Heavy-Truck Drivers’ Following Behavior With Intervention of an Integrated,In-Vehicle Crash Warning System

Introduction

The safety of heavy trucks (large trucks with a gross vehicle weight rating of more than 10,000 pounds) has been of great societal concern because of the rapid growth of the truck industry and the associated high fatal-crash rates. According to the Federal Motor Carrier Safety Administration (FMCSA; 2007), there was a 68% increase in the number of registered heavy trucks from the year 1975 to 2007, equaling an increase of 3.7 million heavy trucks throughout the United States. In 2007, heavy trucks represented approximately 4% of all registered vehicles but accounted for about 8% of all vehicles involved in fatal crashes (National Highway Traffic Safety Administration [NHTSA], 2008). There were more than 310,000 crashes with heavy trucks, which led to 101,000 injuries and 4,808 fatalities in the year 2007 in the United States (FMCSA, 2007). The Insurance Institute for Highway Safety (2006) reported that heavy trucks had a much higher rate of fatal-crash involvement (2.2 per 100 million miles) than did passenger cars (1.3 per 100 million miles).

Rear-end crashes occur when a vehicle’s front bumper strikes the rear of a leading vehicle. It is was one of the most common truck crashes in the United States (Abdel-Aty & Abdelwahab, 2004; Blower & Campbell, 1999; Chang & Mannering, 1999; Zhu & Srinivasan, 2011). It is estimated that approximately one fifth of all heavy-truck crashes that occurred during the period from 1994 to 1999 in the United States were rear-end crashes, among which there were 50% more truck-striking rear-end crashes than there were truck-struck crashes (FMCSA, 2001). Najm and Smith (2007) also reported that about 46,000 rear-end crashes occur annually in which the heavy truck is the striking vehicle. Golob and Recker (1987) found that rear-end crashes that involve heavy trucks were more likely to result in severe injuries and fatalities than were the other multivehicle crashes with heavy trucks.

In studies investigating the causation of rear-end crashes, researchers have determined that driver errors while following other vehicles are among the leading contributing factors. For example, Dingus et al. (1997) indicated that rear-end crashes usually involved drivers who fail to attend to the leading vehicle, follow too closely, or drive when impaired. Kostyniuk and Eby (1998) conducted a focus group study and interviewed 26 participants who had recently experienced rear-end crashes as drivers of striking vehicles. The results of their study showed that drivers believe that their inattention to the leading vehicle and incorrect assumptions about traffic movement are among the dominant reasons for a rear-end crash.

The large-truck crash causation study (LTCCS; Starnes, 2006) conducted by the U.S. Department of Transportation, FMCSA, and NHTSA involved a nationally representative sample of large-truck (heavy-truck) fatal and injury crashes from 2001 to 2003 at 24 sites in 17 states in the United States. The LTCCS determined that truck driver errors (i.e., following too closely, driving too fast, falling asleep, paying improper attention, and performing poorly) were primary causes for approximately 88% of all the crashes when it was determined that a heavy truck was at fault.

The likelihood of rear-end crashes can generally be minimized by maintaining a safe distance from a leading vehicle while driving, which will, therefore, provide drivers with more time to respond to potential traffic hazards. A commonly used measure of the distance between vehicles traveling in the same direction is time headway, which is defined as the value of the distance between vehicles divided by current vehicle speed (Ben-Yaacov, Maltz, & Shinar, 2002; Ervin et al., 2005; Fuller, 1981; Shinar & Schechtman, 2002). It is generally recommended that a minimum of 2 s time headway be maintained on a freeway for passenger car drivers (National Safety Council, 1992).

Maintaining adequate time headway is important because it allows drivers to initiate braking in response to the decelerating leading vehicle. Longer time headway values can be assumed to translate to fewer rear-end crashes. This assumption is especially true for heavy trucks, since trucks actually take about 20% to 40% farther to stop relative to passenger-vehicle stopping distances (NHTSA, 1987). However, drivers commonly make judgment errors on safe time headway and distance from a leading vehicle (Hoffmann & Mortimer, 1994; McLeod & Ross, 1983). Taieb-Maimon and Shinar (2001) found that even alert drivers tend to overestimate and misjudge their time headway from leading vehicles. In their study, the average actual minimum time headway was 0.66 s, even though drivers estimated this headway to be 2.1 s on average.

