Accident Analysis and Prevention 81 (2015) 107–119

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

Driving simulator evaluation of drivers’ response to intersections with dynamic use of exit-lanes for left-turn Jing Zhao a,1, Meiping Yun b, * , H. Michael Zhang c,2 , Xiaoguang Yang b,3 a b c

Business School, University of Shanghai for Science and Technology, 516 Jungong Road, Shanghai, PR China Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, 4800 Cao’an Road, Shanghai, PR China Department of Civil & Environmental Engineering, University of California at Davis, Davis, USA

A R T I C L E I N F O

A B S T R A C T

Article history: Received 21 December 2014 Received in revised form 9 April 2015 Accepted 23 April 2015 Available online 14 May 2015

With the worsening of urban traffic congestion in large cities around the world, researchers have been looking for unconventional designs and/or controls to squeeze more capacity out of intersections, the most common bottlenecks of the road network. One of these innovative intersection designs, known as the exit-lanes for left-turn (EFL), opens up exit-lanes to be used by left-turn traffic with the help of an additional traffic light installed at the median opening (the pre-signal). This paper studies how drivers respond to EFL intersections with a series of driving simulator experiments. In our experiments, 64 drivers were recruited and divided into two groups. One group is trained to use the EFL while the other group is not. In addition, four scenarios were considered with different sign and marking designs and traffic conditions in the experiments. Results indicate that drivers show certain amount of confusion and hesitation when encountering an EFL intersection for the first time. They can be overcome, however, by increasing exposure through driver education or by cue provided from other vehicles. Moreover, drivers unfamiliar with EFL operation can make a left turn using the conventional left-turn lanes as usual. The EFL operation is not likely to pose any serious safety risk of the intersection in real life operations. ã 2015 Elsevier Ltd. All rights reserved.

Keywords: Intersections with exit-lanes for left-turn use Driver behavior Driving simulator Unconventional design

1. Introduction Facing with geometric limitations to expand intersections with heavy traffic demand, researchers have been looking for unconventional intersection designs to squeeze more capacity out of an intersection with oversaturated traffic conditions (Hummer, 1998b, a; Xuan et al., 2011; El Esawey and Sayed, 2013). One of these innovative intersection designs, as shown in Fig. 1, is to use some of the opposing through-lanes for left turns, which is known as the exitlanes for left-turn (EFL) intersection design (Zhao et al., 2013). The EFL intersection is proposed to increase the throughput of left-turn traffic in congested intersections with heavy left turn volumes and no space to add more left-turn bays. The function of lanes in the mixed-usage-area is variable. These lanes can be used

* Corresponding author. Tel.: +86 21 6959 5273; fax: +86 21 6959 5273. E-mail addresses: jing_zhao_traffi[email protected] (J. Zhao), [email protected] (M. Yun), [email protected] (H. M. Zhang), [email protected] (X. Yang). 1 Tel.: +86 21 6571 0430; fax: +86 21 6571 0430. 2 Distinguished Professor of Transportation Engineering, Tongji University, 4800 Cao’an Road, Shanghai, PR China. Tel.: +1 530 754 9203; fax: +1 530 752 7872. 3 Tel.: +86 21 6958 9475; fax: +86 21 6958 9475. http://dx.doi.org/10.1016/j.aap.2015.04.028 0001-4575/ ã 2015 Elsevier Ltd. All rights reserved.

as opposing through-lanes or left-turn lanes during different periods of a signal cycle. At the upstream of the intersection, there is a pre-signal and a median opening. Some of the left-turn vehicles can drive into the mixed-usage-area at the signalized median opening then turn left at the intersection using lanes in the mixedusage-area when they receive green light. This increase of the number of left-turn lanes could shorten the left turn phase, hence releasing green time for use by other conflicting movements, which in turn increases the capacity of the intersection. Fig. 2 shows how the EFL system would operate during one signal cycle. The pre-signal starts its cycle by giving the red to the left-turn vehicles while the left-turn phase of the cross-street is green. The pre-signal turns green for left-turn vehicles a few seconds after the cross-street left-turn green terminated. Then, some of the left-turn vehicles can drive into the mixed-usage-area from the median opening and queue at the intersection stop line. The left-turn vehicles could discharge more efficiently by using either left-turn lanes in the approach or the mixed-usage-area lanes during the left-turn phase. The pre-signal turns red a few seconds earlier than the start of the opposing-through green. In an earlier study (Zhao et al., 2013), an optimization model for the EFL control has been formulated, in which the geometric

108

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

Fig. 1. Exit-lanes for left-turn (EFL) control concept.

layout, main-signal timing and pre-signal timing were integrated. The EFL control could increase the capacity of an intersection with promising preliminary results especially under high left-turn demand compared with conventional intersection control. The highest improvement is about 50%. Moreover, the EFL control could be applied to one or multiple legs simultaneously, thus it is particularly useful for intersections with unbalanced left demand and degree of saturation among different travel directions. Generally, an additional leg using the EFL control leads to about 7% increase in intersection capacity.

Despite the promising results, it is not clear how drivers would respond to such an unconventional intersection when it is implemented in real life and if the benefits would fully materialize. For example, a left-turn driver unfamiliar with EFL operation might queue at the median opening at the upstream of the approach without driving into the mixed-usage-area when the pre-signal turns green, then blocks other left-turn drivers. At this point, careful driver behavior studies and adequate driver education have to be carried out before EFL control is perfected and implemented in the field. The current study was intended to understand motorist

Fig. 2. Operating process of EFL control.

