The primary objective of this project is to define and evaluate
infrastructure-only Intersection Collision Avoidance System (ICAS) concepts
aimed at reducing the number of intersection crashes. System engineering analyses were performed to define and evaluate
the feasibility and effectiveness of alternative infrastructure-based advanced
technology concepts. This included
development of functional requirements, conceptual designs, and testing the
feasibility of those designs at high crash intersections in three states.
A literature review of human factors studies, crash studies,
and countermeasures identified to reduce intersection crashes was
conducted. The review resulted in a
general description of crossing path crashes at intersections and the factors
causing those crashes. The project
identified certain parameters required for characterizing traffic flow based on
current Intelligent Transportation Systems (ITS) applications/concepts for
traffic management. Information on
human factors issues important to the selection and design of
infrastructure-based technology was identified. These included the driver age,
vehicle gap acceptance, and response to emergency events.
BMI worked closely with Virginia, California, and Minnesota
(The Infrastructure Consortium) to select high-priority candidate intersections
where the feasibility of different ICAS concepts could be evaluated. Crash reports for crashes at candidate
intersections were analyzed to identify types of crossing path crashes that
were occurring and potential causes of those crashes. It was determined that Left Turn Across Path of Opposite
Direction (LTAP/OD); Straight Crossing Path (SCP); and Left Turn Across Path of
Lateral Direction (LTAP/LD) crashes were the most frequent types of crash,
regardless of whether or not the intersection was signalized.
Operations concepts were developed based on crash scenarios
and causal factor patterns obtained from crash reports for the candidate
intersections. Six of the original candidate intersections were chosen for
further study to determine the feasibility of implementing an ICAS at each
location. Data was collected on-site
for each intersection. Based on that data, conceptual designs for an ICAS were
developed to address the crashes observed at each intersection. Based on this
work it was determined that implementing an ICAS to address each of the three
most prevalent types of intersection crashes was feasible. In addition, the benefit-cost analysis
showed recouping of ICAS implementation costs to be quick.
The ITS program within the U.S. Department of Transportation
(US DOT) has sponsored research, development, and field-testing of advanced
safety systems that can improve transportation operations and address safety
problems. One of these problem areas
addresses highway intersections and the potential for Intersection Collision
Avoidance (ICAS) systems to reduce crashes.
Intersection crashes account for almost 30 percent of vehicle crashes in
the United States.
The US DOT has sponsored research studies of vehicle-based
ICAS concepts and of infrastructure-based ICAS concepts for unsignalized
intersections. Since 1993, Veridian,
under contract to the National Highway Traffic Safety Administration (NHTSA),
has studied vehicle-based concepts for ICAS including the use of vehicle-based
systems with map databases and GPS that recognizes the presence of stop signs
or signalized intersections, and vehicle-mounted scanning radars that allow
detection and warning of potential conflicts for left turns. These “autonomous” concepts offer the
potential for deployment through incorporation of the new technology into
passenger vehicles.
While under contract to the Federal Highway Administration
(FHWA), Raytheon Company developed a prototype Collision Countermeasure
System. This infrastructure-only system,
which provides active warning signs activated by loop detectors in the roadway,
was pilot tested at an intersection in Prince William County, Virginia.
Although autonomous vehicle-based concepts and
infrastructure-based concepts for ICAS have been found to offer considerable
potential for alleviating some safety problems at intersections, it has been
suggested that systems communicating between vehicles, or between vehicles and
the highway infrastructure, could potentially provide even greater safety benefits. These “cooperative” systems would
potentially allow vehicle-based systems to utilize information from external
sensors or other sources to supplement the information that can be obtained
from the vehicle itself.
Because cooperative systems require the deployment of
systems both within the vehicle and in the highway infrastructure, deployment
involves risks for vehicle manufacturers, drivers who purchase vehicles with
cooperative features, and the government agencies who deploy the needed infrastructure
components. One potential strategy for
overcoming some of the risks of deploying cooperative systems is to first
deploy infrastructure-only systems that communicate to drivers through existing
traffic control devices, variable message signs, or other means so that drivers
can take appropriate actions to avoid crashes.
These infrastructure-only ICAS systems could then be enhanced and
extended later, once motorists and governments are convinced of the potential
benefits of these ICAS systems and a sufficient population of systems is
available to communicate with intelligent vehicles as well.
