You are on page 1of 12

Accident Analysis and Prevention 139 (2020) 105498

Contents lists available at ScienceDirect

Accident Analysis and Prevention


journal homepage: www.elsevier.com/locate/aap

Evaluating the effectiveness of new-designed crosswalk markings at T


intersections in China considering vehicle-pedestrian interaction
Yang Bian, Kun Liang, Xiaohua Zhao*, Haijian Li, Liping Yang
Beijing Key Laboratory of Traffic Engineering, College of Metropolitan Transportation, Beijing University of Technology, Beijing, 100124, China

A R T I C LE I N FO A B S T R A C T

Keywords: Crosswalk markings are a type of facility installed at the vehicle-pedestrian interaction locations and the
Crosswalk markings function is to warn drivers to watch out for pedestrians crossing the street and improve safety for pedestrians. In
Effectiveness Beijing, a type of new-designed crosswalk markings in China (NCMC) was installed. However, evaluating the
Interaction effectiveness of this type of crosswalk markings was not conducted. Accordingly, the objective of this paper is to
TOPSIS
evaluate the effectiveness of this type of new-designed crosswalk markings. During the evaluation process, the
Driving simulator
vehicle-pedestrian interaction was considered; standard crosswalk markings in China (SCMC) were taken as a
control group. In addition, empirical data were collected from a driving simulator, and nine evaluating in-
dicators representing vehicle operating data, drivers’ maneuvering data and drivers’ subjective evaluation were
proposed. In order to combine nine indicators, a Technique for Order of Preference by Similarity to Ideal
Solution (TOPSIS) method was used in this study to achieve the premium degrees of these two types of crosswalk
markings. The evaluation result showed that, for intersections with high or low pedestrian flow, the compre-
hensive effectiveness and influences on drivers’ driving behaviors with presence of NCMC were better than those
with presence of SCMC, no matter where vehicle-pedestrian interactions occurred. For intersections with no
pedestrians, the comprehensive effectiveness and influences on drivers’ driving behaviors with presence of
NCMC were worse than those with presence of SCMC, no matter where vehicle-pedestrian interaction occurred.
These results may provide references for facility installing and future development of standards.

1. Introduction pedestrians at intersections, various pedestrian safety facilities continue


to be developed and implemented. Crosswalk markings are a typical
Pedestrians, as a typical group of vulnerable road users, are exposed type of facility, which has those functions mentioned above.
to a relatively unsafe walking environment. According to the Global Crosswalk markings are a type of pedestrian safety facilities which
Status Report on Road Safety 2018, approximately 1.35 million people are installed at the vehicle-pedestrian interaction locations. Installing
die each year due to crashes, and pedestrian deaths constitute 23 % of crosswalk markings is to mark the path that allows pedestrians to cross
road traffic deaths by type of road users (World Health Organization, the road under certain conditions and to warn drivers to watch out for
2018). Nation Bureau of Statistics of China reported that about 1322 pedestrians crossing the street (China National Standardization
pedestrians were killed and more than 1340 were injured in traffic Management Committee, 2019). In order to ascend the function men-
accidents in 2017 (Nation Bureau of Statistics of China, 2019). The road tioned above, highway departments in different countries and regions
intersection is a specific road section where vehicles and pedestrians’ developed various crosswalk marking patterns by redesigning the
paths in different directions meet. This is one of the critical reasons why geometric shape of crosswalk markings or changing colors of crosswalk
many pedestrian crashes occur. In 2017, there were 5977 pedestrians markings. Several crosswalk marking patterns are shown in Fig. 1.
killed, and 18 % of them died at or near intersections in the USA The crosswalk markings in Fig. 1(d) are new-designed crosswalk
(National Highway Traffic Safety Administration, 2017). Over 30 per- markings which were installed at signalized intersections in Fengtai
cent of the pedestrian fatalities happened at intersections between 1999 District and Haidian District, Beijing, in order to enhance the visibility
and 2014 in Canada (Transport Canada, 2020). of the crosswalk (Beijing Traffic Management Bureau, 2019). However,
In order to decrease pedestrian crashes and enhance the safety of according to the Road Traffic Signs and Markings (China National


Corresponding author.
E-mail addresses: bianyang@bjut.edu.cn (Y. Bian), kkkyleung@emails.bjut.edu.cn (K. Liang), zhaoxiaohua@bjut.edu.cn (X. Zhao), lihaijian@bjut.edu.cn (H. Li),
yangliping@emails.bjut.edu.cn (L. Yang).

https://doi.org/10.1016/j.aap.2020.105498
Received 28 October 2019; Received in revised form 9 March 2020; Accepted 9 March 2020
Available online 13 March 2020
0001-4575/ © 2020 Elsevier Ltd. All rights reserved.
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Fig. 1. Several crosswalk marking patterns in different cities.

Standardization Management Committee, 2019), the pattern of the


new-designed crosswalk markings does not comply with the relevant
provisions of China. The effectiveness evaluation of this type of cross-
walk markings was also not conducted.

2. Literature review

2.1. Vehicle-pedestrian interaction occurring time and location conditions

Vehicle-pedestrian interaction at intersections is pervasive and this


may cause a relatively high risk of crashes or collisions. Besides traffic
flow factors, such as the number of pedestrians, vehicle speed and ve-
hicle average daily traffic volume (Zegeer et al., 2002; Himanen and
Kulmala, 1988; Obeid et al., 2017), when and where the risk of vehicle-
pedestrian interactions at intersections is relatively high should also be
taken into consideration.

2.1.1. Locations where high-risk vehicle-pedestrian interactions occur


In the vast majority of cases, interactions between vehicles and Fig. 2. Illustration of primary and secondary interactions at a four-legged in-
pedestrians at intersections happened on the crosswalks. Taking a four- tersection.
leg intersection as an example, this type of intersection usually has four
crosswalks for pedestrians to cross the road. When the vehicle turns left, pedestrians involved in secondary interactions, compared to primary
right or crosses the intersection, it will pass the crosswalk locations interactions. Some previous studies also indicated that secondary in-
twice, as shown in Fig. 2. According to Fu et al. (2019), the vehicle teractions are more dangerous (Roudsari et al., 2006; Iasmin et al.,
passes the first crosswalk located on the intersection approach where it 2016). Thus, when it comes to vehicle-pedestrian interactions at in-
comes from. Then, on its way to exit the intersection, the vehicle passes tersections, the locations where interactions occur should be both dis-
a second crosswalk on the exiting street (the street that the users take to tinguished and considered.
exit the intersection), regardless of its destination (choice of exiting the
street). Therefore, pedestrian-vehicle interactions at intersections can 2.1.2. Time when high-risk vehicle-pedestrian interactions occur
be put in two categories: The risk level of vehicle-pedestrian interactions is also related to
vehicle real-time operation data, such as speed, distance from the
• Primary interactions (PI), defined as those happening at the first conflict location, etc. These dynamics will influence the time when
pedestrian crosswalk on the approach street where vehicles enter vehicles arrive at the conflict location eventually. When taking the
the intersection; moment that the pedestrian arrives at the conflict location too, Time-
• Secondary interactions (SI), defined as those interactions occurring To-Zebra arrive (TTZarr) was proposed. TTZarr (Varhelyi, 1998) is de-
at exiting streets where vehicles leave the intersection. fined as the time left for the vehicle to arrive at the zebra crossing at the
moment the pedestrian arrives at the curb. TTZarr is obtained by di-
In Fu et al. (2019), drivers’ driving behaviors in the second inter- viding the distance of the vehicle from the zebra crossing by the ve-
action are more dangerous, indicating the significantly higher risk for hicle’s speed when the pedestrian arrives at the curb. According to

