Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Available

Available online
online at
at www.sciencedirect.com
www.sciencedirect.com

ScienceDirect
Available online at www.sciencedirect.com
Transportation
Transportation Research
Research Procedia
Procedia 00
00 (2017)
(2017) 000–000
000–000
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
Transportation Research Procedia 25 (2017) 1342–1351
www.elsevier.com/locate/procedia

World Conference on Transport Research - WCTR 2016 Shanghai. 10-15 July 2016

Analysis on drivers’ parking lot choice behaviors


in expressway rest area
Shinji Tanaka aa, Shinya Ohno aa, Fumihiko Nakamura aa *
a
a Yokohama National
Yokohama National University,
University, 79-5
79-5 Tokiwadai
Tokiwadai Hodogaya-ku
Hodogaya-ku Yokohama,
Yokohama, 240-8501,
240-8501, Japan
Japan

Abstract
Abstract

In
In high
high demand
demand seasons,
seasons, heavy
heavy traffic
traffic causes
causes congestion
congestion not not only
only on
on main
main carriageways
carriageways but but also
also in
in rest
rest areas.
areas. Congestion
Congestion in in rest
rest
areas
areas make vehicles cruising,
make vehicles cruising, crawling,
crawling, standing
standing and and sometimes
sometimes reverse
reverse running
running inside
inside the
the area
area to to find
find aa parking
parking lot.lot. Such
Such
vehicles
vehicles form
form aa queue
queue andand it
it may
may extend
extend out
out to
to the
the main
main carriageway,
carriageway, which
which cause
cause more
more serious
serious congestion.
congestion. At At the
the same
same time,
time, it
it
is
is often
often observed there are
observed there are still
still some
some available
available lots
lots left
left in
in the
the back
back side
side of
of the
the area
area even
even though
though such
such information
information is is provided.
provided. It It
means
means drivers
drivers dodo not
not always
always behave
behave perfectly
perfectly and
and the
the capacity
capacity of of the
the rest
rest area
area is
is not
not fully
fully utilized.
utilized. Therefore,
Therefore, it it is
is important
important to to
understand
understand howhow drivers
drivers make
make decisions
decisions ofof their
their behavior
behavior to to select
select parking
parking lots
lots under
under congested
congested situation
situation to to improve
improve the the
performance
performance of of the
the parking
parking capacity.
capacity. Then,
Then, this
this study
study aims
aims atat revealing
revealing drivers’
drivers’ choice
choice behavior
behavior of of parking
parking lots
lots inin expressway
expressway
rest
rest area.
area.
Four rest areas
Four rest areas were
were picked
picked up up toto collect
collect drivers’
drivers’ choice
choice behaviors
behaviors considering
considering the
the characteristics
characteristics such
such as as geometry,
geometry, information
information
provision etc.
provision etc. Video
Video observation
observation was was conducted
conducted to to obtain
obtain individual
individual vehicles
vehicles trajectories
trajectories asas well
well asas the
the transition
transition of of the
the parking
parking
lots occupancy. VMS data showing parking lots availability was also gathered to check what kind of information was given when
lots occupancy. VMS data showing parking lots availability was also gathered to check what kind of information was given when
aa driver
driver made
made aa choice
choice behavior.
behavior. At At the
the same
same time,
time, questionnaire
questionnaire survey
survey was
was also
also done
done toto understand
understand drivers’
drivers’ preference,
preference,
behavioral factors etc.
behavioral factors etc.
Based
Based onon the
the collected
collected video
video and
and questionnaire
questionnaire data,
data, drivers’
drivers’ choice
choice behaviors
behaviors atat diverging
diverging points
points inside
inside the the area
area were analyzed
were analyzed
in detail. As a result, major influential factors were revealed to be “distance to the facilities”, “availability
in detail. As a result, major influential factors were revealed to be “distance to the facilities”, “availability information on VMS” information on VMS”
and
and “queue
“queue length
length inin sight”.
sight”. Moreover,
Moreover, the the order
order ofof their
their importance
importance waswas not
not always
always the
the same
same butbut varied
varied depending
depending on on the
the
condition
condition ofof the
the parking
parking lots
lots occupancy.
occupancy. Especially,
Especially, at at near-saturated condition, drivers
near-saturated condition, drivers were
were affected
affected moremore by by aa queue
queue inin front
front
rather
rather than
than VMS
VMS information.
information. Finally,
Finally, some
some implications
implications for for rest
rest area
area design
design and
and operation
operation were
were shown
shown based
based onon the
the results.
results.
© 2017 The Authors. Published by
© 2017 The Authors. Published by Elsevier B.V. Elsevier B.V.
© 2017 The Authors.
Peer-review under Published by
responsibility of Elsevier
WORLD B.V.
CONFERENCE ON TRANSPORT RESEARCH SOCIETY.
Peer-review under
Peer-review under responsibility
responsibility of of WORLD
WORLD CONFERENCE
CONFERENCE ON ON TRANSPORT
TRANSPORT RESEARCH
RESEARCH SOCIETY.SOCIETY.

