Printer Friendly

Crash reduction with an advance brake warning system: a digital simulation.


Rear-end collisions account for approximately 25% of all traffic crashes. Although these are typically not the most severe crashes, an analysis reported by the National Highway Traffic Safety Administration of police-reported crashes showed that in 1993 they resulted in more than 1600 fatalities and more than 500 000 injuries (NHTSA, 1994). Much of the research conducted to reduce such collisions focused on means of reducing the lag time between the perception of imminent danger of a lead car driver and the response of a following car driver. The improved brake light arrangements that culminated in the incorporation of the center high-mounted stop light (CHMSL) is the most ubiquitous result of these efforts. Its installation was shown to reduce brake response time of following drivers as well as to reduce rear-end collision probability (see Digges, Nicholson, & Rouse, 1985; McKnight & Shinar, 1992).

Another approach to reducing brake reaction time of a following driver relative to initial obstacle detection by a lead driver is the advance brake warning system (ABWS). This approach focuses on reducing the time from the moment the driver detects an obstacle to the onset of the brake lights. In a previous study, Shinar (1995) showed that whenever drivers release the accelerator pedal abruptly, they are very likely to follow with a depression of the brake pedal. An ABWS, which exploits this tendency, had been developed and is in wide use in Israel. With this device, quick release of the accelerator pedal (0.3 m/s) activates the brake lights for 1.0 s. If the driver in fact follows the release of the accelerator pedal with braking, then the brake lights appear continuous from the moment the gas pedal is released. Otherwise, they go off after 1.0 s.

To evaluate this approach to ABWS, it is necessary to show that (a) the ABWS does not create a safety risk (by having the brake lights come on in response to anything other than brake pedal activation) and (b) it can reduce rear-end crashes.

The first evaluation of the ABWS consisted of a field study with more than 60 000 km of travel by drivers unaware of the presence of the device in their cars (Shinar, 1995). The study showed that the ABWS does not pose any significant safety risk in terms of false alarm rates (the frequency with which the brake lights are turned on without the contingent braking behavior), which averaged 23%. More important, such abrupt releases turned out to be relatively rare - less than 17 per 1000 km - and the 4 per 1000 km false alarms constituted only 2.4% of all true brief braking actions lasting 1.0 s or less.

In the present study we used a digital simulation to demonstrate the expected savings in frequency and severity of rear-end crashes given travel speeds typical on different U.S. highways (Godwin, 1992), different roadway conditions (dry, wet, and icy), different intervehicle headways (from 0.5 to 2.0 s), and different levels of driver expectancy. Because the actual numbers of miles driven under each unique combination of vehicle separations, speeds, roadway conditions, and level of driver expectancy are not known,the simulation was designed to provide separate estimates for each of the studied conditions.

The primary potential benefit of the ABWS under evaluation is that it can save the following driver the response lag caused by the lead driver's movement time from the gas pedal to the brake pedal. Therefore the simulation used empirically derived distributions of driver perception reaction time for the estimated time from appearance of the target (i.e., an obstacle for the lead vehicle and the lead vehicle's red brake light for the following vehicle) and empirically derived distributions of brake movement time (the time potentially saved by the ABWS), both from field studies by Olson and Sivak (1986). To nullify the effects of increased alertness typical in an experimental study, we made further adjustments to those data with a multiplicative coefficient empirically derived from field studies by Johansson and Rumar (1971).


A digital simulation program was written to evaluate the ABWS effectiveness. The simulation assumed the following scenario:

Two vehicles are moving at the same speed in the same direction. When - in response to an emergency situation - the lead driver brakes hard, the car brakes lock and a skidding stop is initiated. The following driver, in response to the onset of the brake lights of the lead car, brakes equally hard to a skidding stop (or collision). When the lead vehicle is equipped with an ABWS, its brake lights come on as soon as the gas pedal is released. When the lead vehicle is not equipped with ABWS, its brake lights come on only once the brake pedal is engaged. The difference between the two is in the movement time from the gas to the brake pedal.

If one of the two cars has an antilock braking system, then the vehicles' movements following the braking action will change in both the distance covered prior to a complete stop and the relevant coefficient of friction. However, if both vehicles have the system, or the likelihood of having it is equal for the lead and the following vehicle, then the results of this evaluation will be essentially the same.