2
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Várhelyi’s (1998) and Bella and Silvestri’s (2015) research, TTZarr can measures on drivers’ speed behavior at pedestrian crossings.
be divided into three situations, and the driver’s behavior and the pe- In summary, the distinguished feature of this paper is to conduct a
destrian’s behavior are different in these three situations: driving simulator experiment, which can control independent variables
strictly, to evaluate the effectiveness of the new-designed crosswalk
• TTZarr < 1 s, when the pedestrian reaches the curb, the vehicle is markings, considering the occurring time and location conditions of
very close to the conflict point. Thus, the driver will not be able to vehicle-pedestrian interactions. The drivers’ maneuvering behavior
stop, and the pedestrian will not start to cross, allowing the vehicle data and vehicle operating data obtained from the driving simulator to
to continue without being forced to brake; explain how crosswalk markings influence driving behavior and then
• 1 s ≤ TTZarr ≤ 4 s, the pedestrian could reach the conflict point evaluate the effectiveness of crosswalk markings. The conclusions of
before the driver and force him to brake. The driver accelerates, in evaluation serve as the basis for this facility installing and standards.
order to take priority in passing the crosswalk before the pedestrian.
• TTZarr > 4 s, the pedestrian has enough time to pass the conflict 3. Methods
point safely. The driver realizes that he cannot pass before the pe-
destrian and, thus, adopts a lower speed. The study was conducted using the advanced driving simulator of
the Beijing University of Technology in China. Three independent
According to the descriptions of three situations of TTZarr, when 1 s variables were considered in this experiment:
≤ TTZarr ≤ 4 s, the driver and the pedestrian both want to take priority
in passing the crosswalk. During this period, the pedestrian’s behavior • Two different crosswalk markings (Standard Crosswalk Markings in
and the driver’s behavior have high uncertainty. Thus, when TTZarr is in China (SCMC) and New-designed Crosswalk Markings in China
this range, the probability of crash between vehicle and pedestrian is (NCMC)).
relatively high. • Two different vehicle-pedestrian interaction locations (primary in-
teraction and secondary interaction).
2.2. Evaluation of the effectiveness of crosswalk markings • Three different levels of the pedestrian number (0, 1 or 5).

Evaluation of the effectiveness of crosswalk markings is an indis- Therefore, this experiment followed a full factorial design: 2
pensable part of traffic facilities using. In Table 1, researches on the (crosswalk markings) × 2 (vehicle-pedestrian interaction locations) ×
effectiveness of crosswalk markings from the 1970s until recently are 3 (levels of the pedestrian number). The hypotheses of the study were
summarized. The data collection method of these researches was the as follows:
field test. Early studies focus on the effectiveness of marked crosswalk
Hypothesis 1. Drivers would be more likely to experience unhealthy
and unmarked crosswalk (crosswalk installed crosswalk markings or
tension (may cause dizziness, nausea, or anxiety) when they see the
not). The effectiveness of marked crosswalk and unmarked crosswalk
NCMC.
obtained by these previous studies is not consistent, which may be
caused by inconsistent external environmental conditions. Recent stu- Hypothesis 2. In respect of drivers, NCMC would be more conspicuous,
dies shifted the focus to the effectiveness of different types of crosswalk which will lead to a better warning, and it would make driving behavior
markings. The researchers found that different types of crosswalk more predictable and drivers’ speed control awareness stronger.
markings caused different drivers’ behavior, and that drivers had dif-
Hypothesis 3. The effectiveness of different crosswalk markings on
ferent preferences for different types of crosswalk markings.
driving behavior is not absolute. Other external conditions (vehicle-
As is shown in Table 1, evaluation indicators, including crash data,
pedestrian interaction locations or levels of the pedestrian number)
traffic conflict data, and accident field observation data, have been
would also influence the effectiveness of crosswalk markings.
extensively used to check the effectiveness of the safety of this type of
pedestrian facilities. However, these data can only be collected after According to hypotheses above, the objective of this paper is to
crashes or conflicts occurred. Due to this reason, it takes a long time to evaluate the effectiveness of this new-designed crosswalk markings at
collect these data, and it may also cause the uncertainty of the sample signalized intersections when vehicle-pedestrian interactions occur in
size. Additionally, these crashes may cause pedestrians or researchers to different time and location.
be injured or even killed.
Compared to field tests, a driving simulation study has some es- 3.1. Scenarios and vehicle-pedestrian interaction design
sential advantages. For example, it is safer and there is more effective
control on experiments. Besides, a driving simulator can collect real- In light of the practical applications and suggestions from relevant
time performance data, such as data of the throttle, the brake and the experts in Beijing Municipal Institute of City Planning & Design, the
steering wheel, which are more difficult to acquire in a field test (Zhao scenarios used in this driving simulator experiment were designed as
et al., 2018a, 2018b;Zhao et al., 2019). the validity of the Beijing follows.
University of Technology driving simulator was validated (Xu, 2012; Twelve factors combined scenarios were developed based on a sig-
Ding et al., 2013, 2014, 2015; Zhao et al., 2018a, b; Huang et al., 2019) nalized intersection, where two crossing roads are composed of four-
These previous studies found that the driving simulator was relatively lane urban roads of totally about 1600 m long. This road was 24 m wide
effective in terms of speeds and achieving drivers’ physiological para- formed by four 3.50 m wide lanes, two 2 m wide bicycle lanes, two 1 m
meters. wide lane separators and two 2 m wide sidewalks. Two of four lanes
Although few studies conducted a driving simulation study to were left turn lanes, and another two lanes were used for straight and
evaluate the effectiveness of different crosswalk marking patterns, the right-turning traffic. The distance between the signalized intersection
effects of some other pedestrian safety facilities were also explored by and the starting point equals to 800 m, which allowed the drivers to
this method. Fisher and Garay-Vega (2012) studied drivers’ behaviors, reach a congruous speed for the simulated urban scenario. The posted
including the yielding behavior and the percentage of drivers looking to speed limit was 60 km/h.
target zone, on advance yield markings at marked mid-block crosswalks According to the Road Traffic Signs and Markings (China National
in multi-threat scenarios. Salamati et al. (2012) explored the effects of Standardization Management Committee, 2019), the design para-
three different crosswalk treatments at the exit leg of multilane meters, including size, installed location and RGB color, of Standard
roundabouts according to drivers’ yielding rate and eye tracker results. Crosswalk Markings in China (SCMC) were confirmed. The design
Bella and Silvestri (2015) studied the effects of three different safety parameters of NCMC are determined based on a field investigation in a