Keywords:
Keywords: Parking
Parking lot
lot choice;
choice; Expressway
Expressway rest
rest area;
area; Logit
Logit model
model

*
* Corresponding
Corresponding author.
author. Tel.: +81-45-339-4032; fax:
Tel.: +81-45-339-4032; fax: +81-45-331-1707.
+81-45-331-1707.
E-mail address:
E-mail address: stanaka@ynu.ac.jp
stanaka@ynu.ac.jp

2214-241X
2214-241X ©© 2017
2017 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier B.V.
B.V.
Peer-review
Peer-review under
under responsibility
responsibility of
of WORLD
WORLD CONFERENCE
CONFERENCE ON ON TRANSPORT
TRANSPORT RESEARCH
RESEARCH SOCIETY.
SOCIETY.

2352-1465 © 2017 The Authors. Published by Elsevier B.V.


Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY.
10.1016/j.trpro.2017.05.158
Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351 1343
2 Author name / Transportation Research Procedia 00 (2017) 000–000

1. Introduction

In high demand seasons, heavy traffic demand causes congestion in expressways not only on main carriageways
but also in rest areas. Congestion in rest areas make vehicles cruising, crawling, standing and sometimes reverse
running inside the area to find a vacant parking lot. It generates delay as well as risk of accidents. And these vehicles
form a long queue and it may extend out to the main carriageway, which cause more serious congestion of the
expressway mainline traffic. At the same time, it is often observed there are still some available lots left in the back
side of the rest area even though such availability information is provided via variable message signs (VMS). It
means drivers do not always behave perfectly according to the provided information and as a result the capacity of
the rest area is not fully utilized. Therefore, it is important to understand how drivers make decisions of their
behaviors to select parking lots under congested situation in order to improve the performance of the parking
operation. Then, this study aims at revealing drivers’ choice behavior of parking lots in expressway rest area.

Fig. 1. (a) Congestion in expressway rest area, and (b) queue to the mainline.

2. Research review

There are several existing researches in the context of drivers’ parking lot choice behaviors.
Young (1986) developed a models to describe drivers’ behaviors and vehicles movement inside parking garage,
and developed a simulation system called PARKSIM. They considered influencing factors, such as travel time to
parking spot, walking time to desired destination, ease of parking and so on.
Thompson et al. (1998) considered parking search behaviors by motorists in a road network and developed a
model that evaluates car park and determines routes. They applied this model to a hypothetical CBD network and
showed the effect by calculating travel time and disutility.
Van der Waerden et al. (2003) evaluated parking lot choice behavior in Eindhoven, the Netherlands. They
developed a nested logit model to describe the choice behaviors, using explanatory variables such as distance to the
ticket machine, distance to the shopping or business area etc.
Chen et al. (2011) assumed parking guidance and information system (PGIS) and developed a model to choose
optimal parking space that employ fuzzy multiple attribute decision making. They showed efficiency of the proposed
model by applying it to an underground parking garage.
However, these researches focus on behaviors in off-street parking garage, generally. There is no work dealing
with expressway rest area parking behaviors, which has some different characteristics from normal off-street parking,
such as its large scale, real-time information by VMS, etc. Therefore it is worth considering parking lot choice
behaviors in expressway rest areas.
1344 Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351
Author name / Transportation Research Procedia 00 (2017) 000–000 3

3. Methodology

3.1. Study area

Four rest areas along a same expressway in the southwest of Tokyo were picked up to collect data regarding
drivers’ choice behaviors. The characteristics such as geometry, information provision etc. are considered to select
these areas. Their locations are shown in Fig. 2 and their major features are shown in Table 1. The major difference
of these areas is existence of recursive lane and VMS (Variable Message Sign). Here, recursive lane means a lane
that connects the exit of a parking area and its entrance to make a loop as shown in Fig. 3. It can offer an additional
opportunity for drivers who could not find any vacant parking lot to go around again. VMS is generally installed at
diverging points of the entering approach so as to and provides information of parking lot availability ahead of the
point. It shows the availability by three levels, that is, “vacant (green)”, “congested (orange)” and “full (red)”.
Rest area 1 and 2 are large scale areas that are located approximately 50km away from Tokyo. There are no
recursive lane in these areas and only area 2 has VMS. Rest area 3 is medium size and it is located approximately
80km away from Tokyo. It has several recursive lanes in different parts of the area but does not have VMS. Rest
area 4 is relatively small scale that is located approximately 100km away from Tokyo. It has both recursive lane and
VMS.

