A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems

Transportation Research Part F xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Transportation Research Part F journal homepage: www.elsevier.com/locate/trf

A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems Yung-Ching Liu a,⇑, Chin Heng Ho b a b

Department of Industrial Engineering and Management, National Yunlin University of Science and Technology, 123 University Rd., Sec. 3, Douliu, Yunlin, Taiwan Acer Incorporated, 26F, 116, Sec. 1, Xintai 5th Rd., Xizhi, New Taipei City, Taiwan

a r t i c l e

i n f o

Article history: Received 8 February 2017 Received in revised form 9 July 2017 Accepted 25 September 2017 Available online xxxx Keywords: Forward collision warning (FCW) Statistical quality control (SQC) chart Car following Driving simulator Aggressive driving Usability

a b s t r a c t This study aims to evaluate the usability of the forward collision warning (FCW) system as adopted by the statistical quality control (SQC) chart design concepts on drivers’ car following behaviors and task performance. A total of 48 highly aggressive and 48 less aggressive drivers participated in a two (aggressive driving: high vs. low; between-subjects) by two (driving workload: high vs. low; within-subjects) by three (the FCW system: fivestages vs. X-bar vs. X-bar plus exponentially weighted moving-average (EWMA) control charts; between-subjects) mixed-factorial simulation experiment. The drivers’ behaviors, response time to divided attention (DA) tasks, as well as subjective workload and trust ratings were collected. Findings showed that drivers with the FCW’s assistance improved their car-following behaviors and that the FCWs with the SQC chart design concepts showed better results than the five-stage system. Drivers who used both SQC FCWs performed correspondingly in their car-following behaviors. However, the X-bar FCW aided drivers in responding to DA tasks much faster, and drivers felt less stressed and had more trust in using the X-bar FCW system than those who used the X-bar + EWMA FCW system. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Previous research found that between a quarter to a third of all road accidents can be attributed to rear-end collisions (National Highway Traffic Safety Administration (NHTSA), 2012, Wierwille, Lee, DeHart, & Perel, 2006). It was shown that rear-end collisions are currently the most common type of vehicle-related accidents that constitute a major traffic and even social problem, requiring urgent worldwide attention. Failure by drivers to maintain a safe time headway is a result of aggressive driving, negative driving conditions (e.g., fatigue and drunk driving), distractions, or inattention. It is therefore a probability that unsafe time headway is playing a factor in traffic collisions (Deffenbacher, Lynch, Filetti, Dahlen, & Oetting, 2003; Ellison-Potter, Bell, & Deffenbacher, 2001). Furthermore, aggressive driving is considered an emotion composed of anger-related feelings and thoughts provoked by various traffic situations. Numerous studies show driving anger to be associated with aggressive driving behaviors that can possibly lead to reckless driving actions, such as speeding, excessive lane changing, tailgating, and improper passing (Deffenbacher et al., 2003; Ellison-Potter et al., 2001; Fernandes, Hatfield, & Soames Job, 2010; Iversen & Rundmo, 2002; Lajunen, ⇑ Corresponding author. E-mail address: [email protected] (Y.-C. Liu). https://doi.org/10.1016/j.trf.2017.09.010 1369-8478/Ó 2017 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

2

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

Parker, & Summala, 1999; Mesken, Hagenzieker, Rothengatter, & Dewaard, 2007; Moore & Dahlen, 2008; Ulleberg, 2002). Such driving behavior can consequently result in a high risk of traffic accidents or injuries. Previous studies did confirm the link between aggressive driving behavior and traffic accidents (Heino, van der Molen, & Wilde, 1996; Iversen & Rundmo, 2002; Ranney, 1999). Information technologies have led to different vehicle manufacturers taking an interest in the development of forward collision warning (FCW) systems. Dingus et al. (1997) used three on-road experiments to determine how headway maintenance and FCW displays influence driver behavior. The authors also indicated that when drivers were provided with salient visual warnings presented by well-designed FCW displays, they effectively increased their time headway between their vehicle and the lead vehicle. In addition, auditory warnings were less effective than visual warnings for maintaining a safe headway, but they were nevertheless helpful for decreasing reaction time for deceleration. Maltz and Shinar (2007) argued that distracted drivers increased their temporal headway with the lead vehicle by using a less reliable collision avoidance warning system and that such drivers, by contrast, maintained shorter headways with warning systems having higher reliability levels. Ben-Yaaco, Maltz, and Shinar (2002) used auditory warnings to alert drivers that the temporal headway between the host vehicle and lead vehicle was too short. They reported that drivers often overestimated their temporal headways to lead vehicles, a behavior resulting in dangerous traffic conditions. In-vehicle collision avoidance warning systems have been proven to effectively aid at maintaining an appropriate headway, as well as at reducing the risk of rear-end crashes. A majority of studies have shown the benefits of FCW systems in reducing the number and severity of rear-end collisions (Lee, McGehee, Brown, & Reyes, 2002; Lee, Ries, McGehee, Brown, & Perel, 2000; Mohebbi, Gray, & Tan, 2009; Scott & Gray, 2008). However, previous studies rarely considered individual differences in car-following behaviors. Since the safety distance of headway for each driver’s perception varies, Jamson, Lai, and Carsten (2008)proposed an adaptive FCW system that adjusted the timing of warnings according to each driver’s reaction time. Although this did not have significant implications for the adaptive and non-adaptive systems for non-aggressive drivers, the benefits of the adaptive system for aggressive drivers were clearly demonstrated. Aggressive drivers reported a preference for the adaptive system: they rated it as less ‘‘stressinducing” and more ‘‘safety-enhancing” compared to the non-adaptive system. We therefore intend to likewise develop an adaptive FCW system by using the statistical quality control chart concept, with the additional aim to determine the benefits of such system. Individual differences and ‘‘monitoring in time” are features of control charts that are consequently well suited to evaluate drivers’ behaviors in following cars while using the FCW systems. Traditionally, statistical quality control charts were developed to determine process capabilities and to evaluate process performance in industries. They have developed into useful SQC techniques for monitoring the stability of manufacturing processes and the quality of products. In addition, applications of control chart concepts in other domains have accurately detected departures from the average in practice cases (Maguad, 2005). Some of these domains included human performance and designing warning systems (i.e., the X-bar control charts) to diagnose work-related asthma (Hayati, Maghsoodloo, Devivo, & Carnahan, 2006), to monitor several channels of electroencephalogram (EEG) and electrooculogram (EOG) signals (Cannon, Krokhmal, Chen, & Murphey, 2012). The concept of control charts has been applied to warning systems. Hwang et al. (2008) used X-bar control charts to design a pre-alarm system for a nuclear power plant control room, and then compared the performances of three types of systems: text, graphic pre-alarm, and the original. The results showed that participants had lower mental workloads and higher secondary-task performances when monitoring either type of pre-alarm system along with an X-bar control chart. Moreover, the pre-alarm systems with X-bar control charts effectively provided warnings to operators’ monitoring tasks. For human performances, Ong, Harvey, Shehab, Dechert, and Darisipudi (2004) used three tasks (i.e., identification of outof-control points, estimates of process means, and identification of process patterns) to investigate the effectiveness of three control charts: the X-bar, EWMA, and cumulative sum (CUSUM) charts. Their research results showed that each chart performed well in the identification of out-of-control points. For the mean estimation task, both the X-bar and EWMA charts yielded similar accuracy, and participants produced the fastest reactions by using the X-bar charts. In addition, participants reported higher subjective preferences for X-bar chart in all tasks compared to the other respective charts. The advantages of the X-bar charts make it widely used in industries, and findings implied that X-bar charts were a very appropriate technique to monitor or evaluate human performances in identification of out-of-control points, estimates of process means. Shehab and Schlegel (2000) also adopted X-bar, EWMA, and CUSUM charts to monitor and classify the cognitive and physical performance. The three control charts were tested in 174 trials involving 10 participants and 23 cognitive performance assessment measures. Results indicated that continuous performance measures, such as reaction time, were best examined with EWMA charts. However, X-bar control charts were only moderately effective for these data, with CUSUM charts proving to be relatively ineffective. In addition, X-bar control charts most effectively detected a single time outlier caused by a sudden event. The EWMA chart was more effective in detecting accumulative performance shift produced by continuing small effects, such as fatigue or sleep deprivation. In light of the aforementioned research findings, the aims of this study are to evaluate the effects of adopting FCW systems on two levels of angry drivers’ car-following behaviors, and to examine the differences among drivers’ on-road performances and subjective workloads using different FCW systems.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