3
Y. Bian, et al.

Table 1
Researches on the Effectiveness of Crosswalk Markings.
Year Researchers Object Location Measure of effectiveness Results

1972 Herms (1972) Marked crosswalk vs. unmarked Unsignalized intersections Crash rates Unmarked crosswalks were better
crosswalk
1994 Gibby et al. (1994) Marked crosswalk vs. unmarked Unsignalized highway Crash rates Unmarked crosswalks were better
crosswalk intersections
2000 Jones and Tomcheck (2000) Marked crosswalk vs. unmarked Unsignalized arterial Crash rates Unmarked crosswalks were better
crosswalk intersections
2000 Knoblauch and Raymond (2000) Marked crosswalk vs. unmarked Unsignalized locations Vehicle speed Drivers respond to crosswalk markings more or less.
crosswalk
2001 Knoblauch et al. (2001) Marked crosswalk vs. unmarked Uncontrolled intersection Yielding behavior No difference between marked crosswalks and unmarked
crosswalk crosswalks.
2001 Nitzburg and Knoblauch (2001) High-visibility crosswalk markings Unsignalized intersections Yielding rate Drivers were more likely to yield at the high-visibility
crosswalk
2002 Zegeer et al. (2002) 1.Marked crosswalk vs. unmarked Uncontrolled intersection Crash rates 1.No statistically significant difference between marked

4
crosswalk crosswalks and unmarked crosswalks
2.Different crosswalk marking pattern 2. No statistically significant difference between different
crosswalk marking pattern
2005 Chicago Department of yellow/green crosswalk markings Elementary school zone Drivers’ speed behavior Had an insignificant effect
Transportation (2019)
2011 Fitzpatrick et al. (2011) Transverse Lines, Continental Experiment locations Drivers’ preference Continental markings and bar pairs were better than transverse
Markings, and Bar Pairs markings.
2011 Iragavarapu et al. (2011) Transverse Lines, Continental Experiment locations Detection distances Bar pairs and continental markings were similar, and they were
Markings, and Bar Pairs better than the transverse markings
2012 Pulugartha et al. (2012) High-visibility crosswalk 1.pedestrians trapped in the street Significant increase in the distance at which drivers yielded to
2.pedestrians looking for vehicles before beginning pedestrians.
to cross
3.pedestrians looking for vehicles before crossing
the second half of the street
4.percent of captured or diverted pedestrians
5.driver yield behavior and distance
6.drivers blocking the crosswalk
Accident Analysis and Prevention 139 (2020) 105498
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Fig. 3. Scenarios and vehicle-pedestrian interaction design.

signalized intersection in Fengtai District. Detailed design parameters of NCMC and SCMC in different levels of pedestrian presence and to ex-
SCMC and NCMC are shown in Fig. 3(a) and (b). plore whether the three evaluation results are different. Considering the
In Beijing, China, most traffic signals at intersections have protected difference between the signal phase when primary interaction occurs
left-turn phases. Pedestrians in all directions are prohibited to cross the and that when secondary interaction occurs, this study set the primary
road when this phase is green. Nevertheless, right-turning traffic signals interaction and the secondary interaction as two of 12 scenarios when
are designed to be either combined with the phase of straight traffic the pedestrian was absent.
movement or kept free of traffic signals. The right-turning vehicles in Concerning the time when vehicle-pedestrian interaction occurs, 1 s
Beijing may have both primary interactions and secondary interactions ≤ TTZarr ≤ 4 s is the most dangerous time left for the vehicle to arrive
with the crossing pedestrians. Thus, this study chose right turn as a at the zebra crossing at the moment the pedestrian arrives at the curb,
typical situation in this driving simulator and two types of vehicle-pe- according to Bella and Silvestri’s research. Besides, they concluded that
destrian interactions at intersections, shown in Fig. 3(c) and (d), were 1 s ≤ TTZarr ≤ 4 s, TTZarr ≤ 1 s and TTZarr ≥ 4 s are three different
both taken into consideration. The color of traffic lights in each sig- time ranges of vehicle-pedestrian interaction and set 1 s, 2.5 s and 4 s as
nalized intersection did not change, in order to avoid drivers seeing the theoretical values of TTZarr. 2.5 s is the median between 1 s and 4 s.
different signals because of different speeds. Accordingly, the signal While, 1 s is the boundary value of the interval [1 s, 4 s] and the in-
phase settings, which can ensure vehicle-pedestrian interactions occur terval [0 s,1 s]. Similarly, 4 s is the boundary value of the interval [1 s,
at primary location or secondary location, are also shown in Fig. 3(c) 4 s] and the interval [4 s,+∞]. Thus, the specific values of TTZarr were
and (d). set as 2.5 s. Therefore, the pedestrian was triggered when the driver was
Another variable that was also controlled in the experiment is the 34.7 m. In addition, the pedestrian speed was set as 1.2 m/s, according
number of crossing pedestrians. Obeid et al. (2017) analyzed driver- to the literature (Shan, 2019).
pedestrian interaction in a mixed-street environment using a driving During the experimental process, each scenario was characterized
simulator. In their study, number of pedestrians crossing (0, 1 and 3) by low traffic volume. In this way, the experimental vehicle would not
was set as an independent variable, and the results show that drivers encounter interference while in operation.
are more cautious and their driving is less aggressive when more pe-
destrians are crossing. Therefore, three levels of pedestrian presence (0, 3.2. Apparatus
1 and 5) were adopted in this study. The aim of selecting three levels of
pedestrian presence in this study was to evaluate the effectiveness of A fixed-base driving simulator at Beijing University of Technology

5
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Table 2
Summary of Dependent Variables.
Variables Description

Average speed Average speed in the key impact area. It reflects the overall effectiveness of the crosswalk markings. Lower average
speed in the key influence range indicated the better effectiveness of crosswalk markings.
The SD of acceleration The average value of standard deviation (SD) of acceleration in the key impact area. It is used to evaluate speed as
an indirect evaluation indicator, which depicts the stability of speed reduction in the key impact area. Lower SD of
acceleration in the key influence range indicated the better stability of speed reduction influenced by crosswalk
markings.
Gas pedal power PGPP = ∫T f(t)dt . Where f(t) was the function of the pressing intensity of the gas pedal over time, ranging from 0 to 1;
t was the travel time, in s; and PGPP was the gas pedal power. The gas pedal power not only measured the intensity or
force of drivers’ pressure on the throttle pedal but also considered the duration and frequency of drivers’ use of the
throttle pedals. More gas pedal power in the key influence range indicated that drivers were aroused vigilantly and
control throttles consciously.
Relative speed contrast (θ ) v 2 − v1
θ= . Where v1 –the entering speed, denoting the vehicle speed upon the entering point of each section, in km/
v1
h; v 2 –the minimum speed, denoting the minimum speed in the key impact area, in km/h; and θ–the relative speed
contrast. The relative speed contrast was used to evaluate the decelerating effects of crosswalk markings. If this
indicator was negative, it meant that the vehicle decelerated while traveling in the section, and vice versa. Higher
relative speed contrast in the key influence range indicated the better effectiveness of crosswalk markings.
Minimum speed (Vmin) The minimum speed value in the key impact area. Lower Vmin in the key influence range indicated the better
effectiveness of crosswalk markings.
Distance from zebra crossing where the speed reduced to Distance from zebra crossing where the speed reduced to the minimum. Higher Dvmin in the key influence range
the minimum (Dvmin) indicated the better effectiveness of crosswalk markings.