The primary objective of this project is to define and
evaluate infrastructure-only ICAS concepts complementary to in-vehicle
autonomous and vehicle/infrastructure cooperative concepts aimed at reducing
the number of intersection crashes. The
scope of this effort is to perform conceptual and analytical work to define and
evaluate the feasibility and effectiveness of alternative infrastructure-based
advanced technology concepts. In
addition to infrastructure-only concepts for all vehicle types, the study will
consider infrastructure-vehicle cooperative systems for transit and emergency
vehicles and intersection crashes involving pedestrian or pedal-cyclists.
To meet the objectives of the project, several tasks were
performed. The first task was to review
previous work. This included a review
of accident studies, human factors work related to crash avoidance, and current
advanced technology intersection safety countermeasures. Included in the literature review was an
examination of technology, sensors and displays capabilities.
The second task was to analyze crashes at selected sites
within the Infrastructure Consortium (IC) states: Minnesota, California, and Virginia. Each IC member state identified 20
high-incident intersections for review and analysis. Police reports for three years of crashes provided a large
database for analysis of crossing path crashes. This database was used to determine primary crash types and
causal factors. A final provision of
this second task was to select two sites from each state that would be
candidates for implementing advanced intelligent countermeasures. The last remaining task was to define and
evaluate ICAS concepts. This task
included developing several concepts for reducing crossing path crashes using
intelligent vehicle systems and sensors, communication displays, etc. The final step was to apply these concepts
to the six candidate intersections located in the IC states to test for
feasibility. This was performed by
collecting field data and applying it to the requirements of the particular
concepts.
The following report provides synopses of the details,
methodology and results of each of the aforementioned tasks. During the completion of these tasks,
several detailed reports were developed and are referenced herein. They are attached as Appendices at the end
of this report.
Studies related to crash data were reviewed
to develop descriptive statistics of accidents occurring at both signalized and
unsignalized intersections. These
reviews were used to identify accident types that could potentially be reduced
by infrastructure-based technology countermeasures. A review of existing intelligent intersection crash
countermeasures was also performed to identify innovative advanced
infrastructure technology concepts that have been proposed and/or implemented
to improve intersection safety.
Finally, a review of human factors studies was conducted to determine
the issues that will ultimately affect and influence the selection and design
of infrastructure-based collision countermeasures. Full details on all studies or reports reference herein can be
found in Appendix A.
Two key documents
were reviewed:
“Analysis
of Crossing Path Crash Countermeasure Systems” by W.G. Najm and Jonathan
Koopmann. September, 2001 (1)
“Analysis
of Crossing Path Crashes” by D.L. Smith and W.G Najm, Project Memorandum
DOT-VNTSC-HS019-PM-99-01, Dec. 1999 (2).
Both reports used
1998 General Estimates System data as the basis of their analysis of
crossing path crashes at intersections.
In both reports,
crashes are classified into one of the following six categories (also shown in
Figure 1, reproduced from the Najm and Koopmann report):
1.
Left Turn Across Path – Opposite Direction (LTAP/OD)
2.
Left Turn Across Path – Lateral Direction (LTAP/LD)
3.
Left Turn Into Path (LTIP)
4.
Right Turn Into Path (RTIP)
5.
Straight Crossing Path (SCP)
6. Other/Unknown
Figure
1: Intersection crossing path crash types (Source:
Najm and Koopmann)
The (GES) vehicle
level Accident Type variable was used to discern crash type.
The Smith and
Najm report provide the following distribution of crossing path crash scenarios
for all vehicles based on 1998 GES data:
· 29.9% SCP
· 27.5% LTAP/OD, and
· 19.7% LTAP/LD, and
· 5.9% LTIP, and
· 5.7% RTIP, and
· 11.3% Other
Both reports
provide descriptive statistics disaggregated by the physical setting of the
crossing path crashes. The GES variable
‘relation to junction’ was used to determine the physical setting of the
crossing path crash. The location of
the crash reported in this variable is determined by the location of the first
harmful event. Both interchange and
non-interchange junctions were included.
Crossing path crashes occurring at intersections were aggregated apart
from crossing path crashes at junctions with a roadway and a driveway, alley,
ramp, or grade crossing. The latter
group was collectively referred to as ‘driveway’ crashes. Crashes coded as ‘intersection-related’ were
included with crashes occurring at intersections.
2.1.4
Vehicle
Types
In the Smith and
Najm report, vehicles are classified into light vehicles (at least one light
vehicle involved and no special use vehicles), commercial trucks (at least one
and no special use vehicle), transit buses, and emergency vehicles such as
police, ambulance, and fire. The GES
variables ‘hot-deck imputed body type’, ‘special use’ and ‘emergency use’ were
used to determine the vehicle classification.
The two main
reports establish causal factors differently.