3

2. Interface designs of the forward collision warning information systems 2.1. Description of the FCW systems To build up the SQC charts used in the FCW systems, this study adopted 3 Hz as the time headway sampling frequency, and each participant was given ten minutes at the start of the experiment to build up the charts, starting from the five minute mark. The EWMA and X-bar control charts are commonly combined to form a robust univariate control charting scheme, as well asto effectively monitor performances in these fields (Montgomery, 2005). The EWMA chart detects small process shifts, while the X-bar control chart can easily detect sudden changes, such as outlier points. Therefore, this study applied these two types of control charts, the X-bar control chart and the X-bar + EWMA charts, to the designed SQC FCW interfaces. The control line limits of an X-bar control chart are defined as follows:

Upper Control Limit ðUCLÞ ¼ lw þ 3rw Central Line ðCLÞ ¼ lw Lower Control Limit ðLCLÞ ¼ lw  3rw ; where lw is the mean of the monitored parameter w, and rw is the SD of the monitored parameter w population. The control limits of the EWMA chart are defined as follows:

Upper Control Limit ðUCLÞ ¼ l0 þ Lrz Central Line ðCLÞ ¼ l0 Lower Control Limit ðLCLÞ ¼ l0  Lrz ; Where l0 is the mean of monitored historical data, and rz is the standard deviation of the EWMA statistic. The parameter L represents the multiple of the rational subgroup SD (standard deviation), is selected before an EWMA chart is applied, and is typically set at 3 to match other control charts. The FCW systems that are currently available on the car market adopt the so-called stage warning style. The car company set the default warning stages, and once the driver’s car following condition is in that stage, a warning will be activated. The number of warning stages the driver is provided with is also referred to as the impact of the acceptance of FCW systems. Previous studies have indicated the potential benefits of two-stage or multistage warnings in traffic safety (Campbell, Richard, Brown, & McCallum, 2007; Dingus et al., 1997; Lee, Mcgehee, Dingus, & Wilson, 1998). General Motors Corporation and Delphi-Delco Electronic Systems (2002) recommended a five-stage warning display with the following additional icons to alert drivers with FCW information: no-vehicle detected, vehicle detected, caution, approach imminent, and imminent. In the five-stage mode of the FCW system, this study adopted the warning headway timings by Dingus et al. (1997), with the safety zone corresponding to a time headway of 1.6 s or greater. Three caution zones indicate time headways between 1.0 s and 1.6 s, whereas the danger zone corresponds to a time headway below 1.0 s. The above systems monitor actual time-related driving performances (i.e., time headway with lead vehicle), and set various warning parameters in different FCW systems that send warnings once the driver follows the lead vehicle into a different warning zone. Table 1 shows the time headway intervals of the five warning zones for the three FCW systems. 2.2. The forward collision warning interface The visual FCW displays were designed to combine an analog-based 4-scale display with a looming vehicle icon. The analog-based scale display provides a graphical representation of continual information with the last warning symbol (zone 5), a pictorial red car, flashing on and off at 4 Hz. Auditory warnings were given in zone 4 (a continuous 2 Hz, 75 dB dong sound) and zone 5 (a mixed of dong and buzzer sounds) and they alerted the driver in a timely manner of urgent and imminent crashes. The multimodal display introduced in this study is a simple combination of auditory and visual displays. 3. Methodology 3.1. Participants Forty-eight participants who scored in the upper quartiles of the U.K. Driving Anger Scale (UKDAS > 84) and a further 48 scoring in the lower quartiles (UKDAS < 63) on the 21-item UKDAS (Lajunen, Parker, & Stradling, 1998) were voluntarily recruited as the respective high-aggressive vs. low-aggressive groups.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