was used in this experiment. The road scenario was projected onto three “Q1: What do you think of the conspicuity of Standard Crosswalk
large screens to provide a 130-degree field of view. The vehicle oper- Markings in China?” (from very inconspicuous to very conspicuous), “Q2:
ating data (e.g., braking force, acceleration, speed, lateral placement, What do you think of the conspicuity of New-designed Crosswalk
lane number, and turning angle of the steering wheel) were recorded 30 Markings in China?” (from very inconspicuous to very conspicuous), “Q3:
times per second. Which level of your vigilance to risk (be aware that a pedestrian may be
crossing the crosswalk, and a collision may occur if they do not slow
3.3. Participant down) ahead do you think Standard Crosswalk Markings in China can
arouse?” (from very low to very high), “Q4: Which level of your vigilance
According to Central Limit Theorem, if a sum of random variables is to risk ahead do you think New-designed Crosswalk Markings in China
normally distributed, a large sample size obtained from those variables can arouse?” (from very low to very high), “Q5: Which level of your
also fits normal distribution. Besides, the sample size of no less than 20 unhealthy tension (may cause dizziness, nausea, or anxiety) do you
is acceptable (Ben-Bassat and Shinar, 2011; Van Der Horst and De think Standard Crosswalk Markings in China can arouse?” (from very
Ridder, 2007; Horberry et al., 2006). Accordingly, a total of 27 healthy high to very low), “Q6: Which level of your unhealthy tension do you
participants were recruited by advertisement, including 9 females and think New-designed Crosswalk Markings in China can arouse?” (from
18 males. The average age was 35.6 years old (SD = 9.1, range = very high to very low). Finally, another questionnaire was also required
23–49 years old). All participants were provided with written informed to be finished by the participant to report his or her subjective eva-
consents before joining the experiment. Drivers could not participate in luation of the driving simulator, together with post-experiment phy-
the training until informed consents are provided. siological and psychological conditions.

3.4. Procedures
3.5. Dependent variables
Participants were required to fill out a questionnaire before the
experiment, which collected their basic information (age, gender, The two different crosswalk markings with six different vehicle-
driving experience, etc.) as well as their physiological and psycholo- pedestrian interactions were defined as the independent variables.
gical conditions. Then participants were required to perform a test Researchers chose six driving behavior indicators, which were closely
driving for 5−10 min on a specific scenario in order to become familiar related to the speed, as the dependent variables: average speed, the
with the driving simulator. standard deviation (SD) of acceleration, gas pedal power, relative speed
After the test driving, participants started the formal driving ex- contrast (θ), minimum speed (Vmin) and distance from zebra crossing
periment, in which the vehicle maneuvering and drivers’ operational where the speed reduced to the minimum (Dvmin) (Obeid et al., 2017;
performance data were collected. Each participant needs to complete Bella and Silvestri, 2015; Zhao et al., 2016, 2018b). According to the
12 scenarios in an entirely random order. In each scenario, participants maximum display distance of crosswalk markings in the driving simu-
were required to drive from the starting point and turn right to arrive at lator, this paper regarded the section from the 100 m prior to crosswalk
the ending point. The entire driving experiment lasted approximately markings to 20 m post to the stop line to crosswalk markings as the key
20 min on average. When the formal experiment was finished, each impact area. The six dependent variables were listed in Table 2
participant left the driving simulator and filled out a scene authenticity
evaluation questionnaire and a questionnaire about the perceived ef-
fectiveness of the two different crosswalk markings. The scene au- 4. Results
thenticity evaluation questionnaire consisted of scale that rated the
authenticity of various elements in the scene, including trees, dividers, From the driving simulator experiment, driving behavior indicators,
street lights, pedestrians, signs and markings, road surface, and traffic including vehicle operating data, drivers’ maneuvering data, and dri-
lights. The answers range from 1 to 10, representing the worst to the vers’ subjective evaluation results, were collected. Then the effective-
best. Another questionnaire consisted of six items and each item was ness of NCMC was analyzed. The evaluation of effectiveness was di-
rated by participants on a 5-point unipolar scale ranging from 1 to 5: vided into the following three aspects.

6
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Table 3 4.2.2. The standard deviation of acceleration


Results of the Scene Authenticity Evaluation. When there was one pedestrian present, the standard deviation of
Element Trees Dividers Street Pedestrians Signs and Road Traffic acceleration was significantly influenced by different crosswalk mark-
lights markings surface lights ings (Ped.1/PI - p = 0.097; Ped.1/SI - p = 0.009). The standard de-
viation of acceleration of NCMC is higher than that of SCMC. When
Result 9.2 9.3 9.3 8.7 9.2 9.5 9.5
there were five pedestrians present or the pedestrian was absent, the
effect of two types of crosswalk markings on the standard deviation of
acceleration was not statistically significant. (Ped.0/PI - p = 0.772;
4.1. Results of scene authenticity evaluation
Ped.0/SI - p = 0.235; Ped.5/PI - p = 0.489; Ped.5/SI - p = 0.888)
The results of the scene authenticity evaluation questionnaire were
showed in the following table: 4.2.3. Gas pedal power
As is shown in Table 3, the evaluation scores of the scene authen- When there was one pedestrian present, gas pedal power was sig-
ticity evaluation are all above 8.5. It means that the experiment scene in nificantly influenced by different crosswalk markings. The gas pedal
this experiment is close to reality and vehicle maneuvering and the power of NCMC was higher than that of SCMC at PI situation (p =
drivers’ operational performance data are valid. 0.028), whereas at SI situation, it is the opposite (p = 0.079). When
there were five pedestrians present, the gas pedal power of NCMC was
significantly higher than that of SCMC at PI situation (p = 0.011).
4.2. Influences on driving behaviors Nevertheless, the gas pedal power was not significantly influenced by
different crosswalk markings at SI situation (p = 0.755). When the
A repeated-measure analysis of variances (rANOVA) of six driving pedestrian was absent, the effect of different crosswalk markings on gas
behavior indicators was performed. The influence of two different pedal power was insignificant.
crosswalk markings was analyzed under different vehicle-pedestrian
interaction locations (PI, SI) and different levels of the pedestrian
4.2.4. Relative speed contrast
number (0, 1, 5) and these results represented the crosswalk markings’
When there were five pedestrians present, relative speed contrast
relative effectiveness under different conditions, rather than the abso-
was significantly influenced by different crosswalk markings, and the
lute effectiveness. Results were shown in Table 4 and Fig. 4.
relative speed contrast of NCMC was higher than that of SCMC at PI
situation (p = 0.025) and SI situation (p = 0.085). When there was one
4.2.1. Average speed pedestrian present or the pedestrian was absent, the effect of two types
When the pedestrian was absent, the effect of two types of crosswalk of crosswalk markings on Relative speed contrast was not statistically
markings on average speed was not statistically significant both at PI significant (Ped.0/PI - p = 0.444; Ped.0/SI - p = 0.317; Ped.1/PI - p =
situation (p = 0.725) and SI situation (p = 0.844). The results in- 0.374; Ped.1/SI - p = 0.174).
dicated that the average speed of NCMC was significantly lower than
that of SCMC when pedestrians were present (Ped.1/PI - p = 0.086; 4.2.5. Minimum speed
Ped.1/SI - p = 0.011; Ped.5/PI - p = 0.022; Ped.5/SI - p = 0.076). When there were five pedestrians present, minimum speed was