The Najm and Koopmann report establishes causal percentages by violation
and then applies the percentages across all crash scenarios. It separates crashes into signal/sign
violation and insufficient gap by 1998 GES violations data (after sorting out the
percentage of crashes that are estimated to have involved drugs or alcohol
based on the Smith and Najm report.)
Based on an analysis of an 81 crash case sample of 1992-1993 CDS data,
they applied the following percentages to signal violations:
· 46% did not see the signal or its status (37 samples).
· 18% tried to beat the amber light (15 samples).
· 36% deliberately ran the red light (27 samples).
A similar analysis
of a 40 case sample of stop sign violations was conducted:
· 90% did not detect the presence of the stop sign (36
samples).
· 10% of drivers deliberately ran the stop sign (4
samples).
The Najm and
Koopmann report does not mention the sampling error associated with applying
this distribution of cause. At a 95%
confidence level, the causal factors have a sampling error of +/- 8.4 to 10.9
percent.
Smith and Najm
establish cause by crash scenario and platform. They re-evaluated the narrative statements and kinematic
evaluations of 498 Crashworthiness Data System (CDS) crashes previously evaluated by the
NHTSA.
Smith and Najm
provide a distribution of causal factors for LTAP/OD by traffic control and
vehicle involvement (turning vehicle or vehicle going straight). Two weighting schemes are used to
represent the distribution. A “Ratio
Weight Factor” was assigned to each crash case in the CDS. Cases were also weighted using a severity
weighting scheme that matches the injury severity profile of the CDS sample to
the GES.
The most common causal factor of turning vehicles involved in these
crashes is ‘Looked, misjudged velocity/gap’ for both the CDS weighting scheme
distribution and the GES weighting scheme distribution, followed by ‘Looked,
did not see’ and ‘View obstructed by intervening vehicles’. The two distributions differ the most for
the causal factors ‘Distracted (Inattention)’ and ‘Attempted to beat other
vehicle’. The most common causal
factors for LTAP/OD, SCP, and LTAP/LD crashes are presented in Table 1 by crash scenario,
intersection control type, and vehicle involved.
Table
1: Most
common causal factor by crash scenario, intersection control, and vehicle
involved developed by Najm and Smith
|
Scenario
|
Intersection
|
Vehicle
Involved
|
Most
Common Causal Factor
|
Weighing
|
|
LTAP/OD
|
Traffic
Signal
|
Vehicle
Turning
|
Looked,
misjudged velocity/gap
|
CDS and GES
|
|
LTAP/OD
|
Traffic
Signal
|
Going
Straight
|
Deliberate
red signal violation
|
CDS and GES
|
|
LTAP/OD
|
No Control
|
Vehicle
Turning
|
Distracted
(inattention)
|
CDS and GES
|
|
LTAP/OD
|
Unsignalized
|
Vehicle
Turning
|
Looked,
misjudged velocity/gap
|
CDS and GES
|
|
SCP
|
Traffic
Signal
|
Going
Straight
|
Distracted
(inattention)
|
CDS and GES
|
|
SCP
|
Traffic
Signal
|
Going
Straight
|
Looked did
not see
|
CDS
|
|
SCP
|
Stop Sign
|
Going
Straight
|
Distracted
(inattention)
|
GES
|
|
SCP
|
Unsignalized
|
Going
Straight
|
Looked did
not see
|
CDS and GES
|
|
LTAP/LD
|
Traffic
Signal
|
Vehicle
Turning
|
Deliberate
red signal violation
|
CDS
|
|
LTAP/LD
|
Traffic
Signal
|
Vehicle
Turning
|
Distracted
(inattention)
|
GES
|
|
LTAP/LD
|
Traffic
Signal
|
Going
Straight
|
Distracted
(inattention)
|
CDS and GES
|
|
LTAP/LD
|
Stop Sign
|
Vehicle
Turning
|
Looked did
not see
|
CDS and GES
|
The Najm and
Koopmann report does not mention pedestrians and bicyclists. In the Smith and Najm report, some
descriptive statistics of pedestrian and bicycle crashes are provided from an
analysis of 1998 GES and Fatal Analysis Reporting System (FARS) data. They note that 39.4 percent of all
pedestrian crashes occurred at intersections or were intersection related. Another 5.7 percent occurred at
driveways. Of the pedalcyclist crashes,
59.4 percent occurred at intersections and 19 percent occurred at
driveways. A distribution by traffic
control type is provided, but only for fatal crashes at intersections. Of
1,144 fatal pedestrian intersection crashes, 45 percent occurred at signalized
intersections, 7.8 percent occurred at Stop sign controlled intersections, and
47.2 percent occurred at uncontrolled intersections. Of the 225 fatal pedacyclist intersection crashes, 35.1 percent
occurred at signalized intersections, 31.1 percent occurred at Stop sign
controlled intersections, and 33.8 percent occurred at uncontrolled
intersections.