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

behavior at EFL intersections and the effects of signing and marking on traffic safety of EFL intersections, so as to develop recommendations for the design, signage, and safe operation of EFL intersections. In this paper, a high fidelity driving simulator is used to study driver behavior at EFL intersections, which is a cost-effective way in examining driver responses under different traffic conditions, signage, and other design factors without posing any risk to drivers (Lee et al., 2003; Montella et al., 2011). The driving simulator technology is not new. It has been used in a wide range of studies on traffic safety, vehicle design, and impaired driving (Allen et al., 2007; Gelau et al., 2011), to name a few. In many cases, the use of an advanced driving simulator has many advantages over similar real-world or on-road driving research, including experimental control, efficiency, expense, safety, and ease of data collection (Nilsson, 1993). Moreover, it has been investigated that no significant differences between the road and the simulator for driving errors, lane maintenance, adjustment to stimuli, and visual scanning errors (Shechtman et al., 2009). It has also been reported that the driving simulator can be used as a valid tool to assess traffic safety at signalized intersections by comparing two perspectives, a traffic parameter (speed) and a safety parameter (crash history) between the data observed from the field and in the simulator (Yan et al., 2008). Last but not least, driving simulators have been used to evaluate the driver comprehension of unconventional or complex designs, such as the flashing yellow arrow indication (Knodler et al., 2006a,b), the diverging diamond interchange (Bared et al., 2007), the partial continuous flow intersection (Inman, 2009), and the guide signing at complex interchanges (Fitzpatrick et al., 2013). Therefore, a driving simulator can be an effective tool for addressing roadway design issues (Lee et al., 2011), explaining interaction between

109

drivers and roadway surroundings, and most importantly, exerting control over confounding factors such as the number or type of vehicles involved or the demographics of the driving population (Montella et al., 2011). The rest of the paper is organized as follows. In Section 2, the sign and marking schemes for an EFL intersection is introduced. The driving simulator experiments are described in Section 3. The findings for the two alternative sign and marking designs are analyzed in Section 4. Section 5 discusses the driver behavior issues of the EFL intersection. Conclusions and recommendations are given at the end of the paper. 2. Sign and marking design for EFL control Two navigation sign and marking design for an EFL intersection, as shown in Fig. 3, are implemented: the Scheme 1 is a simplified design, which includes the colored lane pavements for the mixedusage-area, a ground-mounted sign and two signal installations at the beginning of the mixed-usage-area on the nose of median; while the Scheme 2 is a full design, which adds an overhead reversible lane control sign centered over the opposing throughlanes slightly before the end of the mixed-usage-area and a ground-mounted sign at 80 m before the beginning of the mixedusage-area, using the Scheme 1 as a benchmark. The special geometric design issue of EFL intersection is the location and the length of the median opening. In practice, the location of the median opening should be optimized according to the traffic demand pattern and the geometric characteristics of the intersection (Zhao et al., 2013), while in this experiment, for fair comparison, all the locations of the median opening are 50 m before the intersection stop line. The width of the median is set to be 1.5 m as the minimum value specified in the code for design of

Fig. 3. Two schemes of sign and marking designs for EFL control approach.

110

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

Fig. 4. EFL approach guide sign.

urban road engineering in China (MOHURD, 2012). The length of the median opening could be calculated by the formula L  V 2 W=155 (SAC, 2009) for smooth moving purpose, where L represents the median opening length, m; V represents the design speed, km/h; and W presents the width of the lane changing, m. In this experiment, V is set to be 20 km/h and W is set to be 5 m (median width 1.5m + lane width 3.5 m). Consequently, the median opening length should be no shorter than L = 12.9 m. Meanwhile, it is harmful to setting the length of the median opening too long since it may cause two or more vehicles using the median opening simultaneously. Therefore, the length of the median opening is set to be 15 m in the experiment. One feature of the two designs is the colored lane pavement for the mixed-usage-area to provide information for the drivers that the special right-of-way of the mixed-usage-area. At the left side of the mixed-usage-area, there are conventional opposing throughlanes with the oncoming straight arrows showing their functions. Moreover, the mixed-usage-area is set to be a one-direction nopassing zone with markings, which consists of a normal broken line and a normal solid line where crossing is prohibited for the traffic traveling adjacent to the solid line (the direction of the approach left-turn). Besides these markings, an overhead reversible lane control sign is added in Scheme 2 to designate the allowable uses of the lane during different periods of a signal cycle. For left-turn traffic, the symbol of Red X should be shown when the lane is closed, and the symbol of left arrow should be shown when the lane is open only for left-turn travel. For opposing leaving traffic, the symbol of Red X should be shown when the lane is closed, and the symbol of through arrow should be shown when the lane is open only for leaving travel. At the beginning of the mixed-usage-area, there is a groundmounted EFL approach guide sign. It is a new guide sign designed to inform drivers of the EFL approach and help them understand

Fig. 5. Driving simulator in Tongji University.