Standard time-distance-speed functions characteristic of crash reconstruction calculations were applied to both vehicles to calculate distance traveled by time T from moment of maximum braking, speed at time T, distance traveled to complete stop, and crash severity (defined as impact speed).

Independent Variables

The following independent variables - factors that could affect the likelihood of collision - were included in the simulation:

1. Vehicle speeds prior to braking. We used two empirical speed distributions obtained in a 1988 survey of speeds on two types of U.S. roads (Godwin, 1992):

a. On rural, two-lane roads with a 55-mile/h (88.5 km/h) speed limit, the average speed was 59.3 miles/h (95.4 km/h; SD = 6.8 miles/h, 11.0 km/h).

b. On interstate freeways shortly after the speed limit was increased to 65 miles/h (104.6 km/), the average speed was 63.2 miles/h (101.1 km/h; SD = 6.2 miles/h, 9.9 km/h).

2. Intervehicle headway. The standard recommendation (e.g., in driver license handbooks) for safe following distance is 2.0 s. In one set of observations on a Michigan urban interstate freeway, however, most drivers maintained headways less than 2.0 s and more than 20% maintained a headway less than 1.0 s (Evans & Wasielewski, 1982). Rockwell (1972) found that before passing, drivers may reduce their headways to 0.5 s, and Postans and Wilson (1983) observed that 22% of all gaps less than 1.0 s were 0.5 s or less.

Based on these findings, we chose headways of 0.50, 0.75, 1.00, 1.50, and 2.00 s for the simulation.

3. Expectancy and brake response time. We denote perception reaction time as the time from the appearance of a stimulus (i.e., brake lights of the lead vehicle) until the driver initiates release of the gas pedal. We denote brake movement time as the time from the initiation of the release of the gas pedal to the initial depression of the brake pedal. We define total brake response time as the sum of these two times. Thus the total brake response time is affected by the two component times, which are in turn greatly affected by the driver's level of expectancy and alertness.

Because both perception reaction time and brake movement time are variable rather than fixed, the simulation sampled these two component times from empirical distributions of these times obtained in a field study by Olson and Sivak (1986). An important aspect of that study was that the authors measured and found significant differences between the response times to expected and unexpected obstacles. However, in their study, all of the participants were alert - even when the obstacle was unexpected - at least in that they were aware of being part of a study. We therefore used a correction factor for alertness derived by Johansson and Rumar (1971), who demonstrated that in order to allow for an unexpected target when the participant is not alert to being in a study, the reaction time should be multiplied by a correction factor of 1.35.

Based on the aforementioned considerations and empirical data distributions, we assumed the following: For unalert drivers (not expecting the lead car to brake), perception reaction time would average 0.7005 s (SD = 0.0239 s, minimum = 0.338 s) and brake movement time would average 0.3892 s (SD = 0.0185 s, minimum = 0.250 s). For alert drivers (expecting the lead car to brake),perception reaction time would average 0.5182 s (SD = 0.0177 s, minimum = 0.296 s) and brake movement time would average 0.2238 s (SD = 0.0064 s, minimum = 0.150 s).

4. Brake lag time of the vehicles. All braking systems have a lag time from onset of brake pedal depression until actual engagement of the brakes. These typically vary from 0.07 to 0.15 s (TAAR Safety Engineering, Inc., 1990). We assumed that the lag time would be normally distributed between these limits, with an average of 0.11 s and SD = 0.0133 s.

5. Roadway/tire coefficient of friction. Although more than 70% of crashes happen on dry roads (NHTSA, 1991; with an estimated coefficient of friction [cf] = 0.7 g), the simulation was designed to test the effects of the ABWS on wet roads (cf = 0.25 g) as well as icy roads (cf = 0.1 g).

Dependent Measures of ABWS effectiveness

Two dependent measures of ABWS effectiveness were used: (a) the number of crashes prevented with the ABWS versus the number of crashes that would happen without it under the same conditions and (b) average crash severity (in terms of average dV) with and without the ABWS. In some conditions, on some trials the collision was avoided, so crash severity on those trials was zero. Consequently, the more collisions that were prevented, the lower the average severity.