In the Smith and
Najm report, the crashes are represented by crash event description for
pedestrians and pedacyclists. The most
common (41.7 percent) crash event description for pedestrian crashes at
intersections was ‘Intersection-other’.
The crash event description, ‘Vehicle turn/merge’ accounted for 38.1
percent of the pedestrian crashes at intersections. A similar crash event distribution is provided for pedacyclist
crashes at intersections. The leading
crash event descriptions for pedacyclists include: ‘Motorist drives out into or
in front of cyclist at intersection’, ‘Motorist turns or drives out in front of
cyclist at an intersection controlled by a stop sign or flashing red signal,
motorist obeys the sign but fails to yield to cyclist’, ‘Cyclist fails to yield
to motorist at an intersection controlled
by stop sign or flashing red signal (crossing path)’, ‘Motorist turns
right in front of cyclist proceeding in parallel path, cyclist either
proceeding in same direction or from opposite direction’, and ‘Controlled
intersection-other’.
Since the crashes are not grouped into causal types, countermeasures
cannot be easily applied. The following
documents were reviewed to supplement the pedestrian and bicycle elements of
this task:
“Pedestrian-Vehicle
Crash Types: An Update.” Stutts, J.C.,
Hunter, W.W., and Pein, W.E. In Transportation
Research Record 1538 (3).
“Bicycle-Motor
Vehicle Crash Types: The Early 1990’s.” Hunter, W.W., Pein, W.E., and Stutts,
J.C. In Transportation Research
Record 1502 (4).
Stutts, Hunter, and
Pein developed pedestrian and vehicle crash types based on a sample of over
5,000 pedestrian vehicle crashes from California, Florida, Maryland, Minnesota,
North Carolina, and Utah. Their crash
typing revealed that nearly one-third of the crashes were coded as
intersection-related. (This is
comparable to the Najm and Koopmann report finding.) Alleys and driveways were considered intersections only if they
were controlled by a signal.
The following relevant descriptive statistics of
intersection-related pedestrian and vehicle crashes emerge:
· 30.5% involved the vehicle turning or merging.
· 15.9% involved a driver violation.
· 1.5% involved a pedestrian misjudging a gap when
crossing.
Hunter, Pein, and
Stutts conducted a similar crash typing of approximately 3000 bicycle-motor
vehicle crashes from 1991 and 1992.
Cases were drawn from a population-based sample in California, Florida,
Maryland, Minnesota, North Carolina, and Utah.
They analyzed the crash diagrams, narratives, and other information of
the cases to compile factors pertinent to the crash.
Almost half of the
crashes took place at roadway intersections.
Another 3.6 percent were intersection-related. A stop sign was the traffic control device in approximately 25
percent of the crashes overall. A
traffic signal was the controlling device 16 percent of the time.
Fifty-seven percent
of the crashes were crossing path crashes.
The most frequent crossing path crash involved a motor vehicle failing
to yield to the cyclist (37.7 percent of crossing path crashes and 21.7 percent
of all crashes). The second most
frequent involved a cyclist failing to yield to a motorist at an intersection
(29.1 percent of crossing path crashes, and 16.8 percent of all crashes).
The driver was not
interpreted as contributing any fault in 43.1 percent of the cases. The bicyclist was not interpreted as
contributing any fault in 23.4 percent of the cases. A driver yield violation was noted as the most common
driver-contributing factor (24 percent).
Cyclist failure to yield was also the most common contributing factor of
the cyclist (20.7 percent). Stop sign
violations and traffic signal violations by the cyclist were noted as a
contributing factor 7.8 percent and 4.7 percent of the time, respectively. Only 1.9 percent of the cases involved a
driver violation of a stop sign or traffic signal.
The Smith and
Koopmann, Stutts et al., or Hunter et al. reports did not provide the in-depth
causal type distribution for pedestrians and bicyclists that is available for
motor vehicle crashes. However, the
three reports together reveal the approximate amount of crashes at
intersections, the traffic control at those intersections, and insight into the
causal factors of those crashes.
This review
provides a general description of the crossing path crash types occurring at
intersections. The reviewed documents
provide a sufficient initial description of the circumstances of crossing path
crashes.
The purpose of this task was to identify and summarize
published literature on innovative advanced infrastructure technology concepts
that have been proposed and/or implemented to improve intersection safety. A
search was conducted of TRB’s TRIS website, DOT’s online library, along with a
general Internet search. Literature was
also obtained from trade journals and through contacts with vendors and state
or local agencies.