Designers of advanced in-vehicle crash warning systems aim to reduce rear-end crash rates by alerting an inattentive driver to an upcoming forward hazard and by aiding the driver in making correct judgments of the distance from leading vehicles. Studies that evaluated the impact of in-vehicle crash warning systems on passenger cars have shown beneficial effects on drivers’ headway maintenance. Lee, McGehee, Brown, and Reyes (2002) found that a rear-end collision avoidance system benefited both distracted and undistracted drivers by reducing the time required for drivers to release the accelerator in a driving simulator study. Ervin et al. (2006) conducted a naturalistic driving study to evaluate the effects of an automotive collision avoidance system (ACAS), which included a forward collision warning system and an adaptive cruise control system, on drivers’ safety behavior. A total of 96 passenger car drivers participated in the study and drove research vehicles equipped with the ACAS. The forward collision warnings provided drivers with both visual cautionary and auditory imminent warnings when they were following very closely and approaching too rapidly. The results showed a reduction in short headway following (<1 s) when drivers were on the freeway.

Ben-Yaacov et al. (2002) found that passenger car drivers were able to maintain a longer and safer distance from leading cars after a short exposure to an in-vehicle collision avoidance warning system. An audible warning tone was generated when drivers’ headway time from the leading vehicle was equal to 1 s or less in their study. Dingus et al. (1997) also reported similar findings. Shinar and Schechtman (2002) found that the effect of headway feedback in a passenger car (both a red light and an audible warning) was to reduce the time spent in the short headway (≤0.8 s) by about 25%. Although heavy-truck drivers are generally more professional and experienced and have many more hours on road than do passenger car drivers, driver errors are still the major contributing factor to heavy-truck crashes (Starnes, 2006). Therefore, it is expected that an in-vehicle, forward-crash avoidance warning system would help to alert heavy-truck drivers to upcoming events, thereby enabling safer driving. However, very few studies have been conducted on this issue.

In addition to driver-related factors, environmental factors have also been reported to be associated with heavy-truck crashes. For example, roadway type has been reported to have a great impact on crash rates of heavy trucks; crash rates were lower on freeways than on other roadways (Starnes, 2006). Slippery roadway surfaces (wet or icy) led to higher heavy-truck crash rates than did dry roadway surfaces (Golob & Recker, 1987). Ramírez, Izquierdoa, Fernández, and Méndez (2009) reported that crashes involving heavy trucks increased with a general increase in on-road traffic.

The purpose of this article is to investigate the effects of an integrated, in-vehicle crash warning system on the time headway maintenance and response to the forward traffic conflict behavior of heavy-truck drivers with the use of naturalistic driving data. The integrated warning system evaluated in this study included both longitudinal and lateral driving behavior supports; however, only longitudinal results are discussed in this article. The hypothesis is that the crash warning system will assist heavy-truck drivers in driving more safely: (a) Drivers are expected to maintain longer time headway from leading vehicles, and (b) drivers are expected to have a shorter reaction time in responding to a potential traffic conflict in front of the truck. These differences are also expected to be larger in more challenging environmental conditions, such as adverse weather.

Method

The data used in this study were collected through the naturalistic driving study of the Integrated Vehicle-Based Safety System (IVBSS) program conducted by a research team led by the University of Michigan Transportation Research Institute (UMTRI). The program developed integrated, advanced technologies intended to help drivers avoid or mitigate crashes. The integrated system warns drivers when they are about to leave the roadway, are in danger of colliding with another vehicle while attempting a lane change, or are at risk of colliding with the vehicle in front of them. The integrated system addressed crash types that account for 67% of all motor vehicle crashes in the United States. The integrated crash warning system gathered information via inertial, video, and radar sensors plus a global positioning system module. The system provided only warnings and did not include automatic application of throttle release, retarder(s), or brakes.

The IVBSS program consisted of two platforms: a heavy-truck fleet and a light-vehicle fleet. For this article, the data from the heavy-truck fleet were used. We used ten 2008 International ProStar 8600-series tractors as research vehicles. These vehicles were built to specification for, and purchased by, Con-way Freight, the driving test operator. The tractors were equipped with the integrated safety system, which includes a forward-collision warning system, a lane-change or merge warning, and a lateral-drift warning system. Each truck was instrumented to capture information regarding the driving environment, driver activity, system behavior, and vehicle kinematics, with a data collection frequency of 10 to 50 Hz. These vehicles did not have either conventional or adaptive cruise control.