Design and Procedure

There were 120 combinations of vehicle speeds for the two types of roads (2) x headways between the two vehicles (5) x expectancy/alertness levels of the following driver (2) x coefficients of friction between the road and the tires (3) x the presence or absence of ABWS on the lead vehicle (2). We conducted 60 trials of the simulation on each of these combinations. Each trial consisted of randomly sampled values of perception reaction time, brake movement time, and brake system lag from their respective distributions. In all, the database consisted of 7200 simulation trials.


The initial analysis of all 7200 runs showed that with intervehicle headway intervals of 1.5 and 2.0 s, all collisions were prevented, regardless of the presence or absence of an ABWS. Thus in such cases, assuming the following driver is looking at the car ahead when its brakes are applied, he or she has such a safety margin that the ABWS cannot add any marginal safety benefits.

It is noteworthy that the mean total brake response times used here are slightly shorter than those obtained by McKnight and Shinar (1992) in a road study in which unaware pickup truck drivers were trapped behind an experimental vehicle (with various CHMSL configurations) that braked abruptly (though very briefly). Because we needed component perception reaction time and brake movement time from the same study, these distributions could not be used in the present study. The point is that had longer times been used, the ABWS benefits would have extended to longer intervehicle headways.

Because of the absence of any benefits from the headways [greater than] 1.0 s, the more detailed analyses presented in the rest of this paper involved only the 4320 trials conducted with gaps of 0.50, 0.75, and 1.00 s.

Collisions Prevented

In the tables that follow, the percentage of collisions prevented is presented separately for each of the independent variables, with and without the benefits of an ABWS. The total number of runs from which the percentages are calculated have equal representations of all levels of all other variables.

Across all conditions, with the variables and levels examined here, the ABWS prevented 82% of all rear-end collisions with vehicle headways of 1.0 s or less. In contrast, in the absence of ABWS, only 27% of the crashes were prevented, indicating that under these situations a driver is 3.0 times as likely to avoid a crash with an ABWS than without one (odds ratio = 3.03).

To examine the interactions of ABWS with the other variables, we conducted chi-square analyses. Alertness and vehicle headway interacted strongly with the presence or absence of the ABWS, whereas road condition (coefficient of friction) and road type (vehicle speed) did not. Thus the same savings in accidents was observed for all road conditions and all road types. This is understandable in light of the fact that headway was a fixed variable noted in time (seconds of intervehicle separation) and not in distance. We noted a similar absence of differential effects of speed in another analysis utilizing a wider range of speed distributions derived from measurements conducted in Israel with means ranging from 55 km/h (for urban streets) to 99 km/h (for limited-access, intercity highways).

The initially surprising differential lack of effect of the coefficient of friction is explained by the fact that it affects both vehicles equally, making this situation very different from a single-car crash. Thus although the braking of the following vehicle is less effective on a wet or icy road, the lead vehicle is also braking less effectively, and consequently the separation between them remains similar across all three conditions.

Driver alertness interacted significantly with the ABWS. The ABWS was more beneficial to an unalert driver than to an alert one. Table 1 shows that without the ABWS, unalert drivers would not have prevented any of the crashes with an intervehicle gap of 1.0 s or less, whereas an alert driver could avoid 55% of these crashes. However, an alert driver with the ABWS would be expected to avoid all the rear-end crashes.

Vehicle headway had a most significant interaction with ABWS. As expected (and as can be seen in Table 2), with or without ABWS, the greater the headway, the greater the percentage of crashes prevented. The ABWS's advantage increases as the headway decreases. With a headway of 0.5 s, in the absence of an ABWS none of the crashes was prevented, whereas with an ABWS, even with such short headways, 50% of the crashes were prevented. With a headway of 1.0 s the ABWS reaches maximum effectiveness (preventing all crashes), but in its absence 50% of the crashes still occur.

Percentage of Rear-End Crashes Prevented as a Function of ABWS and
Driver Alertness

                 With        Without
Alertness        ABWS          ABWS         Total

Unalert           64             0            32
Alert            100            55            77
Total             82            27            55

Chi square = 323.75. p [less than] .001. Number of simulation runs
in each cell = 1080. Total number of runs = 4320.

Percentage of Rear-End Crashes Prevented as a Function of ABWS and
Vehicle Headway

                With            Without
Headway         ABWS              ABWS         Total

0.5 s            50                 0            25
0.75 s           95                32            64
1.0 s           100                50            75
Total            82                27            55

Chi square = 159.87, p [less than] .001. Number of simulation runs
in each cell = 720. Total number of runs = 4320.