Table 1. Major characteristics of the study areas.


(1) Ebina (Westbound) (2) Ebina (Eastbound) (3) Ashigara (Westbound) (4) Numazu (Westbound)
Parking
544 441 318 98
capacity (veh)
Recursive lane No No Yes Yes
VMS (num) No Yes (8) No Yes (2)

Fig. 2. Location of the study areas.


Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351 1345
4 Author name / Transportation Research Procedia 00 (2017) 000–000

Fig. 3. Geometry of the study areas (1) Ebina (Westbound); (2) Ebina (Eastbound); (3) Ashigara (Westbound); (4) Numazu (Westbound).

3.2. Data collection

3.2.1. Video recording


Video observation was conducted to obtain overall indices of the parking area as well as individual vehicles
movement. Several video cameras were set to cover from the rest area entrance to the passenger car parking block.
The observation was done from 8:00 to 16:00 in each area. Then, data was extracted from the recorded video
images. The overall area-specific indices are, transition of the parking lots occupancy rate, inflow traffic volume to
each aisle within the area, queue length at the entrance of the area. On the other hand, individual vehicle movement
data are, position of the selected parking lot, route and travel time from the area entrance to the parking lot, parking
direction to a lot (front/reverse).

Fig. 4. Example of video image.


1346 Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351
Author name / Transportation Research Procedia 00 (2017) 000–000 5

3.2.2. VMS history


VMS data showing parking lots availability was also gathered from the record of the traffic control system to
check what kind of information was given when a driver made a choice behavior at a diverging point. Although area
1 and 3 do not have VMS, area 2 has 8 VMSs and area 4 has 2 VMSs, and the historical record of these VMSs were
collected.

Fig. 5. Example of VMS (showing (a) full; and (b) vacant).

3.2.3. Questionnaire
At the same timing of video recording, questionnaire survey was also done to understand drivers’ preference,
behavioral factors etc. The survey was face-to-face interview for drivers who just parked their cars. The questions
are, personal attributes, purpose of the rest area use, desired position of parking lot, factors the driver considered
when making choices at diverging points, etc.

4. Results

4.1. Inflow and occupancy

Fig. 6 shows the transition of the inflow traffic volume into the observed areas. The inflow became high
gradually around lunch time, however it did not increase in the area 4. Fig. 7 shows the transition of the parking lots
occupancy in these areas. Here, parking lots occupancy is defined by Equation (1). The occupancy also became high
as it approached lunch time, however the occupancy of the area 4 was stable around 100%. (It sometimes exceeded
100% since some vehicles parked non-designated space (outside of the lots).) It was almost always full during the
observation period in the area 4, and that’s why the inflow did not increase in this area due to the congestion inside
the area.

234567 89 :;7<6= >6?@AB6C


𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 = ×100 (%) (1)
234567 89 :;7<@DE B8FC
Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351 1347
6 Author name / Transportation Research Procedia 00 (2017) 000–000

Fig. 6. Transition of inflow traffic volume.

Fig. 7. Transition of parking lots occupancy.

4.2. Block occupancy

As the area 1 & 2 are large size areas and they have several rows of parking lots inside, these rows are regarded
as “blocks” and the occupancy of each block was extracted from the video. Fig. 8 and 9 shows the transition of
block occupancies in these areas. In general, occupancies of the blocks close to the rest area facilities were always
high and those far from the facilities were low. As the area became congested, blocks got occupied from those closer
to the facilities one by one. This change (the speed of transition) was slower in the area 2 than in area 1, this may be
because VMS provides block availability information in the area 2.

Fig. 8. Transition of block occupancy (area 1).


1348 Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351
Author name / Transportation Research Procedia 00 (2017) 000–000 7

++9Fig. 9. Transition of block occupancy (area 2).

4.3. Lane choice ratio

Lane choice is one of the most typical choice behaviors among individual vehicles movement inside a rest area
parking. It was extracted from the video image and the lane choice ratio (the percentage of the selected lanes) in
every 30 minutes time period was obtained. Fig. 10 and 11 show the transition of the lane choice ratio in the area 1
and 2, which have typical geometric layout as expressway rest areas. The lane number was defined as 1, 2, …, n
from the facility side of the rest area. The lane choice ratio varied time by time, depending on the condition such as
block occupancy, queue on lanes, and so on. Fig. 10 showed an inverted situation around 11:30 am, which means
the lanes closer to the facility were congested then more vehicles selected the 5th and the 6th lane to avoid queueing.