the two left-turn alternatives, which is altered from the keep-left sign in China (SAC, 2009) (see Fig. 4). There are two reasons for recommending the use of the new EFL guide sign instead of the standardized keep-left sign. Firstly, the keep-left sign does not provide exact information of the EFL operation for drivers. In the EFL controlled approach, vehicles can accomplish a left-turn by using the mixed-usage-area or the regular left-turn lanes. Therefore, the left-turn drivers have two choices, and there is no need for them to keep left all the time. Moreover, the keep-left sign does not provide the information to those vehicles driving into the mixed-usage-area that they have to turn left at the intersection. Secondly, the symbolic sign of keep-left sign is most commonly used in the keep-right configuration, and it is worried that drivers would fail to notice the direction reversal and might be induced to keep right instead. Considering these two reasons, it is believed that the newly designed EFL approach guide sign would better convey the desired message, especially to drivers that do not closely study the sign or do not pay close attention to the driving task. In Scheme 2, an advanced ground-mounted sign is added at 80 m before the beginning of the mixed-usage-area to provide the drivers advanced information. It consists of a signal ahead sign and an EFL approach guide sign. The former is one of standardized warning signs in the national standards of the People’s Republic of China (SAC, 2009), and the latter is the same guide sign as that installed on the nose of the median. In the EFL controlled approach, the left-turn vehicles driving into the mixed-usage-area should be controlled by a traffic signal for safe operation. For this purpose, the signal installations are positioned at the same pole of the signs mentioned above. Two signal faces are

Table 1 Age and driving experience distribution of participants. Item

Groups

Drivers in China

All participants

Participants who completed the experiment

Number of drivers Distribution of age

/ 26–35 years old 36–50 years old 50 years old

286,598,050 100,625,397 (35.11%) 121,090,167 (42.25%) 64,882,486 (22.64%)

80 30 (37.5%) 30 (37.5%) 12 (15.0%) 8 (10.0%)

64 25 (39.1%) 24 (37.5%) 12 (18.8%) 3 (4.7%)

Distribution of driving experience

20 years

24,733,412 (8.63%) 78,556,526 (27.41%) 40,639,603 (14.18%) 78,413,226 (27.36%) 32,844,137 (11.46%) 20,978,977 (7.32%) 10,460,829 (3.65%)

6 (7.5%) 26 (32.5%) 14 (17.5%) 20 (25.0%) 8 (10.0%) 6 (7.5%) 0 (0.0%)

6 (9.4%) 23 (35.9%) 10 (15.6%) 16 (25%) 6 (9.4%) 3 (4.7%) 0 (0.0%)

Distribution of gender

Male Female

81.04% 18.96%

56 (70.0%) 24 (30.0%)

50 (78.1%) 14 (21.9%)

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

provided. The lower one has been shown for vehicles near the beginning of the mixed-usage-area to provide safer operation by reducing red signal violations, while the upper one has been shown for vehicles far from the beginning of the mixed-usage-area to provide advanced information of the signal control. 3. Driving simulator experiments 3.1. Goal This section describes an experimental driving simulator study to evaluate the performance of the sign and marking schemes and examine reaction of the left-turn drivers to the EFL intersection. Specific objectives of the study were to address these questions: a Would drivers navigate the EFL intersection correctly in their

first try? b Were the signs and markings adequate for drivers to comprehend? c How would drivers perform if they were fully informed of the

special operation procedure of the EFL intersection? d How would drivers perform if they were provided cues by other

vehicles? e Would drivers become confused and make specific types of

simulator discomfort includes two parts: subjective measures and objective measures. For the subjective measures, the drivers were asked if they feel dizzy, nauseous or any other uncomfortable just after the training. The evaluation rating scales were divided into three (little, medium and high). If the uncomfortable is medium or high evaluated by himself, the driver had to be excluded. For the objective measures, the postural stability of standing on the preferred leg is computed for each participant. Those whose time of maintaining stability after the training reducing more than 20% were excluded. Finally, 16 out of 80 participants had to be excluded. In order to compare the effects of familiarity on driver performance, the sixty-four participants (50 male, 14 female) who passed the training, were divided into two groups: participants who were fully informed of the special operation procedure of the EFL intersection (25 men, 7 women; with a mean age of 32.1 and having their driving license for 5.8 years averagely) and participants who encountered the EFL intersection for the first time (25 men, 7 women; with a mean age of 31.8 and having their driving license for 5.7 years averagely). For the trained participants, they were provided with a detailed explanation of the special operation procedure of the EFL intersection and taught how to accomplish the left-turn task more efficiently in the EFL approach by blueprints. They were also

errors such as driving too slowly so as to be trapped in the mixed-usage-area?

3.2. Driving simulator Fig. 5 shows the Tongji University driving simulator used in this study. Its dome houses a fully-instrumented Renault Megane III vehicle cab and is mounted on an 8 degree-of-freedom motion system with an X–Y motion range of 20  5 m. An immersive 5 projector system provides a front image view of 250  40 at 1000  1050 resolution refreshed at 60 Hz. LCD monitors provide rear views at the central and side mirror positions. SCANeRTM studio software displays the simulated roadway environment and controls a force feedback system that acquires data from the steering wheel, pedals, and gear shift lever. The validation of vehicle dynamics and visual images for this driving simulator used in this paper was applied in previous research (Chen et al., 2013; Xu et al., 2013; Wang et al., 2015a,b). The overall performance of this driving simulator was validated using three tests: simulator sickness, stop distance, and traffic sign size. The overall test results showed that the driving simulator satisfied the above three validation criteria. 3.3. Participants Eighty drivers (56 male, 24 female) participated in this study, whom are carefully selected according to the distribution of age and driving experience from the driver statistical analysis result in China (MPS, 2014), as Table 1 illustrates. All drivers were trained in the driving simulator in order to avoid simulator sickness which is common in urban road scenario with turning at intersections. The training contained three driving scenarios. In the first scenario, as Fig. 6(a) illustrates, the participants drove on a straight road to become familiar with the simulator dynamic view. In the second scenario, as Fig. 6(b) illustrates, the participants were asked to accelerate until above 100 km/h and then decelerate to stop just before a stop line. Finally, as Fig. 6(c) illustrates, the participants had to turn right and left at a road network. This extensive training was done to make sure that only drivers who are well-trained in urban environments participated in the study. Participants who appeared motion sickness had to be excluded from the experiment. The metrics of

111

Fig. 6. Three driving scenarios for training.