A driver-vehicle interface (DVI) was developed for the integrated system, which included a dash-mounted visual display device and two blind spot indicators, one on each side of the cabin, as shown in Figure 1. Drivers used the center display to input the trip information (such as trailer length) and to adjust the settings of the warning systems, including the volume of the auditory warnings and the brightness of the display. The center display continuously presented the availability of the lane tracking for the lateral position, time headway information, and visual warnings when necessary. The two blind spot indicators each contained a red and a yellow LED. When a vehicle or other object was adjacent to the tractor or trailer, the yellow LED on the corresponding side of the cabin would illuminate. If the driver used the turn signal in the corresponding direction as an indication of lane change, the yellow LED turned off and the red LED became illuminated. Details on the DVI are contained in the IVBSS Human Factors and DVI Summary Report (Green et al., 2008).

OPEN IN VIEWER

Figure 1.

Heavy-truck driver-vehicle interface component locations.

The DVI included both visual and auditory information but relied on auditory warnings when immediate driver actions were required (e.g., time headway is 3 s or less). The visual elements of the DVI conveyed situational information on the visual display, such as data of time headway with yellow LEDs and text of collision alert with red LEDs, more than it conveyed actual warnings. There were three forward sound sources from the DVI, which provided one short tone when time headway dropped to 3 s, 2 s, or 1 s.

Results

The analysis of variance for time headway means and minimum data was conducted for all following events, whereas drivers’ brake reaction time was calculated and compared only for the following events with braking reactions, as described earlier. A total of 162,458 following events were identified and used in this study, which represents 1526.2 driving hours. During all the identified following events, there were 793 braking maneuvers in response to the forward traffic conflicts that triggered warnings in the treatment condition and 218 braking cases in response to the forward traffic conflicts that would have triggered warnings, which were muted from drivers, during the baseline condition. The statistical significance level was set at α = .05.

Mean Time Headway

The analysis of variance for the pooled mean time headway data showed significant main effects for shift, F(1, 16) = 7.41, p = .02; roadway type, F(1, 16) = 106.26, p < .01; wiper state, F(1, 16) = 16.36, p < .01; and traffic density, F(2, 32) = 9.62, p < .01. Mean time headway for the daytime shift (3.10 s) was longer than for the nighttime shift (2.71 s). Longer mean time headway was observed with wipers on (2.93 s) than with wipers off (2.79 s) and on surface roads (3.43 s) than on the freeway (2.35 s). Results showed that mean time headway decreased with on-road traffic. Further pairwise comparison tests found mean time headway in dense traffic conditions (2.61 s) was significantly shorter than in both moderate (2.84 s, p < .05) and sparse traffic conditions (2.94 s, p < .05). No significant differences were observed between sparse and moderate traffic conditions. Trailer configuration effect was found not significant.

Drivers generally maintained longer mean time headway in the treatment condition (2.89 s) than in the baseline condition (2.78 s), although the difference was not statistically significant (p > .05). Two significant interaction effects with treatment condition were observed: Treatment Condition × Traffic Density, F(2, 32) = 4.03, p < .05, and Treatment Condition × Wiper State, F(1, 16) = 10.33, p < .01. For the Treatment Condition × Traffic Density interaction, we found that drivers generally maintained longer mean time headways when warnings were presented to drivers, and the differences were found to be significantly larger in dense traffic conditions (Figure 2). Similarly, drivers had longer mean time headways in the treatment condition, and this difference was found to be significantly higher in the wiper-on state (Figure 3).

OPEN IN VIEWER

Figure 2.

Mean time headway for Treatment × Traffic Density interaction (with standard error).

OPEN IN VIEWER

Figure 3.

Mean time headway for Treatment × Wiper State interaction (with standard error).

The interaction between shift type and traffic density was also found to be significant, F(2, 32) = 4.8, p = .02. Daytime drivers had a longer time headway than did nighttime drivers, and the differences were significantly larger in dense traffic conditions (mean difference of 0.3 s) than in sparse traffic conditions (mean difference of 0.16 s, p < .05). No significant differences were observed between moderate traffic conditions (mean difference of 0.2 s) and dense traffic conditions or between moderate traffic conditions and sparse traffic conditions.

Minimum Time Headway

Minimum time headway of all the pooled following events was also analyzed across the six variables and associated interactions. Results showed significant main effects for roadway type, F(1, 16) = 131.88, p < .01; wiper state, F(1, 16) = 58.11, p < .01; and traffic density, F(2, 32) = 8.98, p < .01. Average minimum time headway was longer with wipers on (1.21 s) than with wipers off (0.82 s) and on surface roads (1.27 s) than on freeways (0.71 s). Pairwise comparison tests on traffic density conditions showed different results from mean time headway analysis. Minimum time headway in the dense traffic condition was significantly longer (1.26 s) than in both moderate (0.92 s, p < .05) and sparse traffic conditions (0.89 s, p < .05). No significant differences were observed between sparse and moderate traffic conditions.