4

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

Table 1 The mean headway conditions, warning icons and display modalities of five warning zones for the three FCW systems. Warning zones

Five-stage

X-bar control chart

EWMA +X-bar control charts

1.6 s or greater

Outside upper control limit or in control

Outside upper control limit or in control for both of the EWMA and X-bar control charts

1.4–1.6 s

Points fall between the centerline and the 1r limit

EWMA: in control1 X-bar: Points fall between the centerline and the 1r limit

1.2–1.4 s

Points fall between the 1r limit and the 2r limit

EWMA: in control X-bar: Points fall between the 1r limit and the 2r limit

1.0–1.2 s

Points fall between the 2r limit and the 3r limit

EWMA: in control X-bar: Points fall between the 2r limit and the 3r limit

Below 1.0 s

Outside lower control limit

EWMA: Outside lower control limit or X-bar: Outside lower control limit

Zone 1 (V*)

Zone 2 (V)

Zone 3 (V)

Zone 4 (V+A)

Zone 5 (V+A) *

The warning display modality: V: visual, A: auditory, V+A: the multimodality. 1 According to the formula of calculating the EWMA control lines mentioned above, the EWMA lower control line is dynamic based on the entire history of output with different weighted values. EWMA weights samples in geometrically decreasing order so that the most recent samples are weighted more than the past samples. In this way, the term ‘‘in control” indicated that the drivers’ car following measure monitored by the EWMA control was normal, and the X-bar + EWMA FCW system made its decision on whether or not to warn drivers based on the X-bar control results. However, if the EWMA control result is out of its lower control line, the X-bar + EWMA FCW system directly provides the drivers with the most serious warning information: the warning zone 5.

Each participant had a valid driver’s license for at least one year, had to meet normal vision requirements (a minimum of 18/20, before or after correction), passed the Ishihara Color Card Blindness Test, and demonstrated normal hearing. Exclusion criteria included prior experience using a driving simulator or head-up display (HUD). Each participant was involved in two experiments with respective high- and low-driving workloads. Each participant who completed the experiment received aUS $10 compensation. 3.2. Apparatus UK driving anger scale. The UKDAS (Lajunen et al., 1998) consists of 21 items describing driving situations that may potentially provoke a driver’s aggressiveness. The items are divided into three subscales: impeded progress (9 items), reckless driving (9 items), and direct hostility (3 items). Respondents rated the amount of aggressiveness generated by each situation on a 5-point Likert scale (1 = not at all; 2 = a little; 3 = some; 4 = much; 5 = very much). The Chinese translations were done according to the meaning of the UKDAS items to render them meaningful to Taiwanese drivers. The STI driving simulator. This study uses the interactive STIÒ driving simulator (Systems Technology Inc., Hawthorne, CA, USA). The simulated vehicle cab, a VOLVO 340 DL, features all standard automotive displays and controls (e.g., steering wheel, brakes, and accelerator) in a vehicle with automatic transmission. The driving scenario is projected onto a 100-in. [(Mocom Power ScreenÒ; screen type: aluminum concave; dimensions (W  H): 200  150 cm; curvature: 900 cm; brightness: 20 gains)] screen located 3.1 m in front of the driver with the sound effects of the vehicle broadcasted by two-channel amplifiers. Head-up display. Information related to driving, such as the vehicle speed, traffic signs, and the FCW visual information, is projected on a screen located 3.1 m in front of the driver by the projector. The vertical projection angle is maintained Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

5

between 6° and 12° below the driver’s horizontal line of vision, and the HUD area measures approximately 45 (w)  30 (h) cm. The display resolution is 1024  768 dpi; the presentation font (icon) size is 12  12 cm2 (approximately 2.0°). Forward collision warning system. The FCW system was developed using Visual Basic programming language. The system is able to communicate with the STI driving simulator using RS232 serial cables and adapters, and it can retrieve and decode related driving performances from the STI driving simulator. The visual warning information was presented on the HUD, and a speaker located right behind the dashboard was used as the system auditory display. Fig. 1 presents the HUD, the visual FCW system, and the driving road scenes. 3.3. Tasks Driving tasks. Participants were instructed to proceed to a destination quickly and to complete the simulated driving course while driving safely by following all traffic rules. Car following tasks. Participants were instructed to drive as they normally would and to follow the lead vehicle at a safe distance (as specified by the in-car FCW display), even if doing so might lead to breaking the speed limit. Four types of sudden events resulted from lead vehicles braking in various manners for each road section: coasting deceleration, normal braking, emergency braking, and rear-end collisions. The following section will describe lead vehicle deceleration causing the respective sudden events: (a) Coasting deceleration. The lead vehicle decelerated at a rate of 16 km/h without illuminating its taillights within 5 s (from 70 to 54 km/h), maintained a speed of 54 km/h for 20 s, and then accelerated to a speed of 120 km/h until it disappeared from the screen. (b) Normal braking. The lead vehicle decelerated at a rate of 16 km/h using the brakes (illuminating the taillights) within 3 s (from 70 to 54 km/h), maintained a speed of 54 km/h for 20 s, and then accelerated to 120 km/h until it disappeared from the screen. (c) Emergency braking. The lead vehicle decelerated at a rate of 40 km/h braking heavily within 3 s (from 70 to 30 km/h), maintained a speed of 30 km/h for 20 s, and then accelerated to 120 km/h until it disappeared from the screen. (d) Stopping ahead of an intersection. The lead vehicle decelerated to 0 km/h by braking heavily within 5 s because the signal light turned red, stopped in front of the intersection for 15 s, first accelerated to 40 km/hr for 5 s, and then accelerated to 120 km/h until it disappeared from the screen. 50% all sudden events occurred on straight roads, while 50% occurred on curved roads. The sudden events happened in random order with no similar events appearing in succession. Divided attention task. The divided attention (DA) task is used for measuring drivers’ capabilities in detecting both roadsides unexpected events. As shown in Fig. 1, the sign ( ) appears on both sides of the screen. In this study, the signs changed randomly every 60 s into red triangles (i.e., or ), and remained on display for 5 s. When a change in either signs’