Table 4
rANOVA Results of Six Driving Behavior Indicators.
Average speed The SD of acceleration Gas pedal power Relative speed contrast Vmin Dvmin

Ped.0 PI SCMC Mean 37.847 0.309 0.297 −0.410 28.532 6.733


SD 6.411 0.146 0.078 0.194 9.149 4.721
NCMC Mean 38.358 0.320 0.268 −0.378 30.364 4.189
SD 7.519 0.163 0.080 0.202 10.691 6.622
p-value   0.725 0.772 0.100 0.444 0.377 0.081*
Ped.0 SI SCMC Mean 39.859 0.312 0.278 −0.387 31.089 8.296
SD 7.726 0.161 0.095 0.197 10.974 9.403
NCMC Mean 39.575 0.269 0.283 −0.358 31.219 4.372
SD 6.844 0.145 0.070 0.226 10.750 4.286
p-value   0.844 0.235 0.800 0.317 0.933 0.060*
Ped.1 PI SCMC Mean 34.761 0.444 0.365 −0.659 17.249 8.825
SD 7.781 0.176 0.137 0.286 14.555 6.248
NCMC Mean 31.610 0.514 0.419 −0.697 14.568 16.425
SD 7.203 0.171 0.129 0.247 11.927 14.379
p-value   0.086* 0.097* 0.028** 0.374 0.262 0.017**
Ped.1 SI SCMC Mean 38.057 0.396 0.375 −0.655 18.243 3.690
SD 7.203 0.132 0.142 0.208 9.454 7.647
NCMC Mean 33.581 0.499 0.325 −0.593 17.383 4.759
SD 4.314 0.161 0.080 0.224 11.166 7.581
p-value   0.011** 0.009** 0.079** 0.174 0.745 0.611
Ped.5 PI SCMC Mean 32.276 0.462 0.421 −0.694 14.092 10.304
SD 8.460 0.158 0.123 0.328 16.054 11.224
NCMC Mean 27.977 0.495 0.495 −0.843 6.899 16.237
SD 8.060 0.183 0.104 0.122 6.706 7.016
p-value   0.022** 0.489 0.011** 0.025** 0.026** 0.022**
Ped.5 SI SCMC Mean 35.657 0.466 0.332 −0.593 21.166 2.406
SD 8.207 0.147 0.108 0.213 11.787 3.946
NCMC Mean 32.898 0.470 0.339 −0.656 17.868 5.105
SD 7.776 0.157 0.093 0.273 13.260 6.858
p-value   0.076* 0.888 0.755 0.085* 0.072* 0.095*

* p < 0.1, **p < 0.05.

7
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Fig. 4. Six driving behavior indicators.

significantly influenced by different crosswalk markings, and the 4.3. Outcomes of the subjective questionnaires
minimum speed of NCMC was lower than that of SCMC at PI situation
(p = 0.026) and SI situation (p = 0.072). When there was one pe- The results of the questionnaire on the perceived effectiveness of the
destrian present or the pedestrian was absent, the effect of two cross- two types of crosswalk markings were shown in Fig. 5. In Fig. 5(a), the
walk markings on relative speed contrast was not statistically sig- conspicuity scores of SCMC and NCMC obtained by participants’ sub-
nificant (Ped.0/PI - p = 0.377; Ped.0/SI - p = 0.933; Ped.1/PI - p = jective evaluation are 3.296 and 3.963. It indicated that drivers thought
0.262; Ped.1/SI - p = 0.745). that the conspicuity of NCMC was better than that of SCMC. The results
of the level of drivers’ vigilance to risk ahead aroused by SCMC and
NCMC, shown in Fig. 5(b), are 3.222 and 3.481. Accordingly, this il-
4.2.6. Distance from zebra crossing where the speed reduced to the lustrates that the warning effect of NCMC is better than that of SCMC.
minimum With respect to the unhealthy tension caused by two different crosswalk
When there were five pedestrians present or the pedestrian was markings, the score of SCMC is higher than that of NCMC. It means that
absent, Dvmin was significantly influenced by different crosswalk NCMC is more likely to cause drivers’ unhealthy nervousness, which is
markings. The Dvmin of NCMC was significantly higher than that of consistent with Hypothesis 1.
SCMC at PI situation (p = 0.022) and SI situation (p = 0.095).
Nevertheless, the Dvmin of NCMC was lower than that of SCMC at PI
situation (p = 0.081) and SI situation (p = 0.060). When there was one 4.4. Comprehensive evaluation of different crosswalk markings in different
pedestrian present, the Dvmin of NCMC was significantly higher than vehicle-pedestrian interactions
that of SCMC at PI situation (p = 0.017), while it was not significant at
SI situation (p = 0.611). In fact, the influences of two types of crosswalk markings on in-
dicators were different in the six conditions of vehicle-pedestrian

8
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Fig. 5. Outcomes of the subjective questionnaires.