112

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

notified of their right to ask questions and to withdraw from the study at any time during the training. 3.4. Experimental design As Fig. 7 illustrates, an exact replica of the road network encircled by Fuxingmenwai Road, Nanlishi Road, Fuchengmenwai Road, and Sanlihe East Road in Beijing, China was generated in the driving simulator. Participants were administered a practice drive

along the Fuxingmenwai Road (approximately 5 min) and a test drive along the assigned path (approximately 10 min). The navigation instructions for the next intersection were given to drivers as soon as they pass the upstream intersection by voice command. The desire speed for the background vehicles in the experiment is set to be 40 km/h. For participants, the speed limit is set to be 60 km/h at the segment and 40 km/h at the approach. In Fig. 7(a), the four numbered intersections (1–4) are EFL intersections which having similar geometry characteristic, with

Fig. 7. Experiment road network.

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

113

Fig. 8. Two traffic condition for experiment.

three or four approach lanes and two or three exit lanes on each leg, as Fig. 7(c–f) illustrates. All the EFL intersections have exclusive left-turn lanes. The distances between them are also similar as illustrated in Fig. 7(a). For fair comparison, the signal timing of the four EFL intersections are the same. The duration of green of leftturn movement for the main-intersection is set to be 30 s; the duration of green of the pre-signal is set to be 30 s; and the presignal turns green for left-turn vehicles 15 s before the start of green at the main-intersection. Along the designed test route, drivers would encounter four EFL intersections and several conventional intersections. The four EFL

intersections consisted of four different driving scenarios, with the consideration of two environmental factors. Firstly, the traffic signs and markings designs with either Scheme 1 or 2 were used. Secondly, the traffic conditions with and without leading vehicles were simulated, as Fig. 8 illustrates. In the scenarios without leading vehicles, the participant’s vehicle would be intended to queue just at the beginning of the mixed-usage-area by setting several vehicles waiting for the green signal of left-turn when the participant’s vehicle arrived at the EFL intersections. Consequently, if the driver understood the design of the EFL intersection, he/she would drive into the mixed-usage-area when got the green pre-

Table 2 Four driving scenarios of EFL intersections. EFL intersection index

Signs and markings design

1

Scheme 1

2

Scheme 2

3

Scheme 1

4

Scheme 2

Traffic condition

Without lead vehicles Without lead vehicles With lead vehicles With lead vehicles

Geometric condition Lane assignment of EFL approaches

Number of receiving lanes

Length of the mixedusage-area

Length of the median opening

1LT + 1TH + 1RT

3

50 m

15 m

1LT + 2TH + 1TR

3

50 m

15 m

1LT + 1TH + 1TR

2

50 m

15 m

1LT + 1TH + 1TR

2

50 m

15 m

Note: the abbreviations LT, TH, TR, and RT indicate the exclusive left-turn lanes, through lanes, shared through-right lanes and exclusive right-turn lanes, respectively.

114

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

Table 3 Number of participants (out of 32) driving into the mixed-usage-area. Participant group Well-trained drivers Untrained drivers

x2

Number Percentage Number Percentage

p

Scenario 1

Scenario 2

Scenario 3

Scenario 4

18 56.3% 2 6.3% 18.62 0.05

travel speed were collected from the simulator experiments for analysis of driver response in relation to EFL design and operation.

signal to accomplish the left-turn task efficiently. In the scenarios with leading vehicles, the participant’s vehicle would be intended to queue behind the beginning of the mixed-usage-area. The vehicles between the beginning of the mixed-usage-area and the participant’s vehicle (lead vehicles) would drive into the mixedusage-area when the pre-signal displays green, and provide cue to the participant as a result. Table 2 shows the four driving scenarios of EFL intersections in a four-field matrix. According to the assigned test route, participants would meet the four scenarios following the order of scenario number. The lead vehicles could provide a stronger and more direct cue to the driver than the signs and markings. Therefore, the two scenarios without lead vehicle were tested first and the two scenarios with lead vehicle were tested later to compare the effect of the lead vehicles. Under the same traffic condition, the full message design may give more help for the drivers to understand the special operation of the EFL intersection. Therefore, the drivers will meet the intersection with simplified message design firstly.

A navigation error was recorded when a participant failed to turn left at the EFL intersection. No participants were observed failing to accomplish the left turn task in either driving scenario through the EFL intersections. This is mainly due to the fact that the EFL control provides two alternative routes to turn left: using the mixed-usage-area or using the left-turn lanes in the approach. Therefore, for these unfamiliar drivers, they can choose the latter route to accomplish the left turn task which is the same as the conventional intersections. From this point of view, the EFL intersection would not cause much adverse impact on unfamiliar drivers.

3.5. Data

4.2. Mixed-usage-area utilization

The main concern with the EFL design was that drivers cannot comprehend the passing mode of left-turn traffic. In the experiment, participants would be intended to queue just at the beginning of the mixed-usage-area as mentioned above. Therefore, if participants totally understood the design of the EFL intersection, they would drive into the mixed-usage-area as soon as they got the green signal at the pre-signal. Contrarily, if they did not understand the design, they would keep on waiting until getting the green signal at the intersection and then pass the intersection by using the left-turn lane in the approach. The participant might also be confused and not quite sure about their understanding, and behave differently (such as hesitating for a long time before turning into the mixed-usage-area or driving very slowly). To measure these tendencies and drivers’ comprehension of the EFL intersection, data related to navigation errors, mixed-usagearea utilization, red-light violations, wrong-way violations, and

Although there were no navigation errors at the EFL intersection, some participants did not have the awareness that the mixedusage-area could be used to make a left turn. Table 3 shows the number of participants driving into the mixed-usage-area at the pre-signal under the four scenarios of the two participant groups.

4. Results 4.1. Navigation errors

Fig. 9. Schematic for the definition of hesitation time.

Table 4 Duration of the hesitation time before driving into the mixed-usage-area. Participant group

Scenario 1

Scenario 2

Scenario 3

Scenario 4

Well-trained drivers

Average Variance Test of normality (1-sample K-S)

4.8 s 0.61 0.961 > 0.05

2.0 s 0.68 0.858 > 0.05

1.3 s 0.53 0.791 > 0.05

1.4 s 0.58 0.603 > 0.05

Untrained drivers

Average Variance Test of normality (1-sample K-S) Average

10.4 s 0.18 0.999 > 0.05

5.5 s 0.64 0.925 > 0.05

5.7 s 0.69 0.775 > 0.05

2.1 s 0.66 0.678 > 0.05

5.6 s

3.5 s

4.4 s

0.7 s

Increase

Test of homogeneity of variance (Levene statistic) Tests of between-subjects effects Model Participant Scenario Participant  scenario

F = 0.442, p = 0.874 > 0.05 F = 141.992, p < 0.001 F = 377.180, p < 0.001 F = 147.174, p < 0.001 F = 62.479, p < 0.001

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

115

Table 5 Multiple comparisons of hesitation time (LSD test). (I) group

(J) group

Mean difference (I

Scenario 1

Scenario 2 Scenario 3 Scenario 4

Scenario 2

J)

Std. error

Sig.

95% Confidence interval

2.3910* 2.0567* 3.6515*

0.33615 0.30198 0.29924

0.000 0.000 0.000

1.7271 1.4603 3.0605

3.0550 2.6532 4.2426

Scenario 1 Scenario 3 Scenario 4

2.3910* 0.3343 1.2605*

0.33615 0.26540 0.26228

0.000 0.210 0.000

3.0550 0.8585 0.7425

1.7271 0.1899 1.7785

Scenario 3

Scenario 1 Scenario 2 Scenario 4

2.0567* 0.3343 1.5948*

0.30198 0.26540 0.21677

0.000 0.210 0.000

2.6532 0.1899 1.1667

1.4603 0.8585 2.0229

Scenario 4

Scenario 1 Scenario 2 Scenario 3

3.6515* 1.2605* 1.5948*

0.29924 0.26228 0.21677

0.000 0.000 0.000

4.2426 1.7785 2.0229

3.0605 0.7425 1.1667

Lower bound

Note: * = the mean difference is significant at the 0.05 level.

Fig. 10. Mean travel speed in the EFL approach distribution.

Upper bound

116

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

Table 6 Travel speed characteristics. Group 1 Sample size Maximum speed (km/h) Minimum speed (km/h) Average speed (km/h) Variance of average Test of normality Test of homogeneity of variance ANOVA

Group 2

66 27 58.5 51.3 18.6 21.8 37.4 38.5 77.5 63.5 0.763 > 0.05 0.597 > 0.05 F = 1.179, p = 0.320 > 0.05 F = 12.306, p < 0.001

From the column comparison, one observes that untrained participants have low percentage entering the mixed-usage-area relative to well-trained participants. When there was no lead vehicle (under scenarios 1 and 2), the difference between participant groups was significant, x2 (2) = 18.62, p < 0.01, and x2 (2) = 10.66, p < 0.01, respectively. Once having the lead vehicle (under scenarios 3 and 4), there was no longer a significant difference between treatment groups (p > 0.05). Therefore, it is hard for the drivers to understand the special operation procedure of EFL intersections when encountering them for the first time. However, the situation would become much better once drivers were introduced the characteristic of the EFL control and explained how to use the mixed-usage-area. The knowledge can be gained by driver education or by cue provided from other vehicles. Moreover, comparing the performance of untrained drivers between scenarios 1 and 2, one observes that more participants drive into the mixed-usage-area under the full message design of signs and markings situation (signs and markings Scheme 2) than under the simplified design (Scheme 1). The difference in the number of drivers failing to drive into the mixed-usage-area as a function of signing for untrained drivers was statistically significant (x2 (2) = 4.27, p < 0.05). It might due to the fact that the overhead reversible lane control signs are familiar to most of drivers in Shanghai which has been implemented along Siping Road. for more than ten years. Therefore, the simplified design of signs and markings is not adequate for drivers' comprehension and the full message design is quite important. Besides the number of participants driving into the mixedusage-area, the hesitation time (as in Fig. 9), which means the time difference between the time drivers start the vehicle then going toward the mixed-usage-area and the usual response time at the intersection (taken as 1 second in the paper), was another factor to evaluate the level of driver comprehension. As Table 4 illustrates, the hesitation time of each group was examined for normality and homogeneity of variance using Kolmogorov–Smirnov and Levene’s tests, respectively. The tests of between-subjects effects show that there was a significant difference in the hesitation time under different participants (well-trained and untrained), different scenarios (scenarios 1–4). Along the same line as the number of participants driving into the mixed-usage-area, those untrained drivers hesitated for much longer time before driving into the

Group 3

Group 4

Group 5

Group 6

62 41.5 19.1 29.8 38.0 0.293 > 0.05

101 55.8 20.2 35.7 63.0 0.366 > 0.05

64 61.3 19.5 37.6 75.7 0.554 > 0.05

64 57.1 20.3 38.0 58.7 0.225 > 0.05

mixed-usage-area. Some of the untrained drivers did not even notice that they could use the opposing through-lanes to turn left. Since there are four scenarios groups, the post-hoc tests should be used to identify whether there were significant differences of hesitation time among four groups of scenarios and where the differences lay. Results (see Table 5) show that there was a significant difference between scenarios 1 and 2, and a significant difference between scenarios 3 and 4. It indicated that more navigation signs were necessary. Therefore, in practice, more measures could be used to remind the drivers to drive into the mixed-usage-area, such as flashing or sounding traffic signs. 4.3. Red-light violations A red-light violation was recorded when a participant drove into the mix-usage-area during the red signal. The simulation scenarios were designed such that the drivers would come up to red signals at all the four EFL intersections. No participants failed to stop and violated the red signal in all the 256 instances. There may be two main reasons to explain this result. First, the driving scenarios were generated based on the real-world urban streets in China, and provides a sense of realism and familiarity to the participants. Second, when the participants arrived at the EFL intersections initial queuing would make them decelerate before arriving at the intersection, greatly reducing the chance of red light violations. Regardless of the reasons, an EFL intersection seems not to cause additional red-light violations. 4.4. Wrong-way violations A wrong-way violation was recorded when a participant drove into the conventional opposing through-lanes (not the mix-usagearea) while making a left turn. These drivers understood that the opposing through-lanes could be used to make a left turn, which was the main design idea of EFL intersections, but failed to recognize that they can only use the mixed-usage-area, a onedirection no-passing zone, to make the left turn. In the total of 162 instances in which the participant turn left by using the mixedusage-area, only 2 drivers (about 1.2%) drove into the conventional opposing through-lanes. Both of them were in the untrained drivers group. One occurred in scenario 1 and the other in scenario

Table 7 Acceleration characteristics. Group 1 Sample size Maximum (m/s2) Minimum (m/s2) Average (m/s2) Variance of average Test of normality Test of homogeneity of variance ANOVA

Group 2

66 27 2.78 2.56 1.43 1.66 2.13 2.11 0.077 0.098 0.249 > 0.05 0.656 > 0.05 F = 1.910, p = 0.092 > 0.05 F = 0.989, p = 0.424 > 0.05

Group 3

Group 4

Group 5

Group 6

62 2.71 1.39 2.12 0.083 0.677 > 0.05

101 2.93 1.55 2.21 0.111 0.221 > 0.05

64 2.83 1.44 2.19 0.117 0.587 > 0.05

64 2.79 1.47 2.15 0.118 0.428 > 0.05

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

117

Table 8 Deceleration characteristics. Group 1 Sample size Maximum (m/s2) Minimum (m/s2) Average (m/s2) Variance of average Test of normality Test of homogeneity of variance ANOVA

Group 2

66 27 3.12 2.95 1.66 1.64 2.37 2.28 0.155 0.169 0.255 > 0.05 0.368 > 0.05 F = 0.569, p = 0.724 > 0.05 F = 1.460, p = 0.202 > 0.05

3. It may be owing to the colored lane pavements and the overhead reversible lane control sign, which clearly informing the special right-of-way of the mixed-usage-area. 4.5. Driving behaviors at the mix-usage-area Three performance indices, the mean travel speed, the acceleration, and the deceleration, were collected in the experiment to exam if participants would feel confused when driving in the mixed-usage-area. Six different types of groups are considered: group 1: untrained drivers who did not drive into the mixed-usagearea; group 2: well-trained drivers who did not drive into the mixed-usage-area; group 3: untrained drivers who drove into the mixed-usage-area; group 4: well-trained drivers who drove into the mixed-usage-area; group 5: untrained drivers at the conventional intersections; and group 6: well-trained drivers at the conventional intersections.

Group 3

Group 4

Group 5

Group 6

62 2.96 1.56 2.29 0.137 0.503 > 0.05

101 3.23 1.58 2.34 0.134 0.376 > 0.05

64 3.45 1.71 2.44 0.168 0.791 > 0.05

64 3.32 1.63 2.42 0.164 0.589 > 0.05

The distribution of mean travel speed for different driver groups are shown in Fig. 10. Results of analysis of variance (ANOVA) for the three performance indices are shown in Tables 6–8, respectively. All data were assessed for normality and homogeneity of variance prior to ANOVA. Results of ANOVA shows that the difference in travel speed under different groups was statistically significant (F = 19.717, p < 0.001, see Table 6), while there is no significant difference in the acceleration (F = 0.989, p = 0.424 > 0.05, see Table 7) and the deceleration characteristics (F = 1.460, p = 0.202 > 0.05, see Table 8). Furthermore, pairwise multiple comparisons were used to identify whether there were significant differences of travel speed among the six groups of drivers and where the differences lay, as Table 9 illustrates. Results indicated that only the group of untrained drivers who drove into the mixed-usage-area differs from the other five groups, as Table 6 illustrates. The mean speed for untrained drivers who drove into the mixed-usage-area was 5.9 km/h lower

Table 9 Multiple comparisons of travel speed (LSD test). (I) group

(J) group

Group 1

Group Group Group Group Group

2 3 4 5 6

Group 2

Group Group Group Group Group

Group 3

Mean difference (I

J)

Std. error

Sig.

95% Confidence interval

1.0909 7.6059* 1.7061 0.1565 0.5425

1.81026 1.40149 1.25423 1.39015 1.39015

0.547 0.000 0.175 0.910 0.697

4.6503 4.8502 0.7600 2.8899 3.2759

2.4685 10.3616 4.1723 2.5769 2.1909

1 3 4 5 6

1.0909 8.6968* 2.7970 0.9344 0.5484

1.81026 1.82713 1.71678 1.81845 1.81845

0.547 0.000 0.104 0.608 0.763

2.4685 5.1042 0.5786 2.6412 3.0271

4.6503 12.2894 6.1727 4.5099 4.1240

Group Group Group Group Group

1 2 4 5 6

7.6059* 8.6968* 5.8997* 7.7624* 8.1483*

1.40149 1.82713 1.27847 1.41206 1.41206

0.000 0.000 0.000 0.000 0.000

10.3616 12.2894 8.4135 10.5389 10.9248

4.8502 5.1042 3.3859 4.9859 5.3719

Group 4

Group Group Group Group Group

1 2 3 5 6

1.7061 2.7970 5.8997* 1.8627 2.2486

1.25423 1.71678 1.27847 1.26603 1.26603

0.175 0.104 0.000 0.142 0.077

4.1723 6.1727 3.3859 4.3520 4.7379

0.7600 0.5786 8.4135 0.6267 0.2408

Group 5

Group Group Group Group Group

1 2 3 4 6

0.1565 0.9344 7.7624* 1.8627 0.3859

1.39015 1.81845 1.41206 1.26603 1.40080

0.910 0.608 0.000 0.142 0.783

2.5769 4.5099 4.9859 0.6267 3.1403

2.8899 2.6412 10.5389 4.3520 2.3684

Group 6

Group Group Group Group Group

1 2 3 4 5

0.5425 0.5484 8.1483* 2.2486 0.3859

1.39015 1.81845 1.41206 1.26603 1.40080

0.697 0.763 0.000 0.077 0.783

2.1909 4.1240 5.3719 0.2408 2.3684

3.2759 3.0271 10.9248 4.7379 3.1403

Lower bound

Note: * = the mean difference is significant at the 0.05 level.

Upper bound

118

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119

than that for the well-trained drivers who drove into the mixedusage-area, 7.6 km/h lower than that for untrained drivers who did not drive into the mixed-usage-area, and 7.8 km/h lower than that for untrained drivers at the conventional intersections. The difference in mean speed shows that the well-trained participants have more selfconfidence in having the right-of-way in the mixed-usage-area during this period of time.

concern of vehicle being trapped in the mixed-usage-area is also not warranted since only a slight reduction of travel speed was measured which could be addressed by coordinating of mainsignal and pre-signal. Moreover, the EFL intersection does not cause higher safety risk to unfamiliar drivers since they can also accomplish the left turn task by using the regular left-turn lanes in the approach as usual. Therefore, the EFL intersection is not likely to adversely affect the safety of the intersection.

5. Discussions The results of the experiments show that it is difficult for participants to comprehend the special operation procedure of the EFL intersection if having no training or previous experience. However, it is quite easy for drivers to handle EFL operations once they know how to accomplish the left turn task more efficiently by using the mixed-usage-area. Therefore, adequate driver education should be carried out before EFL controls are implemented in the field. Especially during the early days of the operation, a traffic police is recommended to manage the EFL approach to guide drivers unfamiliar with EFL operations. With respect to the signs and markings of EFL intersections, the full message scheme performs much better in mixed-usage-area utilization, wrong-way violations and travel speed analysis. Especially, the overhand reversible lane control sign plays an important role in guiding participants to drive into the mixedusage-area and not to pass into the conventional opposing through-lanes. It is uncertain whether the colored lane pavement for the mixed-usage-area was responsible for providing the special right-of-way information to drivers in this study. However, this pavement marking treatment is relatively inexpensive and should be considered, particularly at the first time using the EFL intersection in a city. One of the greatest safety concerns with EFL intersection is the possibility of red-light violations at the pre-signal, which may result in head-on collisions. The results of the experiment suggest that this concern is not warranted. No participants drove into the mixed-usage-area during the red signal at the median opening. Furthermore, even though some participants were unfamiliar with such operations, they can also accomplish the left turn task by using the left-turn lanes in the approach as they used to do at the conventional intersections. From this point of view, the EFL intersection would not cause additional safety problems. 6. Conclusions Through the driving simulator study of the design of and driver response to EFL intersections, the following conclusions are drawn. (1) It is a ubiquitous comprehension problem when drivers are

confronted with an EFL intersection for the first time, although the EFL operation is not difficult to understand. However, this can be easily overcome by driver training or by learning from cues provided from other vehicles that are using EFL. Therefore, adequate driver training/education is recommended prior to field EFL operations. (2) Considering the special nature of EFL operation, new EFL guide signs are recommended for smooth operation. Among the two signage schemes studied, the full signing scheme is recommended in which the overhead reversible lane control sign is effective in improving mixed-usage-area utilization and reducing wrong-way violations since it provides a clearer guidance and the right-of-way information of the mixedusage-lanes. (3) The results of our experiments suggest that the concern of headon collision in the mixed-usage-lanes may not be warranted since no participants ran red light at the pre-signal. And the

Due to the limited number of participants, experiments concerning other signal operation parameters, such as the saturation flow rate and headway at the pre-signal and the intersection approach, were not performed in the current study. These parameters are perhaps best observed in field operations and be used to perfect EFL control. Furthermore, for this driving simulator based research, some limitations should be mentioned. Primarily, the driving environment in the simulator generally differs from that in the real world. The negative factor such as violations of other road users, marking degradation, limited sight distance, etc., are not considered in this experiment. Moreover, in real-world, the drivers' attention may be declined by talking, phoning, fatigue driving, etc., which increase the uncertainty of driving behavior. Acknowledgements The research is supported by the National Natural Science Foundation of China under Grant Nos. 51178344 and 51238008, and the PhD research startup foundation of University of Shanghai for Science and Technology under Grant No. BSQD201504.