Fig. 1. Example of road scene from the driver’s viewpoint. This figure illustrates the road scene including the FCW information (e.g., zone one) and the current vehicle speed 54 in green on the HUD, and the surrounding environment. Notably, two red diamond symbols, designed for the divided attention (DA) task, appeared on both right and left sides (see divided attention task for details). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

6

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

appearances was detected, participants were asked to turn the signal handle in the same direction as the sign that changed. The participants were furthermore instructed to respond as accurately and quickly as possible. 3.4. Scenario descriptions The road driving scenarios were developed using STI Scenario Definition Language (SDL) V.8.0 and were divided into lowand high-load conditions. Factors considered for designing the high- and low-driving load environments included: number of lanes, lane width, number of intersections, density of roadside buildings, and density of approaching vehicles (Liu & Wen, 2004). All participants were instructed to finish two separate driving experiments with high- and low-driving workloads. Each driving experiment lasted for approximately 60 km and was divided into three equal road sections: (1) the first 20 km section (driving without the warning system). This section constituted the baseline. In addition, for participants using the SQC FCW systems, the time headway data of the road section between 5 km and 15 km were collected to build up the central, and the lower boundary lines of the charts; (2) the second 20 km section (driving with one of the FCW systems); (3) the last 20 km section (driving without the warning system). The latter section was designed to investigate the possible post effects on the drivers’ car following their behaviors after using the FCW systems. In addition to the four types of lead vehicle decelerations (the sudden events mentioned above), this study manipulated the other four types of driver’s aggressiveness-induced events based on daily driving experiences: (1) (2) (3) (4)

The lead vehicle proceeds more slowly than is acceptable for ensuring a smooth traffic flow; A bicyclist is crossing the intersection, forcing the driver to slow to a stop; A vehicle is driving close to a participant’s rear bumper and is honking repeatedly; A pedestrian walks slowly across the street, forcing participants to slow in order to avoid a collision. In each road section, they appeared at random every 1.33 km with no similar events occurring in succession.

For the entire drive, participants were asked to maintain a speed of 90 km/h and to follow the lead vehicle at a safe distance. During each 20-km road section, four types of sudden events appeared 14 times, the driver’s aggressiveness-induced events 8 times, and DA tasks 14 times. 3.5. Experimental design This study conducted a 2 (driving aggressiveness: high vs. low; between subjects)  2 (driving workload: high vs. low; within-subjects)  3 (the FCW system: five-stage vs. X-bar vs. X-bar and EWMA control charts; between subjects) mixedfactorial experiment. Counter balance was considered to prevent possible order effects, such as learning or fatigue. Each aggressive group included 48 participants and was divided into three sub-groups with a five-stage, an X-bar control chart, or an X-bar + EWMA control charts FCW system respectively. Data collected for this study included both objective (i.e., brake response time, minimum time to collision, minimum range between the driver and the lead vehicle, Maximum longitudinal deceleration due to brakes, DA response time) and subjective measures (i.e., NASA-TLX workload rating, trust in the FCW system, the preference for the FCW system). Data were analyzed by the analysis of variance (ANOVA) and the post hoc analyses were performed using the Tukey method. The level of significance for all analyses was set at a < 0.05. 3.6. Procedures Participants with high- (UKDAS > 84) or low- (UKDAS < 63) driving anger scales were invited to take part in the driving simulation experiment. Before the experiment, participants had to meet the normal requirements for vision and hearing. They received instructions on the operation of the simulator, the experimental procedure, and the tasks at hand. After participants declared to have fully understood the study, they were invited to signa consent form. Participants were given approximately 10-min to practice driving, to familiarize themselves with the simulator controls and to experience lead vehicle decelerations, all of which formed part of the experimental trials. They were instructed to drive as they normally would and to follow the lead vehicle at a safe distance. Participants drove on the first road section without FCW system in order to set up the SQC FCW system for each of the SQC FCW participants. At the end of each road section, NASA-TLX questionnaires were scaled to evaluate participants’ subjective workload. Subjective preferences and trust regarding the FCW system were scaled after drivers reached the end of the road section with the FCW. All experiments lasted for approximately 60 min and each participant completed one of the two workload roads first. To avoid a possible fatigue effect caused by one load road experiment, participants were asked to conduct the left load road experiment approximately one week later, with the same procedures as for the first experiment. Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