interactions. Thus, the Technique for Order of Preference by Similarity 1 − Hj


Wj = n ,
to Ideal Solution (TOPSIS) was used in this research. TOPSIS was a ∑ j = 1 (1 − Hj ) (3)
method for multi-criteria decision analysis, which could evaluate a few
schemes with many indicators by calculating the premium degree of Index weights (Wj) were acquired on account of the former matrix
each scheme (Qin et al., 2018). In order to determine weights of nine X *ij. The values of these weights demonstrated that nine indicators had
indicators used in TOPSIS, this paper adopted the entropy weight different effects on the comprehensive efficiency of the two different
method, which was the most objective method of weight assignment. crosswalk markings. Then, the weighted normalized matrix was ob-
According to the literature (Li et al., 2014), information entropy was tained according to the Eq. (4):
introduced into information theory in 1948 by C. E. Shannon based on Uij = W*X *
j ij, (4)
entropy (a thermodynamic concept). Shannon information entropy,
which is an objective and applicable method for the determination of The maximum and minimum indicators from matrix Uij were used to
weight value, was introduced into the comprehensive assessment. form two ideal solution vectors, U+ and U-. For the evaluation units of
Therefore, TOPSIS with the combination of coefficient of entropy two crosswalk markings, their distance to the U+ formed the best dis-
weight was used to get premium degrees of the two types of crosswalk tance vector Di+; their distance to the U- formed the worst distance
markings in the six situations of vehicle-pedestrian interactions. vector Di-. Finally, the premium degrees (Ci*) of the two types of
crosswalk markings were obtained according to the Eq. (5):
4.4.1. TOPSIS algorithm Di+
The nine indicators were used as the evaluation indices for the two Ci* = ,
Di+ + Di− (5)
crosswalk markings in the six situations of vehicle-pedestrian interac-
tions. Thus, the multiple objective decision matrix of TOPSIS was The value of premium degrees was 0-1. A premium degree closer to
formed as X = (X)mn (m = 2, n = 9) in the six situations of vehicle- 1 was associated with better efficiency of the scheme.
pedestrian interactions.
TOPSIS method required that all indicators should work in the same 4.4.2. Results of TOPSIS in different crosswalk markings in different
direction so that it could help explain the final result. Therefore, vehicle-pedestrian interaction situations
transformation consistency was performed on the indicators. A new The weights of nine indicators in the six situations of vehicle-pe-
multiple objective decision matrix (X #IJ ) was formed. Then, the nor- destrian interactions were obtained, shown in Table 5.
malized matrix X *ij was obtained according to the Eq. (1) The premium degrees of two crosswalk markings in different ve-
hicle-pedestrian interactions were also calculated, as shown in Fig. 6.
X IJ#
X ij* = n
, When there were five pedestrians present, the premium degrees of
∑i = 1 (Xij# )2 (1) NCMC are higher than those of SCMC, no matter where the vehicle-
pedestrian interaction occurred. When there was one pedestrian present
The entropy weight method was used. The information entropy Hj
in scenarios, which was basically consistent with five pedestrians’ si-
was obtained according to the Eq. (2)
tuations, the premium degrees of NCMC are higher than those of SCMC.
m
While the difference of the premium degrees between SCMC and NCMC
Hj = −k∑ pij lnpij,
i= 1 (2) when there was one pedestrian is relatively lower than that when there
xij 1
were five pedestrians. When the pedestrian was absent, the premium
Where pij = ∑im= 1 xij
,k= lnm degrees of NCMC are lower than those of SCMC. The results mentioned
The entropy weight was obtained according to the Eq. (3) above is consistent with Hypothesis 3.

9
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Table 5
The Weights of Nine Indicators in the Six Situations of Vehicle-pedestrian Interactions.
Ped.0 / PI Ped.0 / SI Ped.1 / PI Ped.1 / SI Ped.5 / PI Ped.5 / SI

Speed 0.00057 0.00010 0.01715 0.06444 0.02407 0.00995


SD of acceleration 0.00388 0.04403 0.04041 0.21881 0.00542 0.00016
Gas pedal power 0.03287 0.00051 0.03576 0.08303 0.03098 0.00062
θ 0.02036 0.01234 0.00600 0.04146 0.04439 0.01532
Vmin 0.01215 0.00004 0.05409 0.00960 0.56605 0.04394
Dvmin 0.68817 0.78740 0.70002 0.26486 0.23808 0.81159
Conspicuity 0.10609 0.06820 0.06425 0.13931 0.03989 0.05191
Vigilance 0.01879 0.01208 0.01138 0.02468 0.00707 0.00920
Unhealthy tension 0.11713 0.07530 0.07094 0.15381 0.04405 0.05731

crosswalk markings ahead, and then they were alerted by crosswalk


markings and turned their attention to pedestrians on the street. When
drivers did not find pedestrians, who were crossing or tended to cross
the road, their behavior may not be influenced by crosswalk markings,
and they may maintain a speed, which they thought was relatively safe
and efficient to drive across intersections. Similar speed behavior was
also found in Bella and Silvestri’s study (2015). In addition, as is shown
in Fig. 5 (c), NCMC was more likely to arouse drivers’ unhealthy ten-
sion, which supports Hypothesis 1. This phenomenon may be raised by
the striking color of NCMC and an optical illusion of stereoscopic vision
caused by the combination of blocks. In order to eliminate the dis-
comfort as soon as possible, drivers may tend to drive across the in-
Fig. 6. Premium degrees of two crosswalk markings in different vehicle-pe-
tersection faster. Therefore, the situation where approaching drivers
destrian interactions. cannot see pedestrians at intersections should be taken into con-
sideration when NCMC is installed in the practical application. For
example, drivers' views may be obscured at intersections.
5. Discussion
In terms of the situation where there were five pedestrians present,
the premium degrees of NCMC were better than those of SCMC when
In the current study, a driving simulator experiment was conducted
vehicle-pedestrian interaction occurred at two interaction locations. It
to evaluate the effectiveness of this new-designed crosswalk markings
indicated that NCMC has a relatively better influence on the speed
at signalized intersections, considering when and where vehicle-pe-
behaviors of drivers at intersections when the pedestrian flow was high,
destrian interaction occurs. Based on the results of driving behaviors,
compared with SCMC. The rANOVA results of driving behaviors in-
subjective questionnaires, and comprehensive evaluation, there were
dicated that average speed, relative speed contrast, gas pedal power,
some implications about the effectiveness of NCMC.
minimum speed and Dvmin of NCMC were significantly better than those
From the results of subjective questionnaires, NCMC are more
of SCMC. When pedestrian flow was high, the NCMC affected drivers to
conspicuous, and it can arouse drivers’ vigilance to risk better. As for
decelerate to a lower minimum speed and decelerate to minimum speed
driving behavior, Hypothesis 2 is proved in most cases. Drivers’ pre-
earlier, due to the striking color of NCMC and an optical illusion of
dictability of potential threats (crossing pedestrians), which was re-
stereoscopic vision caused by the combination of blocks. Although
presented by Dvmin, was better when NCMC were installed. Besides,
NCMC’s characteristics aroused drivers’ unhealthy tension, they had a
drivers’ speed control awareness under NCMC (represented by average
much more positive effect on drivers’ behaviors of approaching inter-
speed, the standard deviation of acceleration, gas pedal power and
sections. The mean speed profiles in Fig. 8 also depict the vehicle op-
minimum speed) was relatively better, compared with that under
eration process influenced by NCMC and SCMC. The mean speed of
SCMC.
NCMC at 100 m from crosswalk was almost the same as that of SCMC.
The results of TOPSIS supports Hypothesis 3 that other external
The mean speed difference between NCMC and SCMC gradually in-
conditions (vehicle-pedestrian interaction locations or levels of the
creased until it reached the minimum value of both. Besides, drivers’
pedestrian number) would also influence the effectiveness of crosswalk
subjective evaluation of NCMC’s conspicuity and level of drivers’ vigi-
markings. Interpretation of TOPSIS results under different other ex-
lance to risk ahead aroused by NCMC were also better compared with
ternal condition is shown below. Two conditions when pedestrian was
SCMC.
absent were illustrated in Fig. 7. As is shown in Figs. 3 and 7, scenario
In the case of one pedestrian present, the premium degrees of NCMC
elements in Condition A shown in Fig. 3(c) are same with those in
were also better than those of SCMC. The rANOVA results of driving
Condition PI shown in Fig. 7(a), except the pedestrian numbers. Also,
behaviors indicated that average speed of NCMC was better than that of
scenario elements shown in Fig. 3(d) in Condition B are same with those
SCMC, while the standard deviation of acceleration achieved the op-
show in Fig. 7(b) in Condition SI, except the pedestrian numbers.
posite. The difference between NCMC’s premium degrees and SCMC’s
Therefore, two conditions when pedestrian was absent were also named
premium degrees at primary interaction conditions is large. Whereas,
as PI and SI in this study
the difference between NCMC’s premium degrees and SCMC’s premium
According to the results shown in Fig. 6, the premium degrees
degrees at secondary interaction conditions is relatively small. This may
showed that the comprehensive effectiveness of NCMC was worse than
be due to the fact that in the case of PI, the direction of pedestrians
that of SCMC in the case of Ped.0/PI and Ped.0/SI. The rANOVA results
crossing is approximately perpendicular to the vehicle’s direction.
of driving behaviors indicated that most driving behaviors of NCMC
While, the direction of pedestrians crossing is along the vehicle’s di-
were not significantly different from those of SCMC. While Dvmin of
rection in the case of SI. This may make it harder for drivers to see
NCMC in these two situations was significantly lower than that of
crossing pedestrians in the case of SI. Therefore, whether or not to in-
SCMC. This result can be interpreted as follows. If there was no pe-
stall NCMC at intersections with the low pedestrian flow should be
destrian walking on the crosswalk, drivers might first see different