Analysis of the minimum time headways revealed no statistically significant differences between the treatment and baseline conditions. Drivers maintained slightly longer minimum time headways in the treatment condition (0.98 s) than in the baseline condition (0.96 s). In a further exploratory analysis of the data, we found that the proportion of following events during which the minimum time headways reached 1 s or less was 16% (16.3% in the baseline condition; 15.9% in the treatment condition), whereas the proportion of following events with mean time headways of 1 s or less was 3% (3.2% in the baseline condition; 2.9% in the treatment condition).

Three significant interaction effects were observed with Wiper State × Traffic Density, F(2, 32) = 4.47, p < .05; Wiper State × Roadway Type, F(1, 16) = 10.82, p < .01; and Traffic Density × Roadway Type, F(2, 32) = 11.74, p < .01. Although no three-way interactions were observed, the average minimum time headways for Wiper State × Traffic Density × Roadway Type interaction were summarized in Table 1 for the purpose of explanation. For the interaction of Wiper State × Traffic Density, it was found that drivers maintained significantly longer minimum time headways with wipers on than with wipers off across all three traffic density conditions. The differences between two wiper states in the dense traffic condition were significantly larger than in the other two traffic conditions. Similar results were found for the Traffic Density × Roadway Type interaction. Differences between the two roadway types in the dense traffic condition were significantly larger than in the other two traffic conditions. For the interaction of Wiper State × Roadway Type, we found that differences between two roadways were larger with wipers on than with wipers off.

OPEN IN VIEWER

Table 1:

Aggregate Minimum Time Headways in Traffic Density × Wiper State × Roadway Type (in seconds)

	Wipers Off		Wipers On
Traffic Density	Freeway	Surface	Freeway	Surface
Sparse	0.55	0.91	0.78	1.49
Moderate	0.54	1.11	0.72	1.62
Dense	0.93	1.35	1.35	2.7

Driver Brake Reaction Time

Approximately 11% of all the braking maneuvers occurred on freeways and 89% on surface roads, as summarized in Table 2. Approximately 8% of braking maneuvers occurred in the dense traffic condition, whereas 52% occurred in the sparse traffic condition. Only 3% of the braking maneuvers occurred in the wiper-off state, and about 18% were from dual-trailer trucks.

OPEN IN VIEWER

Table 2:

Distribution of Braking Maneuver Counts

Variable	Treatment	Baseline
Road type
Highway	92	25
Surface	701	193
Wiper state
On	23	9
Off	770	209
Traffic density
Sparse	416	119
Moderate	333	67
Dense	44	32
Trailer
Single	666	161
Dual	127	57

Drivers’ brake reaction time was evaluated for the following maneuvers with braking reactions. Main effect of treatment condition was found to be significant, F(1, 16) = 4.87, p < .05; shorter brake reaction time was observed in the treatment condition (mean = 1.62 s) than in the baseline condition (mean = 1.88 s). The data indicated that drivers were able to respond 0.26 s faster, a 15% enhancement compared with the baseline condition, by braking in response to forward traffic conflicts in the treatment condition. Drivers were also found to have a significantly different reaction time in different traffic density conditions, F(2, 32) = 11.91, p < .01). Conditions with more on-road traffic led to significantly shorter driver reaction time. The further pairwise comparison tests showed that reaction time in dense traffic was significantly shorter than when traffic was sparse (mean = 1.18 s vs. 1.83 s) and when traffic was moderate (mean = 1.18 s vs. 1.61 s). Daytime drivers were observed to have a shorter reaction time than were nighttime drivers (mean = 1.69 s vs. 1.88 s), but the differences were not statistically significant. No other significant differences were observed.

Discussion and Conclusion

This study was conducted to evaluate whether an integrated, in-vehicle warning system would improve heavy-truck drivers’ time headway maintenance and shorten their reaction time when there were potential forward traffic conflicts. To achieve the objective, we collected data on following behavior of 18 heavy-truck drivers across a 10-month naturalistic driving study. Results of the analysis confirmed the hypothesis that the warning system does have a positive impact on truck drivers’ safety performance. All the results and discussions in this study are related to the longitudinal driving safety support of the integrated warning system.

Crash studies reported that crash rates for heavy trucks were higher in challenging environmental conditions, such as driving on wet surface roads (Golob & Recker, 1987) or with more traffic on the road (Ramírez et al., 2009). The significant positive interaction effects observed in this study suggest that the warning system could be more beneficial to truck drivers in challenging driving situations. We found that the presence of warnings resulted in an increase of 0.28 s mean time headways (a 11% enhancement) in conditions of dense traffic. Drivers generally maintained shorter time headways when there was more on-road traffic, but they were able to keep a comparably safer distance from leading vehicles with the warning system in such conditions. When weather was adverse (indicated by wipers-on state), there was a 0.20-s increase of mean time headways associated with the system, equaling a 7% enhancement compared with the baseline condition.

The increased time headways observed in this study are shorter than the 0.5-s main effect found in Dingus et al.’s (1997) passenger car study. In addition to the behavior differences between truck drivers and passenger car drivers, another possible explanation for the different effects is that drivers’ behavior in natural driving situations may be different than in controlled experimental situations. In Dingus et al.’s study, drivers’ performance was measured in an on-road experiment in which drivers were asked to follow a designated leading vehicle with no challenging environmental factors included. Drivers may be more attentive and more likely to keep a larger distance from the leading cars in such experimental circumstances. However, it is consistent in both studies that the warning systems help drivers maintain a safer distance from the leading vehicles regardless of the effect size.

It has been recommended that drivers maintain a minimum of a 2-s headway time to reduce the rear-end crash risk for passenger car drivers (National Safety Council, 1992), whereas trucks actually take about 20% to 40% farther to stop relative to passenger vehicle stopping distances (NHTSA, 1987). Therefore, a minimum of a 2.5-s headway time seems more appropriate to recommend to truck drivers. Previous studies on passenger car drivers have observed that drivers usually spend a significant portion of their time at unsafe distance (i.e., with 1-s time headway or less) and maintained a less-than-2-s mean time headway while following other vehicles (Shinar & Schechtman, 2002; Taieb-Maimon & Shinar, 2001). In this study, we found that heavy-truck drivers generally drive more conservatively than do passenger car drivers, but there were still hazardous following behaviors observed. For example, the mean time headway values of truck drivers in both baseline and treatment conditions are greater than 2.5 s. However, still about 3% of the following events were observed with less-than-1-s mean time headways and 16% observed with less-than-1-s minimum time headways.

An interesting finding of this study is that drivers maintained longer time headways in challenging situations, such as driving on surface roads or when the weather was bad. The results suggest a possible compensatory behavior of truck drivers in such conditions. Drivers were also found to have shorter mean time headways and longer minimum time headways when there was more on-road traffic. The results indicate that truck drivers are more likely to follow other vehicles in less-varying headways in conditions of dense traffic. Results showed that daytime drivers maintained longer time headways than did nighttime drivers. This result may be because daytime drivers mainly drove on surface roads, whereas nighttime drivers drove mostly on freeways in this study. No trailer configuration effect was observed, suggesting that it has limited impact on truck drivers’ following behavior.

Drivers responded to the forward conflicts 0.26 s faster, a 15% enhancement in the treatment condition compared with the baseline condition. The ability to respond to a forward traffic conflict (i.e., when an imminent auditory warning was issued) is evaluated in terms of how fast drivers intentionally brake the vehicle. Although there is an argument that shorter headways do not necessarily involve a higher level of risk since drivers may have quicker reaction time (Taieb-Maimon & Shinar, 2001), we found a significant improvement in driver reaction time for all the drivers with the integrated warning system. These findings have significant implications for road safety and suggest that truck drivers can benefit from the integrated warning system by not only maintaining a safer distance from leading vehicles but also by reacting quicker to traffic conflicts. The installation of such warning systems may help improve drivers’ reaction ability and awareness of following distances between themselves and the leading vehicles.

McKnight and Shinar (1992) reported that drivers responded to forward conflicts much faster with shorter time headways than with longer time headways. We observed a similar tendency that truck drivers had a significantly shorter reaction time in conditions of dense traffic (with shorter mean time headway) than in conditions of less traffic (with longer time headway), suggesting that drivers may be more attentive to the situations (i.e., respond quicker) with more traffic on the road. The low proportion of braking maneuvers observed after warnings in the dense traffic conditions also supports such conclusions in that drivers may be more likely to brake before warnings when there is more traffic on the road.

As a result of the limitations in the nature of naturalistic driving studies and driver recruitment methods, some other factors that relate to truck drivers’ longitudinal safety, such as age and gender, were not examined with these data. Therefore, the conclusions of this study are related to the following behavior of a general, skilled truck driver population. Individual differences could be included in future studies.

In summary, the presence of an integrated, in-vehicle warning system with time headway feedback has a statistically significant, positive impact on heavy-truck drivers’ longitudinal driving safety: longer following distance in difficult driving situations and shorter driver brake reaction time. Installation of such a warning system is expected to be beneficial to heavy-truck drivers.

Key Points