7

4. Results 4.1. The objective measures 4.1.1. Braking response time Braking-response time is defined as the total time recorded from sudden events causing drivers to step on the brakes. Results showed that driving workload [F(1,90) = 162.682, p = .000], FCW system [F(2,90) = 10.167, p = .000] and driving aggressiveness [F(1,90) = 4.124, p = .045] had a significant effect on braking response time after occurrence of sudden events, while other interactions were not significant. The results showed that the braking response times performed by participants in a high workload road conditions (1.867 s) were longer than those in alow workload road (1.676 s). As for the braking response time during sudden events, drivers’ response times with a five-stage FCW (1.833 s) proved longer than those in FCWs using the X-bar + EWMA (1.718 s), as well as those using the X-bar (1.732 s). The difference in braking response time between using the X-bar + EWMA and the X-bar was not significant. In addition, the braking response time for low-aggressive drivers (1.738 s) was shorter than for high-aggressive drivers (1.784 s). 4.1.2. Minimum time to collision (TTC) The minimum TTC with the lead vehicle is recorded through a STI driving simulator and represents the urgency of braking. Results indicated that driving workload [F(1,90) = 28.851, p = .000], FCW system [F(2,90) = 3.756, p = .027] and driving aggressiveness [F(1,90) = 19.479, p = .000] had significant effects on minimum TTC with the lead vehicle. Other interactions did not prove to be significant. Participants’ minimum TTC with the lead vehicle in low load driving road sections were comparatively long (3.152 s), whereas in complex driving road sections, minimum TTC decreased significantly (2.927 s). Drivers using the X-bar + EWMA as well as X-bar performed longer minimum TTC with the lead vehicle (X-bar: 3.083 s; X-bar + EWMA: 3.129 s) than drivers with a five-stage FCW (2.908 s). Furthermore, low aggressive drivers achieved favorable minimum TTC with the lead vehicle during sudden events (3.193 s), whereas high aggressive drivers significantly decreased minimum TTC with the lead vehicle (2.886 s). 4.1.3. Response time of the DA task Response time of the DA tasks, collected by the STI driving simulator, is defined from the time the signs changed into red triangles to the time the indicator was used. Results revealed that the driving load condition [F(1,90) = 94.128, p = .000) and the FCW system (F(2,90) = 3.500, p = .034) had a significant effect on response time of DA tasks. Other effects of interactions were found statistically insignificant. In a low driving load road, drivers took a short (1.366 s) response time in detecting the roadside symbol changes, whereas in a high driving load road, the response times increased significantly (1.562 s). In addition, drivers using a five-stage FCW took longer to respond (1.516 s) in the DA tasks than drivers using the X-bar (1.412 s). Drivers using the X-bar + EWMA showed a response time of 1.464 s in performing the DA task, not resulting insignificant differences with drivers using the five-stage or the X-bar FCWs. 4.1.4. Advantages of the FCW systems The study designed three road sections for each driving load road. By comparing any two of the three sections of the drivers’ driving performance while approaching those sudden events, we can obtain: (1) the FCWs effect: the changing rate of the drivers’ car following performance between the road section using the FCWs and the baseline road; (2) the learning effect of using the FCWs: the changing rate of the drivers’ car following performance between the road section after using the FCWs and the baseline road; (3) the decline effect after using the FCWs: the decline rate of the drivers’ car following performance between the road section after using the FCWs and the road section using the FCWs. In Table 2, overall, results showed that (1) the drivers using the FCWs in the FCW effect [F(2,93) = 20.698, p = .000], (2) the decline rate [F(2,93) = 5.631, p = .005], and (3) the learning effect [F(2,93) = 17.805, p = .000] all proved to be significant in measuring the drivers’ braking response times. Driving with SQC FCWs enhanced the drivers’ braking response times significantly during sudden events (X-bar + EWMA: 15.3%, X-bar:15.0%) compared to those drivers using the five-stage FCW system (11.0%). The drivers using the X-bar + EWMA FCWs showed a significantly larger decline rate (12.3%) in braking response times than those using the X-bar FCW (10.4%), while the five-stage FCWs resulted in the lowest decline rate (8.8%). The drivers’ experiences of using the X-bar FCW produced the largest learning rate of braking response times (6.3%) compared to those using the X-bar + EWMA FCW system (5.0%). The user experiences of five-stage FCWs led to a mere modest learning effect of braking response time (3.3%) while sudden events occurred.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

8

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

Table 2 Effects of using the three FCWs on the drivers’ braking RTs during sudden events. High driving load road

Five-stage X-bar X-bar + EWMA

Low driving load road

FCW1 effect

Learning effect

Decline effect

FCW effect

Learning effect

Decline effect

9.56%2 14.70% 14.80%

2.42% 5.57% 3.17%

8.14% 10.81% 13.83%

12.36% 15.32% 15.80%

4.19% 6.94% 6.86%

9.52% 10.02% 10.84%

1 The FCW effect = [(drivers’ braking response time in road section 2)-(drivers’ braking response time in road section 1)/(drivers’ braking response time in road section 1)]  100%; the learning effect = [(drivers’ braking response time in road section 3)-(drivers’ braking response time in road section 1)/(drivers’ braking response time in road section 1)]  100%; the decline effect = [(drivers’ braking response time in road section 3)-(drivers’ braking response time in road section 2)/(drivers’ braking response time in road section 2)]  100%. 2 ‘‘” in the FCW and the learning effects indicated the positive improvement of quickening the brake response times; thus, the smaller the negative value the larger the improvement, while, ‘‘+” in the decline effect indicated the decreasing rate, and the larger the positive value the larger the declination.

Table 3 shows that drivers using the three FCWs caused their minimum time to collision with the lead vehicle to significantly differ in measuring the FCW effort [F(2,93) = 9.894, p = .000], the decline rate [F(2,93) = 7.573, p = .001], and the learning effect [(F(2,93) = 7.804, p = .000). The drivers using the X-bar and the X-bar + EWMA FCWs produced significant improvement rates in minimum time to collision (17.5% and 17.9% respectively), while using the five-stage FCW led to the lowest rate (12.9%). Using the X-bar + EWMA FCW caused a significant larger decline rate in the drivers’ minimum times to collision (9.9%) than those using the X-bar (6.8%) and five-stage FCWs (6.8%). The drivers’ experiences with the X-bar FCW led to a significant larger learning effect rate (9.4%) than those using the X-bar + EWMA (6.0%) and the five-stage (5.1%) FCWs. In addition to revealing the inferential statistical analysis results mentioned above, this study also recorded the mean headway (in seconds) to the lead vehicle with the sudden events occurring in 5-s intervals for the two groups of angry drivers in both conditions using the three FCWs on the respective three road sections (Figs. 2 and 3). As can be seen in Fig. 3, the drivers with the FCWs assistance maintained longer mean time headways to the lead vehicle (the FCW effect). Moreover, after using the systems, the drivers’ car following behaviors did make some changes favorable to safe driving (the learning effect). For the three FCWs, driving conditions with one of the two SQC FCWs more efficiently alerted the drivers to maintain longer headways than did those of the five-stage FCW system (as shown in Fig. 2). 4.2. The subjective measures 4.2.1. The NASA-TLX workload ratings The main effect of driving mental workload due to driving load environment caused significantly different mental workloads [F (1,90) = 293.224; p = .000] and no significant interaction was found. The drivers felt a lower workload in the lowload driving road experiment (42.912) compared to those in the complex driving road one (55.296). However, the NASATLX questionnaire scores did not vary significantly regarding the variables of the FCW systems [F(2,90) = 1.209; p = .303] and driving aggressiveness [F(1,90) = 1.280; p = .261]. We also conducted comparisons of the three road sections in each driving load environment. The three road sections led to significantly different mental workloads [F(2,190) = 352.178; p = .000], and the NASA-TLX questionnaire ratings in the FCW road stretch (49.104) were significantly lower than those in baseline (63.607) and post-learning (59.073) road stretches. 4.2.2. The trust and preference in the FCWs The driving load condition [F(1,90) = 50.965, p = .000], FCW system [F(2, 90) = 54.568, p = .000], and driving aggressiveness [F(1,90) = 7.287, p = .006] had significant impacts on subjective trust in using the FCW systems.

Table 3 Effects of using the three FCWs on the drivers’ minimum TTCs during sudden events. High driving load road 1

Five-stage X-bar X-bar + EWMA

Low driving load road

FCW effect

Learning effect

Decline rate

FCW effect

Learning effect

Decline rate

12.50%2 17.10% 17.50%

5.75% 8.52% 5.65%

-5.89% -7.26% -9.93%

13.32% 17.80% 18.30%

4.61% 10.20% 6.43%

7.69% 6.39% 9.79%

1 The FCW effect = [(drivers’ minimum TTC in road section 2)-(drivers’ minimum TTC in road section 1)/(drivers’ minimum TTC in road section 1)]  100%; the learning effect = [(drivers’ minimum TTC in road section 3)-(drivers’ minimum TTC in road section 1)/(drivers’ minimum TTC in road section 1)]  100%; the decline effect = [(drivers’ minimum TTC in road section 3)-(drivers’ minimum TTC in road section 2)/(drivers’ minimum TTC in road section 2)]  100%. 2 ‘‘+” in the FCW and the learning effects indicated the improvement of lengthening the minimum TTC, and thus the larger the positive value, the longer the minimum TTC improvement, ‘‘” in the decline effect indicated the time to collision shortening rate, and thus the smaller the negative value the more the declination of the minimum TTC.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

9

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

Fig. 2. Trend analyses of mean headway to the lead vehicle when (a) the highly aggressive and (b) the less aggressive drivers used the three FCWs during the sudden events occurring in 5-s intervals.

Time headway (sec)

Time headway (sec) 1.6

1.6

1.4

1.4 Baseline road

1.2

Baseline road

1.2

FCW road Post learning road

1.0

FCW road Post learning road

1.0 0.8

0.8 0.6

Time (sec)

0

1

2

3

4

(a)

5

6

0.6

Time (sec)

0

1

2

3

4

5

6

(b)

Fig. 3. Trend analyses of mean headway to the lead vehicle when (a) the highly aggressive and (b) the less aggressive drivers drove on the three section roads during the sudden events occurred in 5-s intervals.

The drivers’ trust for using the FCW systems in the low-load driving road were comparatively low (77.948), whereas trust in high-load driving road conditions increased significantly (81.906). Drivers’ trust using the X-bar FCW (85.375) was greater than that of using the X-bar + EWMA (74.297) and five-stage FCWs (77.109). The low-aggressive drivers’ trust on those FCWs was higher (82.969) than the relevant trust of the highaggressive drivers (76.885). Two 2-way interactions of driving load condition  FCWs [F(2,90) = 4.932, p = .009] and FCWs  driving aggressiveness [F(2,90) = 11.425, p = .000] showed significant effects on subjective trust in using the FCW system. In the high-load driving road, the drivers using the FCWs showed significantly different subjective trust [F(2,93) = 46.091; p = .000]. The drivers felt an increased trust in using X-bar FCW (91.500) than in using the X-bar + EWMA (75.313) and the five-stage FCWs (78.906). While the participants drove in the low-load road environment, different FCWs still caused significantly different trust [F(2,93) = 21.015; p = .000]. The drivers felt an increased trust in using the X-bar FCW (85.25) than in using X-bar + EWMA (73.281) and the five-stage FCWs (75.313). The high-aggressive drivers felt a higher subjective trust in using the X-bar FCW (86.906) than in using X-bar + EWMA (73.594) and the five-stage FCWs (70.156). Similar results were found for the low-aggressive drivers: the highest subjective trusts in using X-bar FCW (89.844), while the five-stage FCWs (84.063) came second and the X-bar + EWMA FCW last (75.00) in the drivers’ trust ratings. For the preference ratings of using the three FCWs [(F(2,90) = 59.788, p = .000], drivers significantly preferred to use the X-bar control chart FCW (88.203) compared to those using the X-bar & EWMA control chart FCW (70.703) and the five-stage FCWs (71.875), with the latter two control charts showing no significant differences. 5. Discussion SQC charts in the manufacturing process control have been widely used in industries since the 1920s. They have proven to be valuable in monitoring the process stability and in controlling product quality within acceptable limits. However, using

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

10

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

the SQC techniques in monitoring human behaviors and performance were rare. Continuous and monitoring in time based on different product characteristics are the main features of the SQC, and the drivers’ car following behaviors are very suitable for the SQC applications. In the car following procedure, different personalities as well as various perceptions of speed and distance made that each driver had, to some degree, different car following behaviors. Without consideration of these differences, the current FCW systems all set the warning zone based on absolute criteria, e.g., less than 1.6 TTC, resulting drivers’ annoying, false alarm, or even distrust in the systems. This, we believe, illustrates the potential value of adopting SQC charts in the design of FCW systems. Driving with the in-vehicle FCW assistance made the drivers aware of their car following behaviors, and thus produced a longer time headway than was the case on the baseline road for both low- and high-aggressive driver groups (Fig. 3). In addition, the drivers driving with the FCWs felt the least psychological workload compared to those not using the respective systems. These findings indicate the usefulness of equipping vehicles with the FCW systems in order to avoid rear-end collisions (Ben-Yaacov et al., 2002; Lee, Ries, McGehee, Brown, & Perel, 2000; Lee et al., 2002; Maltz & Shinar, 2007; Mohebbi et al., 2009; Scott & Gray, 2008). As such, even drivers with different tempers can be made to follow the in-vehicle warning information to some extent. Therefore, the issue at hand is not whether the vehicle should be equipped with the FCWs, but how to provide drivers with an efficient in-vehicle warning system, with the key factor depending mainly on the interface design. In comparing the three FCWs, results revealed the substantial benefits of adopting the SQC design concepts. The X-bar + EWMA FCW system, in particular, assisted the drivers in improving the braking response time by 15.3% from occurring sudden events (X-bar: 15.0%; five-stage: 11.0%), while lengthening the minimum time to collision by 17.9% (X-bar: 17.5%; five-stage: 12.9%) (Tables 2 and 3). Although the differences between the two FCWs with the X-bar vs. the X-bar + EWMA were not significant, better behavioral values in drivers’ car following measures by using the X-bar + the EWMA FCW imply that, in contrast to the X-bar being sensitive in large shifts, the sensitivity of EWMA in monitoring small shifts (e.g., the minimum time headway) might have the FCW system alert the drivers more frequently/probably as is the case with the FCW with the X-bar design concept, and therefore makes drivers more aware of the car following conditions. Yet, drivers felt the same psychological load feelings in using the two SQC FCWs (X-bar: 47.29; X-bar + EWMA: 48.82). However, the drivers trusted and preferred using the X-bar FCW system (trust: 85.375; preference: 88.203), in contrast to using the X-bar + EWMA FCW system (trust: 74.297; preference: 70.703). The X-bar FCW system received the highest trust ratings from both angry driver groups. In revealing the DA task evaluation, using the X-bar FCW system made the drivers’ have faster responding times (1.412 s) than those using the other two FCWs (the X-bar + EWMA: 1.464 s; the five-stage: 1.516 s). Findings show that, compared to the X-bar FCW system, frequently presenting the warning information by the X-bar + EWMA FCW system does not have a negative effect on the drivers’ mental load in this study. Nevertheless, in the long term, it can be presumed that this system could possibly annoy drivers as well as raise their arousal levels. This might consequently cause a negative impact on their driving performance, e.g., the mental load. The fact that drivers slowed their response times in detecting the road side DA tasks appeared to compensate, to some degree, for performing better in car following behaviors while using the X-bar + EWMA FCW system. The X-bar FCWs proved very effective in detecting a large outlier, the abnormal behavior, and would gradually warn the drivers on the various safeties to danger stages. The X-bar + EWMA effectively detected small shifts and their accumulation with different weights, the most recent samples are weighted most highly while the most distant samples contribute very little (Shehab & Schlegel, 2000). Therefore, in addition to presenting the warning information frequently, the X-bar + EWMA system might not gradually warn the drivers from zone one to zone five. Consequently, the respective drivers were easily startled when the serious warning information suddenly appeared. These characteristics made that the X-bar + EWMA system might prove to be problematic to drivers, since the car following condition as perceived by the drivers would not be consistent with the degree of danger provided by the FCW warning, even if valid warning information were given. In other words, the main question remaining was why the drivers trusted and favored the X-bar FCW system the most. Indeed, users’ trust has been recognized as an important determinant of system performance and it is one of the main predictors of efficiency in automation use (Lee & Moray, 1992; Muir, 1987). 6. Conclusion In order to thoroughly investigate the effects of the three FCW systems, this study divided three test road sections: the baseline section of not using the FCWs, the section of using the FCWs, and the section after using the FCWs, for each driving load road sequentially. The road section using the FCWs caused the longest time headway, the section after using the FCWs the second shortest, and the baseline section the shortest headway (Fig. 2). Clearly, the FCWs were helpful and the experiences of using the FCWs improved the drivers’ behaviors towards much safer driving conditions. As compared to the baseline and under the two respective load conditions, the drivers with the X-bar FCW system produced the greatest improvements in car following performance, e.g., the braking response times and the minimum TTC, when encountering the sudden events manipulated by the leading vehicle, i.e., the learning effect. In addition, using the

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

11

X-bar FCW mitigated the drivers’ performance decline ratios after using the FCWs, i.e., the decline effect (Tables 2 and 3). These findings indicated that, by using the X-bar FCW, drivers were more likely to comply with the warning direction and thus changed their driving behaviors accordingly. In summary, vehicles equipped with the FCW system did improve drivers’ awareness and responses in their car following behaviors, irrespective of whether the driver’s emotion was angry. The warning is therefore considered to be effective. However, when each driver’s car following habit was taken into consideration, the FCW system adopting the SQC concepts showed better warning effects on drivers’ responses with respect to braking, since they maintained a longer time headway and minimum TTC to the leading vehicle compared to those using the five-stage FCW system. In addition, drivers using the X-bar FCWs responded to the unexpected roadside events faster than those who used the five-stage FCW system. The most important advantage seems to be that drivers’ trust and preference of the X-bar FCW system makes the system itself serviceable, and drivers become more willing to comply with the system without increasing their mental load. We therefore believe that applications of SQC to monitor human performance and behavior with the purpose of providing warning information deserves more follow-up research in the field of interface design. Acknowledgement This study was financially supported by the Ministry of Science and Technology, Taiwan (contract no. MOST-101-2221-E224-001-MY2). The authors deeply express their gratitude to all participants for their valuable efforts in completing this research. References Ben-Yaacov, A., Maltz, M., & Shinar, D. (2002). Effects of an in-vehicle collision avoidance warning system on short- and long-term driving performance. Human Factors, 44(2), 335–342. Campbell, J. L., Richard, C. M., Brown, J. L., & McCallum, M. (2007). Crash warning system interfaces: Human factors insights and lessons learned, DOT HS 810 697. National Highway Transportation Safety Administration. Cannon, J., Krokhmal, P. A., Chen, Y., & Murphey, R. (2012). Detection of temporal changes in psychophysiological data using statistical process control methods. Computer Methods and Programs in Biomedicine, 107(3), 367–381. Deffenbacher, J. L., Lynch, R. S., Filetti, L. B., Dahlen, E. R., & Oetting, E. R. (2003). Anger, aggression, risky behavior, and crash-related outcomes in three groups of drivers. Behaviour Research and Therapy, 41, 333–349. Dingus, T. A., McGehee, D. V., Manakkal, N., Jahns, S. K., Carney, C., & Hankey, J. M. (1997). Human factors field evaluation of automotive headway maintenance/collision warning devices. Human factors, 39(2), 216–229. Ellison-Potter, P., Bell, P., & Deffenbacher, J. (2001). The effects of trait driving anger, anonymity, and aggressive stimuli on aggressive driving behavior. Journal of Applied Social Psychology, 31, 431–443. Fernandes, R., Hatfield, J., & Soames Job, R. F. (2010). A systematic investigation of the differential predictors for speeding, drink-driving, driving while fatigued, and not wearing a seat belt, among young drivers. Transportation Research Part F, 13(3), 179–196. General Motors Corporation & Delphi-Delco Electronic Systems (2002). Automotive collision avoidance system field operation test, warning cue implementation summary report (DOT HS 809 462). Washington, DC: National Highway Traffic Safety Administration. Hayati, F., Maghsoodloo, S., Devivo, M., & Carnahan, B. (2006). Control chart for monitoring occupational asthma. Journal of Safety Research, 37(1), 17–26. Heino, A., van der Molen, H. H., & Wilde, G. J. S. (1996). Differences in risk experience between sensation avoiders and sensation seekers. Personality and Individual Difference, 20(1), 71–79. Hwang, S.-L., Lin, J.-T., Liang, G.-F., Yau, Y.-J., Yenn, T.-C., & Hsu, C.-C. (2008). Application control chart concepts of designing a pre-alarm system in the nuclear power plant control room. Nuclear Engineering and Design, 238(12), 3522–3527. Iversen, H., & Rundmo, T. (2002). Personality, risky driving and accident involvement among Norwegian drivers. Personality and Individual Differences, 33(8), 1251–1263. Jamson, A., Lai, F., & Carsten, O. (2008). Potential benefits of an adaptive forward collision warning system. Transportation Research Part C, 16(4), 471–484. Lajunen, T., Parker, D., & Stradling, S. G. (1998). Dimensions of driver anger, aggressive and highway code violations and their mediation by safety orientation in UK drivers. Transportation Research Part F, 1(2), 107–121. Lajunen, T., Parker, D., & Summala, H. (1999). Does traffic congestion increase driver aggression? Transportation Research Part F, 2, 225–236. Lee, J. D., McGehee, D. V., Brown, T. L., & Reyes, M. L. (2002). Collision warning timing, driver distraction, and driver response to imminent rear-end collisions in a high-fidelity driving simulator. Human Factors, 44(2), 314–334. Lee, J. D., Mcgehee, D. A., Dingus, T. A., & Wilson, T. B. (1998). Collision avoidance behavior of un alerted drivers using a front-to-rear-end collision warning display on the Iowa driving simulator. Transportation Research Record, 1573, 1–7. Lee, J. D., & Moray, N. (1992). Trust, control strategies and allocation of function in human-machine systems. Ergonomics, 35, 1243–1270. Lee, J. D., Ries, M. L., McGehee, D. V., Brown, T. L., & Perel, M. (2000). Can collision warning systems mitigate distraction due to in-vehicle devices? Retrieved from . Liu, Y.-C., & Wen, M.-H. (2004). Comparison of head-up display (HUD) vs. head-down display (HDD): Driving performance of commercial vehicle operators in Taiwan. International Journal of Human-Computer Studies, 61(5), 679–697. Maguad, B. A. (2005). Monitoring teachers’ performance using statistical process control charts. International Journal of Quality and Productivity Management, 5(1), III-1–III-10. Maltz, M., & Shinar, D. (2007). Imperfect in-vehicle collision avoidance warning systems can aid distracted drivers. Transportation Research Part F, 10, 345–357. Mesken, J., Hagenzieker, M., Rothengatter, T., & Dewaard, D. (2007). Frequency, determinants, and consequences of different drivers’ emotions: An on-theroad study using self-reports, (observed) behaviour, and physiology. Transportation Research Part F, 10(6), 458–475. Mohebbi, R., Gray, R., & Tan, H. Z. (2009). Driver reaction time to tactile and auditory rear-end collision warnings while talking on a cell phone. Human Factors, 51(1), 102–110. Montgomery, D. C. (2005). Introduction to statistical quality control (5th ed.). New Jersey: John Wiley. Moore, M., & Dahlen, E. R. (2008). Forgiveness and consideration of future consequences in aggressive driving. Accident Analysis & Prevention, 40(5), 1661–1666. Muir, B. M. (1987). Trust between humans and machines, and the design of decision aids. International Journal of Man-Machine Studies, 27, 527–549. National Highway Traffic Safety Administration (2012). Traffic safety facts 2012: A compilation of motor vehicle crash data from the fatality analysis reporting system and the general estimates system (DOT HS 812 032). Washington, DC: U.S. Department of Transportation.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010

12

Y.-C. Liu, C.H. Ho / Transportation Research Part F xxx (2017) xxx–xxx

Ong, K. H., Harvey, C., Shehab, R., Dechert, J., & Darisipudi, A. (2004). The effects of three statistical control charts on task performance. Production Planning & Control, 15(3), 313–323. Ranney, T. A. (1999). Psychological factors that influence car-following and car-following model development. Transportation Research Part F, 2, 213–219. Scott, J. J., & Gray, R. (2008). Comparison of visual, auditory and tactile warnings for rear-end collision prevention in simulated driving. Human Factors, 50, 264–275. Shehab, R. L., & Schlegel, R. E. (2000). Applying quality control charts to the analysis of single-subject data sequences. Human Factors, 42(4), 604–616. Ulleberg, P. (2002). Personality subtypes of young drivers. Relationship to risk-taking preferences, accident involvement, and response to a traffic safety campaign. Transportation Research Part F, 4(4), 279–297. Wierwille, W. W., Lee, S. E., DeHart, M. C., & Perel, M. (2006). Test road experiment on imminent warning rear lighting and signaling. Human Factors, 48(3), 615–626.

Please cite this article in press as: Liu, Y.-C., & Ho, C. H. A comparison of car following behaviors: Effectiveness of applying statistical quality control charts to design in-vehicle forward collision warning systems. Transportation Research Part F (2017), https://doi.org/10.1016/j. trf.2017.09.010