10
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

Fig. 7. Two conditions when pedestrian was absent.

pedestrian interaction occurring time and location conditions. The


above hypothesis was verified, and this study included the following
findings:

• Compared with SCMC, NCMC have significant impacts on most of


the six indicators more or less. For intersections with the high pe-
destrian flow, the effect of NCMC on average speed, relative speed
contrast, Vmin and Dvmin was significantly better than that of
SCMC. For intersections with the low pedestrian flow, the effect of
NCMC on average speed was significantly better than that of SCMC.
While the effect of NCMC on gas pedal power was significantly
worse than that of SCMC. For intersections with no pedestrians, the
effect of NCMC on Dvmin was significantly worse than that of
SCMC.
• For intersections with the high or low pedestrian flow, the com-
prehensive effectiveness and influences on driving behavior of
NCMC were better than those of SCMC, no matter where vehicle-
pedestrian interactions occurred. The comprehensive effectiveness
of NCMC was slightly better than that of SCMC at Ped.1/SI condi-
tion.

Fig. 8. Mean speed profile when there were five pedestrians present.
• For intersections with no pedestrians, the comprehensive effective-
ness and influences on driving behavior of NCMC were worse than
those of SCMC, no matter where vehicle-pedestrian interactions
considered. occurred.
In brief, the changes in driving behavior and driving psychology
brought by NCMC may allow drivers to better judge the behavior of The main contributions of this study included two aspects: First, the
pedestrians crossing the road. Meanwhile, this driving behavior may comprehensive effectiveness of NCMC was evaluated, which may pro-
give pedestrians a signal that the vehicle will yield to them, thereby vide a reference for the future development of standards. Second, the
enhancing their sense of safety across the road. comprehensive evaluation of driver dynamic operation data combined
In future studies, there is still much work to do. First, this research with subjective perception data is applied, which may lay a foundation
only examined the effects of NCMC, considering different vehicle-pe- for evaluating the effectiveness of pedestrian safety facilities from dif-
destrian interactions. In future researches, external environmental ferent subjects in the future.
conditions, including the number of lanes, vehicle volume, high or low
visibility, day or night time light conditions etc., should also be taken
Author contributions
into consideration. Second, pedestrians’ reactions to oncoming vehicles
are more complex. The future studies on the effectiveness evaluation of
The authors confirm contribution to the paper as follows: study
pedestrian safety facilities should improve the interaction between
conception and design: Yang Bian, Kun Liang; data collection: Kun
virtual pedestrians and vehicles, in order to achieve a more realistic
Liang, Liping Yang; analysis and interpretation of results: Yang Bian,
evaluation result. In addition, crosswalk markings affect not only dri-
Kun Liang, Xiaohua Zhao, Haijian Li, Liping Yang; draft manuscript
vers but also pedestrians. Future studies may consider the combined
preparation: Yang Bian, Kun Liang, Xiaohua Zhao, Haijian Li. All au-
effect of these two different groups.
thors reviewed the results and approved the final version of the
manuscript.
6. Conclusion
Declaration of Competing Interest
This study conducted a driving simulator experiment to examine the
effectiveness of NCMC at signalized intersections in different vehicle- The authors declare that they have no known competing financial