Fig. 10. Transition of lane choice ratio (area 1).


Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351 1349
8 Author name / Transportation Research Procedia 00 (2017) 000–000

Fig. 11. Transition of lane choice ratio (area 2).

5. Modeling

5.1. Data

Using the vehicles movement data extracted from the video image, drivers’ lane choice behaviors at diverging
point were modeled since lane choice is influenced by so many factors among parking behaviors. As there are
several diverging points in the observed areas, the first divergence near the entrance of the rest area was considered
in order to capture more number of observed samples. Observed data in the area 2 and 4 were used since it was
assumed that lane choice behaviors were greatly influenced by VMS information. Fig. 12 shows the focused lane
choice behavior in these areas. The area 2 has 4 lane options and the area 4 has 2 lane options.

Fig. 12. Focused lane choice behaviors in the area 2 and 4.

Lane choice data in different time period was picked up in each area in order to consider different congestion
levels. As the area 4 was more congested than the area 2 throughout the observed period, data of free flow period
(9:30 – 10:00) and little congested period (12:00 – 12:30) was used in the area 2, whereas less congested period
(8:30 – 9:00) and more congested period (10:00 – 10:30) was used in the area 4.

5.2. Model structure

It is assumed that drivers choose a lane considering various factors. For example, they usually like to park their
vehicles in a parking lot as close to the facilities as possible. However, such closer blocks usually attract more
vehicles and require more time for drivers to park, so they also have to consider the level of congestion. In that sense,
we can say that drivers try to maximize their utility under given condition, which is described by several lane choice
factors. Therefore, we employed random utility maximization modeling that is known by logit model, typically.
A simple logit-based model was assumed to describe the lane choice behaviors as shown in Equation (2). It
became a multinomial logit model for the area 2 and a binary logit model in the area 4. It is possible to assume a
series of binary logit model for the area 2, however, we assumed the choice out of 4 lanes is done at the same timing
since drivers can look at all these lanes from the rest area entrance and they generally hope to get closer to the
facilities if there is no obstacles.

6M: NO
𝑃𝑃@ = , 𝑉𝑉@ = 𝑎𝑎R 𝑋𝑋R + ⋯ + 𝑎𝑎D 𝑋𝑋D + 𝑐𝑐 (2)
6M: NO

Pi: Probability to select lane i, Vi: Utility of lane i, Xi: lane choice factors, ai: parameters, c: constant

The following variables were used as lane choice factors (Xi). “Distance (m)” is the average distance from that
lane to the rest area facilities. “Occupancy (%)” is the block occupancy along that lane. “Queue (m)” is the queue
length of that lane that is visible to drivers. “VMS” is the availability information of parking lots on that lane ahead
of the diverging point. Here, VMS shows three types of information depending on the block occupancy, that is,
1350 Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351
Author name / Transportation Research Procedia 00 (2017) 000–000 9

“Available” (occupancy: 85% or less), “Congested” (occupancy: 85 – 95%) and “Full” (occupancy: 95% or more).
Then, we converted these availability information into 1, 0.15 and 0.05, respectively. The reason why both “VMS”
and “Occupancy” were used is, “VMS” is not just information of the occupancy but it includes drivers’ perception
and interpretation of the displayed sign, which may be different from continuous value of the occupancy. All
possible combination of the variables were tried and significant one was identified.

5.3. Model estimation results

Table 2 shows the results of the model parameter estimation for the model in the area 2. From this result of t-
values, we can understand that the “distance” was the most influential factor and the next was the “occupancy” in
free flow condition, however, “VMS” became more important than “distance” when it was a little bit congested. As
for 12:00 – 12:30 time period, the model became invalid when the variable “occupancy” was included.

Table 2. Estimated parameters (area 2).


9:30 – 10:00 (free flow) 12:00 – 12:30 (little congested)
Variables
estimator t-value estimator t-value
Distance (m) X1 -0.056 -5.61 -0.024 -3.42
VMS (0 or 1) X2 -3.170e-11 -3.01e-10 0.326 4.14
Occupancy (%) X3 -1.234 -2.99 – –
Constant c -3.368e-11 -1.57e-10 1.496e-11 4.44e-16
Likelihoood ratio 0.0307 0.0124
Hit ratio 0.404 0.313
Num of samples 146 259

Table 3 shows the result for the area 4. As it was already congested than the area 2 in 8:30 – 9:00 period, “VMS”
was the most influential factor. Then when it was congested more, “queue” length in sight in front of each lane
became more important than other factors.

