For 55 years, a concept known as stopping sight distance (SSD) has figured
significantly in the design of roadways. The American Association of State
Highway Officials (AASHO, known now as AASHTO) first proposed the common
model for predicting SSD, and "the model remains a simple chaining of constant
deceleration after an allowance of lag time for the driver to detect a hazard and
initiate the braking maneuver." While the model itself has remained relatively
unchanged, the term "lag allowance" has been changed to "brake reaction time"
(the estimate of the perception-reaction time (PRT) for braking to occur), and the
initial time of 2 to 3 seconds has been changed to a constant 2.5 seconds.
Recent research suggests, however, that this time-tested model is ready for
change, to a model "that has its roots not in theory or engineering judgement but
in actual performance of real people in real vehicles on real roads." Like
virtually all experimental researchers studying human behavior, those
researching driver behavior are confronted by the problem of subject
awareness--the idea that subjects in a research experiment behave differently
when they know they are being observed and evaluated. To overcome that
problem, researchers have tried pure "covert observation" (when drivers do not
know they are being observed), but success in measuring driver performance
has been limited at best.
The new braking research prompts the question: "How do real drivers behave in
emergency situations when SSD is a significant factor?" For an answer,
researchers launched "a comprehensive braking performance study . . . .
[whose] objective was to evaluate driver braking response to an unexpected
hazard encountered in the roadway--one that would tend to trigger an extreme
braking maneuver . . . ." In their article entitled "Measuring Driver Performance
in Braking Maneuvers" (Transportation Research Record 1550), Rodger J.
Koppa, Daniel B. Fambro, and Richard A. Zimmer highlighted the methodology
of that study.
METHODS
The instrumentation device used in the study at hand measured driver braking
response to a surprise hazard. The instrument had significant advantages over
previous designs. It proved reliable and could be quickly and easily installed
and removed (in less than 30 minutes and 10 minutes, respectively). The
experimenter (onboard) signaled the driver of a hazard by way of a concealed
"button box." Signals were either a red LED light on top of the dash or a horn.
All data were processed into a Compaq laptop computer through an interfacing
data acquisition unit (the DaqBook/100 manufactured by IOtech, Inc., of
Cleveland, Ohio). For each trial run, the Compaq Model SLT 386z20 laptop
computer performed a data acquisition program. After entering the date, subject,
etc., the researcher logged an initiation command to begin data acquisition and
later a second command to end the experiment. The program (written in
QuickBasic) then automatically wrote to a hard-drive file. The instrument was
conveniently powered from the host vehicle's 12-volt electrical supply, an
optimum source that ensured no power loss. Car owners were comfortable
having the instrumentation installed on their own vehicles.
Researchers made more than 3,000 test runs, with a significant success rate;
they discarded only about 100 runs because of instrumentation or computer
problems. A variety of braking scenarios were studied using nine test drivers; a
later phase of the study involved volunteer drivers using either their own or a
test vehicle.
Surprise Braking Maneuvers
Drivers were given a few practice runs to acquaint themselves with the course
and its conditions before experiencing "a completely unexpected barrier that
suddenly sprang up from the pavement in their path." The "barrier" hung from an
arm concealed in a two-inch (5-cm) wide trench in the pavement. Attached to
the arm was a monofilament line. When pulled tight, the line unfolded a piece of
cloth displaying four stop signs. Researchers activated the barrier device with a
garage door opener, and the hydraulic unit that operated the equipment was
hidden behind traffic barrels at the side of the road.
The drivers' approach was at 55 miles per hour (88.5 km/hr), with the barrier
timed to be visible 210 feet (64 m) ahead of the vehicle "using a 1-sec latency
[response time] and pavement friction of 0.80." By allowing such a short time in
which to respond, researchers hoped drivers would brake rather than try to
evade the barrier. The barrier gave way without damage to the vehicle if a driver
hit it. Ten of the drivers did just that, and two drivers showed no reaction and
drove right through the barrier--one mistook it for a finish line; the other had no
explanation.
Expected Braking Maneuvers
All drivers experienced the unexpected braking scenario before researchers
exposed them to expected braking scenarios. Twenty-six drivers used a test
vehicle from Texas Transportation Institute (TTI), and twelve used their own
vehicles. Males and females of various ages participated. For the expected
braking-maneuver experiments, subjects drove 55 mph. Researchers asked
them to stop as quickly as possible if they saw the bright LED light come on--an
event that might or might not occur. Both wet and dry pavement surfaces were
used, as well as straight roadway and horizontal curves. Drivers knew only that
braking was likely to occur on most trials.
On-Road Braking Maneuvers
A section of rural two-lane roadway ("asphaltic concrete in moderate to poor
condition") was used for the on-road portion of the testing. Researchers asked
drivers to drive as they normally would on such a road. A pickup truck was
parked perpendicular to the road in the entrance drive to a pasture; the pickup
was loaded with cardboard drums. At first, drivers drove past the pickup. Later
they were instructed to turn around and travel back the same way. Upon a
signal from the test vehicle, one barrel rolled from the pickup onto the roadway.
To lend credibility to the scenario, a researcher posing as a farmer was
unloading the barrels when this "accident" occurred. The barrel was released
(on the driver's right side) when the test vehicle was 75 feet from the pickup.
The posted speed in this section of the roadway was 45 mph, which again
allowed approximately a one-second response time for the driver to begin
braking.
RESULTS
Table 1 below shows results for the three test situations. The table indicates
mean PRTs ("lag from first onset of signal or appearance of obstacle to initiation
of the braking or other response") and gives a baseline (the baseline condition
involved a stationary-vehicle test in which drivers were told to apply the brakes
as soon as they saw the LED).
Of note are the shorter PRTs for the on-road scenario than the surprise one.
Drivers later said the barricade in the surprise scenario was startling, but they
knew it was controlled by the test. However, none of the drivers thought the
cardboard-drum situation was controlled. Perhaps drivers had a heightened
awareness and response time to a "real," as opposed to contrived, obstacle; in
addition, they may have been more alert and reactive because they knew they
were being tested. Also of note is that all of the 95th percentile estimates and all
but one of the 99th percentile estimates fell within the 2.5-second PRT from the
AASHTO model.
BRAKING PERFORMANCE
Steady Deceleration
Under expected-stop conditions, research shows drivers generally exert an
average steady braking force of -0.35 g (g = acceleration of gravity; 1 g is about
32 feet per second per second). This amount of braking force seems
comfortable for drivers. Computing constant braking force (deceleration) over
the length of the stopping distance in these tests, researchers found that under
Surprise conditions drivers maintained an average of -0.63 g (standard deviation
0.08) in TTI vehicles and -0.55 g (standard deviation 0.07 g) in their own
vehicles.
(Editor's note: Many wet pavement surfaces will not provide the high levels of
braking force cited above. AASHTO assumes a braking force (coefficient of
friction) of 0.28 in its formula for computing stopping sight distance at 70 mph
and a pavement friction of 0.40 for 20 mph.)
Maximum Braking
Analysis of a typical braking run revealed that drivers reached a maximum
braking force on wet pavement of almost -0.6 g within 5 seconds. By scanning
data files, the researchers found drivers of TTI cars averaged -0.91 g (with a
standard deviation of 0.08 g) maximum deceleration, while those driving their
own vehicles averaged a peak deceleration of -0.74 g (with a standard deviation
of 0.09 g). The authors speculated that drivers may have perceived the braking
tests as "severe" and, consequently, avoided putting that kind of wear and tear
on their own vehicles. On the other hand, they were more willing to subject test
vehicles to severe treatment.
CONCLUSIONS/IMPLICATIONS
The instrumentation used in this study proved practical, reliable, and easy to
install; in addition, "the data for each maneuver sequence is amenable to
analyses for a wide variety of purposes." Of particular interest are "the
considerable differences" between performances when drivers were at the wheel
of their own car as opposed to driving a test vehicle. Drivers may inherently be
more conservative with their own vehicles, but the unfamiliarity of a new vehicle
may also come into play. These differences "may disappear under simulated
(but seen to be genuine) emergency conditions," and future studies are
warranted to make further determinations. Expendable research vehicles might
be well suited to "extreme maneuver studies," but a driver's own vehicle is the
vehicle of choice when studying everyday driving conditions. The driver's own
vehicle provides the benefits of both a familiar vehicle and a natural setting--the
"covert ideal" for this type of empirical research.