Crash Severity

Crash severity, measured as dV, was calculated for all trials. Trials in which the crash was prevented yielded dV = 0.0 km/h. Consequently the greater the percentage of crashes prevented, the smaller the average dV. An important implication is that severity here is for all incidents, not only those ending in a rear-end collision. If all of the preventable crashes were excluded from the data, and the database for the calculation of average severity included only the trials ending in a collision, the average crash severity would be much higher.

Because it is a continuous measure, severity is a much more sensitive indicator of the ABWS's effectiveness than crashes prevented. Consequently it is not surprising that in a five-way analysis of variance, all main effects except road type were significant at p [less than] .001. Of interest here is the main effect of the ABWS and its significant interactions with other variables. The ABWS had two-way significant interactions with driver alertness, road-tire coefficient of friction, and the intervehicle headway, as well as all three-way interactions with these variables.

Table 3 shows the effects of ABWS, driver alertness, and road-vehicle coefficient of friction on the average collision speed. Three patterns are conspicuous in this table.

First, crash severity increases with increasing coefficient of friction. The reason is that the greater the rate of deceleration of the lead vehicle, the greater the impact of the delayed braking of the following vehicle caused by the driver's and brake system's response lags.

Second, an alert driver following an ABWS-equipped vehicle has a crash severity of 0.00 under all road conditions because all the rear-end crashes are avoided. This means that the elimination of the brake movement time from the total response time function, provides the following driver with enough of a safety margin to avoid all crashes.

Third, for an unalert driver, crash severity is a relatively stable 4.0-4.5 times as high without the benefit of ABWS on the lead vehicle as with it.

Crash Severity (dV) as a Function of ABWS Alertness and Coefficient
of Friction

                  Alert Driver             Unalert Driver

Road                         Without                   Without
Condition     With ABWS       ABWS       With ABWS      ABWS

Dry road         0.00         1.20         1.00         4.11
Wet road         0.00         0.47         0.39         1.60
Icy road         0.00         0.19         0.14         0.64

F(2, 4318) = 538.38, p [less than] .001. Number of simulation runs
in each cell = 720. Total number of runs = 4320.

Crash Severity (dV) as a Function of ABWS Alertness and Headway

Alert Driver Unalert Driver

Vehicle                     Without                   Without
Headway     With ABWS        ABWS      With ABWS       ABWS

0.50 s        0.00           2.55         2.41         3.74
0.75 s        0.00           0.58         0.14         3.76
1.00 s        0.00           0.00         0.00         3.08

F(2, 4318) = 1824.06, p [less than] .001. Number of simulation runs
in each cell = 360. Total number of runs = 4320.

Table 4 shows the effects of ABWS, driver alertness, and vehicle headway on the average collision speed. Two effects are noticeable in this table. First, an alert driver benefiting from an ABWS can avoid all crashes even when the headway is reduced to 0.5 s. Second, the greatest benefits of the ABWS are for unalert drivers maintaining intervehicle separation of 0.75-1.00 s. In these cases the difference in vehicle severity is over 3 km/h.

Table 5 shows the effects of ABWS, driver alertness, and vehicle headway on the average collision speed. The data show how crash severity increases with increasing coefficient of friction and with decreasing headway. Furthermore, the significant interaction demonstrates the synergistic effect of the coefficient of friction and vehicle separation, so that the benefit of the ABWS increases the most as the headway decreases on a dry road and increases the least as the headway decreases on an icy road. On a dry road, with headways of 0.5-1.0 s, the ABWS reduces crash severity by an average of 4 km/h.

Crash Severity (dV) as a Function of ABWS Coefficient of Friction
and Headway

             Dry Road            Wet Road          Icy Road

Vehicle    With    Without    With    Without    With    Without
Headway    ABWS     ABWS      ABWS     ABWS      ABWS      ABWS

0.50 s     2.42     6.28      0.87     2.25      0.35      0.90
0.75 s     0.09     4.22      0.10     1.66      0.02      0.63
1.00 s     0.00     2.78      0.00     1.31      0.00      0.53

F(4, 4318) = 504.09, p [less than] .001. Number of simulation runs
in each cell = 360. Total number of runs = 4320.