Table 3. Estimated parameters (area 4).


8:30 – 9:00 (less congested) 10:00 – 10:30 (more congested)
Variables
estimator t-value estimator t-value
Distance (m) X1 -0.161 -3.23 -0.064 -0.98
VMS (0 or 1) X2 23.688 4.14 7.496 1.56
Queue (m) X3 -0.615 -3.42 -0.537 -3.13
Constant c 1.05e-11 4.44e-16 6.84e-11 5.77e-15
Likelihoood ratio 0.182 0.0739
Hit ratio 0.712 0.689
Num of samples 111 122

As the overall estimation indices are not so good for both models, we may need more samples or further refining
to improve the models. However, considering the results of the models in the both areas, drivers’ parking lane
choice behaviors were summarized as follows. When vehicles can move in free flow condition, drivers are
conscious of the distance to the rest area facilities. Then, as traffic increases and the area become congested, they
become more sensitive to the VMS that provides parking lots availability information ahead of their way. Finally,
the area is very congested such that vehicles have to stop frequently and VMS shows only “full” drivers pay more
attention to the queue in front of them rather than the VMS information.
Shinji Tanaka et al. / Transportation Research Procedia 25 (2017) 1342–1351 1351
10 Author name / Transportation Research Procedia 00 (2017) 000–000

5.4. Verification by questionnaire answers

The result of the model estimation was verified using the collected answers by the questionnaire survey. The
questionnaire asked drivers to put weights on five factors that might influence parking lot choice. Then it was
revealed that the order of the importance between the factors by drivers’ answers were the same even in different
rest areas, that is, (1) distance to the facilities, (2) necessary time to park, (3) width of parking lot, (4) occupancy of
the block, and (5) existence of shade. This result was consistent with the model estimation result, which showed the
distance to the facilities was the most important and the block occupancy along a lane was less important. Therefore,
the model estimation result seems reasonable from the aspect of the questionnaire answers, too.

6. Conclusions

This study aimed at revealing and modeling drivers’ parking choice behavior. Video observation and
questionnaire survey were conducted in four rest areas and the collected data was utilized to develop a model to
describe drivers’ parking lane choice behavior. The result of the model parameter estimation showed that distance to
the facilities, VMS information of parking availability, and visible queue in sight in front of a lane were significant
influential factors to lane choice behaviors, and their importance became different depending on the traffic condition.
This result was also supported by the answers from the questionnaire survey to drivers.
This study revealed some parts of drivers’ behaviors and these findings can be utilized for rest area parking
design and operation as follows. VMS information greatly affects drivers’ choice behaviors, especially when the
parking lots occupancy is changing. However, VMS become less influential in heavy congested situation because
the display does not change any more. This is because VMS shows parking lot availability only, not showing
congestion in aisle. Therefore, one possible improvement is to provide aisle information together through VMS.
Another influential factor in congested condition is visible queue length. Then, it is worth considering to make
visible queue shorter by arranging queue formation space etc.
There are some further challenges that are not considered in this study. One is to consider recursive movement
vehicles explicitly in the lane choice modeling. Their behaviors might be different because they have experienced
the same choice previously. It can be considered as drivers’ experience, but not only that, it also provides the rough
picture of the whole parking area already from the first round. Another challenge is to develop a model of individual
parking lot choice behaviors. It has much larger choice options and wider range of conditions. It may also be
necessary to consider a situation that there are more than one facilities located separately in a rest area, such as
restrooms, a shop, a restaurant and a gas station etc. Then the choice behaviors become more complicated since the
final destination of each driver is different.
By considering these behaviors, it becomes possible to describe the whole driving behaviors inside a rest area
parking and it can be further utilized for simulation models, evaluation tools, and so on. Then, these outputs can
contribute to make a design guideline of expressway rest area parking that ensures smooth and balanced operation.

References

Young, W., 1986. PARKSIM/1: A Network Model for Parking Facility Design. Traffic Engineering and Control 27, 606-613.
Warerden, P., Borgers, A., Timmermans, H., 2003. Travelers Micro-Behavior at Parking Lots: A Model of Parking Choice Behavior. TRB
Annual Meeting, CD-ROM.
Thompson, R., G., Richardson, A., J., 1998. A Parking Search Model. Transportation Research part A 32, 159-170.
Chen, M., Hu, C., Chang, T., 2011. The Research on Optimal Parking Space Choice Model in Parking Lots. International Conference on
Computer Research and Development, 93-7.

You might also like