11
Y. Bian, et al. Accident Analysis and Prevention 139 (2020) 105498

interests or personal relationships that could have appeared to influ- Iragavarapu, V., Fitzpatrick, K., Chrysler, S.T., 2011. Driver preference for crosswalk
ence the work reported in this paper. marking patterns. Transp. Res. Rec. 2250 (1), 57–64.
Jones, T.L., Tomcheck, P., 2000. Pedestrian accidents in marked and unmarked cross-
walks: a quantitative study. ITE J. 70 (9), 42–46.
Acknowledgments Knoblauch, R.L., Raymond, P.D., 2000. The Effect of Crosswalk Markings on Vehicle
Speeds in Maryland, Virginia, and Arizona. Publication FHWA-RD-00-101. Center for
Applied Research, Inc., Great Falls, VA.
This study was supported by International Science & Technology Knoblauch, R.L., Nitzburg, M., Seifert, R.F., 2001. Pedestrian Crosswalk Case Studies:
Cooperation Program of China (NO.2017YFE0134500). Richmond, Virginia; Buffalo, New York; Stillwater, Minnesota. Publication FHWA-
RD-00-103, FHWA, U.S. Department of Transportation.
Li, Y., Zhao, L., Suo, J., 2014. Comprehensive assessment on sustainable development of
References highway transportation capacity based on entropy weight and TOPSIS. Sustainability
6 (7), 4685–4693.
Beijing Traffic Management Bureau, 2019. Three-dimensional Crosswalk Appeared Nation Bureau of Statistics of China, 2019. China Statistical Yearbook-2018. Accessed
Beijing Street. White, Blue and Yellow 3 Color Markings Were Conspicuous. jtgl. Feb. 9, 2019. www.stats.gov.cn/tjsj/ndsj/2018/indexch.htm.
beijing.gov.cn/jgj/95332/527905/index.html. National Highway Traffic Safety Administration Pedestrians, 2017. Data (Traffic Safety
Bella, F., Silvestri, M., 2015. Effects of safety measures on driver’s speed behavior at Facts). Accessed May 23, 2019. crashstats.nhtsa.dot.gov/Api/Public/
pedestrian crossings. Accid. Anal. Prev. 83, 111–124. ViewPublication/812681.
Ben-Bassat, T., Shinar, D., 2011. Effect of shoulder width, guardrail and roadway geo- Nitzburg, M., Knoblauch, R., 2001. An Evaluation of High-Visibility Crosswalk
metry on driver perception and behavior. Accid. Anal. Prev. 43 (6), 2142–2152. Treatments—Clearwater. Publication FHWA-RD-00-105, FHWA, U.S. Department of
Chicago Department of Transportation, 2019. Evaluation of School Traffic Safety Program Transportation, Florida.
Traffic Control Measure Effectiveness. Accessed May 15, 2019. mutcd.fhwa.dot.gov/ Obeid, H., Abkarian, H., Abou-Zeid, M., Kaysi, I., 2017. Analyzing driver-pedestrian in-
resources/policy/ygcrosswalkmarking/chicagostudy/index.htm. teraction in a mixed-street environment using a driving simulator. Accid. Anal. Prev.
China National Standardization Management Committee, 2019. National Standards of the 108, 56–65.
People’s Republic of China: Road Traffic Signs and Markings-Part 3: Road Traffic Pulugartha, S.S., Vasudevan, V., Nambisan, S.S., Dangeti, M.R., 2012. Evaluating the
Markings. Accessed April 15, 2019. www.gb688.cn/bzgk/gb/newGbInfo?hcno= effectiveness on infrastructure-based countermeasures on pedestrian safety. In:
DC7BD0612CD48B8B3C81DD3EFCEA18D7. Presented at the 91st Annual Meeting of the Transportation Research Board.
Ding, H., Zhao, X., Rong, J., Ma, J., 2013. Experimental research on the eff ;ectiveness of Washington, D.C..
speed reduction markings based on driving simulation: a case study. Accid. Anal. Qin, Z., Chen, G., Li, T., Sun, W., Fu, B., 2018. CW-TOPSIS mine internal caused fire
Prev. 60, 211–218. evaluation model of “AHP + entropy weight method”. J. Xi’an Univ. Sci. Technol. 38
Ding, H., Zhao, X., Rong, J., Ma, J., 2014. Experimental Research on the Eff ;ectiveness of (2), 193–201.
Speed Reduction Markings Based on Driver’s Operating Performance: A Driving Roudsari, B., Kaufman, R., Koepsell, T., 2006. Turning at intersections and pedestrian
Simulation Study. Research Report in Transportation Research Center in Beijing injuries. Traffic Inj. Prev. 7 (3), 283–289.
University of Technology. Salamati, K., Schroeder, B., Rouphail, N.M., Cunningham, C., Zhang, Y., Kaber, D., 2012.
Ding, H., Zhao, X., Rong, J., Ma, J., 2015. Experimental research on the eff ;ectiveness and Simulator study of driver responses to pedestrian treatments at multilane round-
adaptability of speed reduction markings in downgrade sections on urban roads: a abouts. Transp. Res. Rec. 2312 (1), 67–75.
driving simulation study. Accid. Anal. Prev. 75, 119–127. Shan, D., 2019. Study on Configuration and Type-selecting of Pedestrian Crossing
Fisher, D., Garay-Vega, L., 2012. Advance yield markings and drivers’ performance in Facilities. www.doc88.com/p-131713600860.html.
response to multiple-threat scenarios at mid-block crosswalks. Accid. Anal. Prev. 44 Transport Canada, 2020. National Collision Database (NCDB). open.canada.ca/data/en/
(1), 35–41. dataset/1eb9eba7-71d1-4b30-9fb1-30cbdab7e63a. Accessed May 23, 2019. .
Fitzpatrick, K., Chrysler, S.T., Iragavarapu, V., Park, E.S., 2011. Detection distances to Van Der Horst, R., De Ridder, S., 2007. Influence of roadside infrastructure on driving
crosswalk markings: transverse lines, continental markings, and bar pairs. Transp. behavior: driving simulator study. Transp. Res. Rec. 2018 (1), 36–44.
Res. Rec. 2250 (1), 1–10. Varhelyi, A., 1998. Drivers’ speed behaviour at a zebra crossing: a case study. Accid. Anal.
Fu, T., Hu, W., Miranda-Moreno, L., Saunier, N., 2019. Investigating secondary pedes- Prev. 30 (6), 731–743.
trian-vehicle interactions at non-signalized intersections using vision-based trajectory World Health Organization, 2018. Global Status Report on Road Safety. Accessed Feb. 9,
data. Transp. Res. Part C Emerg. Technol. 105, 222–240. 2019. www.who.int/violence_injury_prevention/road_safety_status/2018/en/.
Gibby, A.R., Stites, J.L., Thurgood, G.S., Ferrara, T.C., 1994. Evaluation of Marked and Xu, S., 2012. The Study on the Discriminating Method of Driving Fatigue Based on
Unmarked Crosswalks at Intersections in California. Publication FHWA/CA/TO-94/1. Physiological Signal. Master Thesis. Beijing University of Technology, Beijing.
Chico State University. Zegeer, C.V., Stewart, R., Huang, H., Lagerwey, P., 2002. Safety Effects of Marked Versus
Herms, B.F., 1972. Pedestrian crosswalk study: accidents in painted and unpainted Unmarked Crosswalks at Uncontrolled Locations: Executive Summary and
crosswalks. Highway Res. Rec. 406, 1–13. Recommended Guidelines. Publication FHWA-RD-01-075, FHWA, U.S. Department of
Himanen, V., Kulmala, R., 1988. An application of logit models in analysing the beha- Transportation.
viour of pedestrians and car drivers on pedestrian crossings. Accid. Anal. Prev. 20 (3), Zhao, X., Li, J., Ma, J., Rong, J., 2016. Evaluation of the effects of school zone signs and
187–197. markings on speed reduction: a driving simulator study. SpringerPlus 5 (1), 789.
Horberry, T., Anderson, J., Regan, M.A., 2006. The possible safety benefits of enhanced Zhao, N., Zhao, X., Lin, Z., Song, Y., 2018a. Advance cuide sign design considering the
road markings: a driving simulator evaluation. Transp. Res. Part F Traffic Psychol. guidance demand of indirect reached freeway information. J. Beijing Univ. Technol.
Behav. 9 (1), 77–87. 44 (8), 1129–1135.
Huang, L., Zhao, X., Li, Y., Rong, J., 2019. Evaluation research of the effects of dia- Zhao, X., Ding, H., Lin, Z., Ma, J., Rong, J., 2018b. Effects of longitudinal speed reduction
grammatic guide signs with different complexities on driving behavior. Cogn. markings on left-turn direct connectors. Accid. Anal. Prev. 115, 41–52.
Technol. Work. 1–18. Zhao, X., Xu, W., Ma, J., Li, H., Chen, Y., 2019. An analysis of the relationship between
Iasmin, H., Kojima, A., Kubota, H., 2016. Safety effectiveness of pavement design treat- driver characteristics and driving safety using structural equation models. Transp.
ment at intersections: left turning vehicles and pedestrians on crosswalks. Iatss Res. Res. Part F 62, 529–545.
40 (1), 47–55.

12

You might also like