The results of this study point to the clear-cut benefits of the ABWS whenever the driver of a lead vehicle brakes abruptly and the intervehicle gap is less than 1.0 s. These benefits are relatively constant independent of the vehicles' traveling speeds. These benefits are realized provided that three assumptions are met: (a) At the moment of braking, the two vehicles are traveling at the same speed; (b) the following driver is actually directing his or her eyes at the lead vehicle's brake lights at the time of their onset; and (c) the following driver's response is to brake as soon as he or she perceives the brake lights.

Unfortunately, there are no data to indicate the joint probability of these events. For this reason, the savings in crash severity and occurrence presented here must be considered as upper limits of the potential benefits.

One of the primary benefits of the simulation program and the data presented here is the ability to establish the situations under which the benefits would be maximized. The statistical analyses of the simulation runs indicate that these are short headways and dry roads, situations typical of most driving in urban high-speed highways and of rush hour driving on major high-speed arteries. Interestingly, the benefits are greater on dry roads than on wet and icy roads - at least insofar as two-vehicle rear-end collisions. Also, the benefits are fairly constant across a wide range of speeds (provided the two vehicles are proceeding at the same speed prior to the onset of braking).

The field study of the false alarm rates of the ABWS (Shinar, 1995) and this simulation study provide the scientific basis for establishing the potential benefits of the ABWS. To estimate the expected benefits of the ABWS in the real world across the different driving situations and with the added uncertainty of drivers' looking and response behaviors, a fleet field study is necessary. Such studies were used to support the requirement of the center high-mounted stop light as standard equipment in all U.S. cars (Digges et al., 1985). A fleet study to evaluate the ABWS has been initiated by Israel's Highway Safety Administration and one of the country's leading insurance companies.


This work was partially supported by the Center for Research for Ergonomics and Safety, Ben Gurion University.


Digges, K. H., Nicholson, R. M., & Rouse, E. J. (1985). The technical basis for the center high mounted stop lamp (SAE Technical Paper Series, No. 851240). Warrendale, PA: Society of Automotive Engineers.

Evans, L., & Wasielewski, P. (1982). Do crash involved drivers exhibit riskier everyday driving behavior? Accident Analysis and Prevention, 14, 57-64.

Godwin, S. R. (1992). Effect of the 65 mph speed limit on highway safety in the U.S.A. Transport Reviews, 12(1).

Johansson G., & Rumar, K. (1971). Drivers' brake reaction times. Human Factors, 13, 23-27.

McKnight, A. J., & Shinar, D. (1992). Brake reaction time to center high-mounted stop lamps on vans and trucks. Human Factors, 34, 205-213.

National Highway Traffic Safety Administration. (1991, November). General estimates system, 1990 (Report DOT HS 807 781). Washington, DC: U.S. Department of Transportation, NHTSA.

National Highway Traffic Safety Administration. (1994, October). Traffic safety facts, 1993 (Report DOT HS 808 169). Washington, DC: U.S. Department of Transportation, NHTSA.

Olson, P. L., & Sivak, M. (1986). Perception-response time to unexpected roadway hazards. Human Factors, 28, 91-96.

Postans, R. I., & Wilson, W. T. (1983). Close following on motorways. Ergonomics, 26, 317-327.

Rockwell, T. H. (1972). Skills, judgment, and information acquisition in driving. In T. W. Forbes (Ed.), Human factors in highway traffic safety research (pp. 133-164). New York: Wiley.

Shinar, D. (1995). Field evaluation of an advance brake warning system. Human Factors, 37, 746-751.

TAAR Safety Engineering, Inc. (1990). Traffic accident analysis and reconstruction. Silver Spring, MD: Author.

David Shinar is a professor of ergonomics in the Department of Industrial Engineering and Management, Ben Gurion University of the Negev, Israel. He received his Ph.D. in human performance and human factors engineering from Ohio State University in 1973.

Eli Rotenberg received his B.Sc. in industrial engineering and management from Ben Gurion University in 1995.

Tal Cohen received her B.Sc. in industrial engineering and management from Ben Gurion University in 1995.
COPYRIGHT 1997 Human Factors and Ergonomics Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1997 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Shinar, David; Rotenberg, Eli; Cohen, Tal
Publication:Human Factors
Date:Jun 1, 1997
Previous Article:The impact of signal bandwidth on auditory localization: implications for the design of three-dimensional audio displays.
Next Article:Automotive HUDs: the overlooked safety issues.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters