEPTABLE WAITING TIME


The minimum and maximum acceptable waiting time for different pedestrians is
calculated based on 15th and 85th percentile from cumulative distribution curve. It was found that
a waiting time of 4 s will be accepted to 85 percent pedestrians. Female pedestrian preferred less
waiting time than male. The 85 percent female pedestrians preferred a lower waiting time (3.9 s)
than male pedestrians (4.1s). Further acceptable waiting time increased from 3.2s to 6.2s with
increase in age of the pedestrians. This shows that middle and older pedestrians have more
patience to wait for safer gaps than the lower age groups. It was also found that 85 percent
pedestrians experienced waiting time less than 9.4 s. The acceptable waiting time for different
pedestrian category is given in Table 6. The 15th and 85th percentile values indicate towards the
ranges within which the waiting time should vary (4.0 s for most ideal crossing conditions and
9.4 s for most delayed condition). These should be considered while implementing traffic
management policies at midblock locations mostly used by pedestrians for crossing the road.

Table 6 Acceptable waiting time (s)

Pedestrian Category 15th percentile 85th percentile


Overall: 4.0 9.4
Male 4.1 9.8
Gender :
Female 3.9 9.0
Young - Overall 3.2 8.8
Young - Male 2.0 9.0
Young - Female 4.3 8.5
Middle Aged - Overall 3.7 9.8
Age : Middle Aged- Male 4.0 10
Middle Aged - Female 3.4 9.5
Old - Overall 6.2 10.5
Old - Male 6.2 10.5
Old - Female Nil Nil

R Das, Chandra, Rastogi


10
3rdConference of Transportation Research Group of India (3rd CTRG)

6.3 PEDESTRIAN PERCEPTION ANALYSIS


The perception of pedestrians for each type of crossing condition with respect to different
waiting times is given in Table 7. It was found that as waiting time condition changes from very
low to very high, the waiting time shift diagonally from left to right (i.e. from low value to high
value). Higher interval (15s – 58s) is observed for condition satisfactory and lower interval (0s –
9s) for very low waiting time. The cumulative frequency distribution for each type of crossing
was plotted and the 85th percentile waiting time for each condition is taken as the perceived
maximum waiting time for that condition of crossing. These values are given in Table 8.
Table 7 Percentage pedestrian perception of waiting time in each type of crossing
conditions
Crossing Waiting time (s)
Condition
(waiting
time)
0 4 9 15 25 32 45 58 63 76 102
Very low 50 41 9 0 0 0 0 0 0 0 0
Low 0 9 40 39 12 0 0 0 0 0 0
Percentage
Satisfactory 0 0 0 4 28 39 23 6 0 0 0 Values
High 0 0 0 0 0 0 13 31 37 19 0
Very high 0 0 0 0 0 0 0 0 8 34 58
The 85th percentile waiting time (perceived) for each condition is used to define the upper limit
of waiting time for each category of modified LOS. The decimal values are rounded to nearest
integer. The five conditions are divided into six category of LOS from A to F. The details are
given in Table 8. For lower LOS criteria it was found that the HCM method values are higher
where as for higher LOS criteria the proposed criteria seems to be higher. With respect to cluster
analysis it was found that the precieved LOS is less.

Table 8 Perceived 85th percentile values of waiting time and Pedestrian’s Perception
modified LOS
HCM Method
th
85 Modified LOS
Condition
percentile LOS criteria
(waiting time)
perceived (Waiting time,
value (sec) s)
Very low 2.8 A ≤3 <5
Low 13 B > 3 and ≤ 13 ≥ 5 -10
Satisfactory 38 C > 13 and ≤ 38 >10-20
High 64 D > 38 and ≤ 64 >20-30
Very high 90 E > 64 and ≤ 90 >30-45
F and > 90 >45

R Das, Chandra, Rastogi


11
3rdConference of Transportation Research Group of India (3rd CTRG)

7. SUMMARY
A criterion based on cluster analysis and pedestrian perception for development of LOS
is explained in this paper. Cluster analysis was used to develop LOS criteria initially. Details of
cluster analysis and its algorithm were explained. Silhouette value is used for the validation and
selection of clusters. The LOS based on cluster analysis is further modified with pedestrian‟s
perception. Details of the survey conducted and the preliminary analysis are also explained. It
was found that average waiting time at all locations is below 23 s. . The 85 percent female
pedestrians preferred a lower waiting time (3.9 s) than male pedestrians (4.1s). Further
acceptable waiting time increased from 3.2s to 6.2s with increase in age of the pedestrians. The
level of service criteria developed with respect to pedestrian perception can be used in the
analysis of crossing locations in India.

8. REFERENCES
[1] Chu, X. and Baltes, R. M. (2003), “Measuring Pedestrian Quality of Service for Midblock
Street Crossings Selection of Potential Determinants,” Transportation Research Record:
Journal of Transportation Research Board, No 1828, pp.89 -97.
[2] Chu, X., Gutten Plan,M. and Baltes, R. M. (2003) “Why people cross when they do – The
Role of Street Environment”, Transportation Research Record: Journal of Transportation
Research Board, No 1878, pp.3-10.
[3] Crider, L., Burden, J., and Han, F. (2001), Multi-modal level of service – „point‟ level of
service project – final report. FDOT – Florida Department of Transportation, 2001.
[4] Griffiths, J. D., Hunt, J. G. and Marlow, M. (1984), “Delays at Pedestrian Crossings: Site
Observations and the Interpretation of Data”, Traffic Engineering and Control, 25, pp. 365–
371.
[5] Highway Capacity Manual (2000), Special Report No. 209, Transportation Research Board.
National Research Council, Washington, DC, USA.
[6] Muraleetharan, T., Takeo, A., Toru, H., Kagaya, S., and Kawamura, S. (2004), Method to
determine overall level-of-service of pedestrians on sidewalks and crosswalks based on total
utility value. In Proceedings of the 83rd TRB annual meeting, transportation research board,
Washington, 2004
[7] Phillips, R., Karachepone, J., & Landis, B. (2001). Multi-modal quality of service project –
Final Report. FDOT – Florida Department of Transportation, FDOT Contract #BC205, 2001.
[8] Roddin, M. (1981), “A manual to determine benefits of separating pedestrians and vehicles,”
NCHRP Report 240, Transportation Research Board, Washington DC, USA.
[9] Raghuram Kadali, P. Vedagiri (2015) Evaluation of pedestrian crosswalk level of service
(LOS) in perspective of type of land-use Transportation Research Part A: Policy and
Practice, Volume 73, March 2015, Pages 113-124
[10] Sarkar, S. (1995). Evaluation of safety for pedestrians at macro- and microlevels in urban
areas. Transportation Research Record No. 1502.
[11] Smith, S. A., Opeila, K. S., Impett, L. L., Pietrucha, M. T., Knoblauch, R. and Kubat, C.
(1987) “Planning and Implementing pedestrian facilities in suburban and developing rural
areas,” NCHRP Report 249A. Transportation Research Board, Washington DC, USA.
[12] Virkler, M. (1996), “Pedestrian Compliance effects on signal delay,” Annual Meeting
TRB, Transportation Research Board, Washington D C, USA.

R Das, Chandra, Rastogi


12

You might also like