References Allen, R., Cook, M., Rosenthal, T., 2007. Application of driving simulation to road safety. Adv. Transport. Stud. 5–8 (special issue). Bared, J.G., Granda, T., Zineddin, A., 2007. Drivers’ Evaluation of the Diverging Diamond Interchange. Federal Highway Administration, Washington, D.C. Chen, M., Wang, X., Zhang, H., Wang, L., You, S., 2013. Driving simulator study for drivers’ response times under rear-end scenarios. Proceedings of the 4th International Conference on Transportation Engineering, Chengdu, China, pp. 1812–1819. El Esawey, M., Sayed, T., 2013. Analysis of unconventional arterial intersection designs (UAIDs): state-of-the-art methodologies and future research directions. Transportmetr. A: Transport Sci. 9 (10), 860–895. Fitzpatrick, K., Chrysler, S.T., Nelson, A.A., Iragavarapu, V., 2013. Driving simulator study of signing for complex interchanges. Proceedings of the Transportation Research Board 92nd Annual Meeting, Washington, D.C., pp. 1–22. Gelau, C., Sirek, J., Dahmen-Zimmer, K., 2011. Effects of time pressure on left-turn decisions of elderly drivers in a fixed-base driving simulator. Transport. Res. F 14 (1), 76–86. Hummer, J.E., 1998a. Unconventional left-turn alternatives for urban and suburban arterials—part two. ITE J. 68 (11), 101–106. Hummer, J.E., 1998b. Unconventional left-turn alternatives for urban and suburban arterials—part one. ITE J. 68 (9), 26–29. Inman, V.W., 2009. Evaluation of signs and markings for partial continuous flow intersection. Transport. Res. Rec. 2138 (1), 66–74. Knodler, M.A., Noyce, D.A., Kacir, K.C., Brehmer, C.L., 2006a. Analysis of driver and pedestrian comprehension of requirements for permissive left-turn applications. Transport. Res. Rec. 1982 (1), 65–75. Knodler, M.A., Noyce, D.A., Kacir, K.C., Brehmer, C.L., 2006b. Potential application of flashing yellow arrow permissive indication in separated left-turn lanes. Transport. Res. Rec. 1973 (1), 10–17. Lee, H.C., Cameron, D., Lee, A.H., 2003. Assessing the driving performance of older adult drivers: on-road versus simulated driving. Accid. Anal. Prev. 35 (5), 797–803. Lee, J.D., Mcgehee, D.V., Brown, J.L., Richard, C.M., Ahmad, O., Ward, N.J., Hallmark, S., Lee, J., 2011. Matching simulator characteristics to highway design problems. Transport. Res. Rec. 2248 (1), 53–60. MOHURD, 2012. Code for Design of Urban Road Engineering. Ministry of Housing and Urban–Rural Development of the People’s Republic of China, Beijing. Montella, A., Aria, M., D’ambrosio, A., Galante, F., Mauriello, F., Pernetti, M., 2011. Simulator evaluation of drivers’ speed, deceleration and lateral position at rural intersections in relation to different perceptual cues. Accid. Anal. Prev. 43 (6), 2072–2084. MPS, 2014. China Vehicle and Driver Statistical Analysis in April, 2014. The Ministry of Public Security of The People's Republic of China, Beijing.

J. Zhao et al. / Accident Analysis and Prevention 81 (2015) 107–119 Nilsson, L., 1993. Behavioural research in an advanced driving simulatorexperiences of the VTI system. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, Seattle, Washington, pp. 612–616. SAC, 2009. Road Traffic Signs and Markings. Standardization Administration of the People’s Republic of China, Beijing. Shechtman, O., Classen, S., Awadzi, K., Mann, W., 2009. Comparison of driving errors between on-the-road and simulated driving assessment: a validation study. Traffic Inj. Prev. 10 (4), 379–385. Wang, X., Wang, T., Tarko, A., Tremont, P.J., 2015a. The influence of combined alignments on lateral acceleration on mountainous freeways: a driving simulator study. Accid. Anal. Prev. 76, 110–117. Wang, X., Xu, C., Chen, X., 2015b. Driver drowsiness detection based on nonintrusive metrics considering individual specifics. Washington, D.C.Proceedings of the Transportation Research Board 94th Annual Meeting, 015. , pp. 1–17.

119

Xu, C., Wang, X., Chen, X., 2013. Modeling fatigue level by driver’s lane-keeping indicators. Chengdu, ChinaProceedings of the 4th International Conference on Transportation Engineering, 013. , pp. 2282–2288. Xuan, Y., Daganzo, C.F., Cassidy, M.J., 2011. Increasing the capacity of signalized intersections with separate left turn phases. Transport. Res. B: Methodol. 45 (5), 769–781. Yan, X., Abdel-Aty, M., Radwan, E., Wang, X., Chilakapati, P., 2008. Validating a driving simulator using surrogate safety measures. Accid. Anal. Prev. 40 (1), 274–288. Zhao, J., Ma, W., Zhang, H.M., Yang, X., 2013. Increasing the capacity of signalized intersections with dynamic use of exit lanes for left-turn traffic. Transport. Res. Rec. 2355 (1), 49–59.