Literature exists on evaluations of various detectors based
on requirements of ITS traffic management applications. A summary of evaluation results is
provided. The bulk of the information
obtained was from the Hughes Report - a multi-year evaluation of
several detector technologies, with a bias towards their use in ITS, traffic
management-based applications (5).
A literature search identified nine novel
infrastructure-based Intersection Collision Avoidance System (ICAS) concepts. Seven of the concepts relate to prevention
of vehicle-vehicle accidents. Two
concepts are intended to prevent vehicle-pedestrian crashes. The last concept discussed is intended to
prevent vehicle/emergency-vehicle crashes at intersections. The concepts, along with the type of crashes
they are intended to avoid, are listed in Table
2.
Table
2: ICAS
deployment concepts
Deployment Concept
|
Crashes Deterred
|
|
Minor Road Intersection Warning
|
SCP, LTAP/LD, RTIP, LTIP
|
|
Major Road Intersection Warning
|
SCP, LTAP/LD, RTIP, LTIP
|
|
Left Turn/Oncoming Traffic Warning
|
LTAP/OD
|
|
Dilemma Zone Control –Signalized Intersection
|
Rear-end Crashes, SCP, LTAP/LD, RTIP, LTIP
|
|
Red Light Run Photo Enforcement
|
SCP, LTAP/LD, RTIP, LTIP, Pedestrian
|
|
Red Light Hold
|
SCP, LTAP/LD, RTIP, LTIP
|
|
Pedestrian
Light-in-Pavement Crosswalks
|
Vehicle-Pedestrian Crashes
|
|
Pedestrian
Detection in Crosswalk
|
Vehicle-Pedestrian Crashes
|
|
Emergency
Vehicle Signal Prioritization
|
Emergency Vehicle-Vehicle Crashes
|
Two additional, common infrastructure-based ICAS are briefly
discussed – Protected Left Turn Signal Phasing (PLTSP) and Active Advanced
Warning Signs. A summary of the ICAS
listed above ensues, with examples found in Appendix A.
An Intersection Collision Warning (ICW) is designed to alert
minor road vehicles stopped at an intersection of the presence of oncoming
major road vehicles. Alternatively, it
can also warn major road vehicles that there is a minor road vehicle stopped at
the intersection (and therefore, a potential for a collision exists). An ICW employs detector (speed or presence)
for major road vehicles and active warning signs that warn minor road motorists
of approaching major road vehicles. Figure 2a shows a layout of sensors and warning signs that
comprise an experimental ICW system in Prince William County, VA. Minor road vehicles have a warning sign that
indicates the presence of vehicles coming from either the left or right side,
as detected by loop detectors labeled 1 through 6 in the Figure 2. Detectors
1,2 and 4,5 comprise two closely spaced loops that measure major road vehicles’
speed and activate a timer that controls the activation length of the minor
road warnings.
Additionally, major road vehicles are warned of minor road
traffic detected by loops labeled 7 through 10 in Figure 2. The warning
signs are activated by the presence of a minor road vehicle stopped at the
intersection – detectors 7 and 8 – and by approaching minor road
vehicles – detectors 9 and 10. The
system requires installation of detectors in every lane of the major road. Sensor locations are based on American Association
of State Highway and Transportation Officials (AASHTO) sight distance values
and/or the geometric design of the road and design speed of the road. Additional sensors, not shown, can detect a
vehicle slowing down to turn left and provide a warning to major road vehicles,
similar to the warning of presence of minor road vehicles.
Figure 2
(a) and (b): Two examples of ICWs –
comprehensive ICW in Prince William County, VA and proposed ICW in Amberstone
National Park, WY, respectively.
An ICW has been proposed for the rural greater Amberstone,
Wyoming area. The concept is similar to
the deployment described above in that it detects the presence of vehicles on
major roadway and relays information to vehicles waiting to cross on minor
roadways at two-way, stop-sign controlled intersections. This ICW would only serve minor road
vehicles.
It has been proposed to apply this technology along rural
high-speed highways where intersection control consists of two-way stop sign or
where there is a statistically high number of crossing-path accidents and
traditional countermeasures are unable to mitigate the problem. Passage detectors, in the form of magnetic
loops are installed in every major road lane on either side of the intersection. Again, placement of the detectors is based
on AASHTO safe distance values for crossing, left turn, and right turn
maneuvers. Safe crossing information
is relayed to the driver through stop sign-mounted beacons. The timing between the detector and the
active sign for this deployment is not based on the speed of the approaching
major-road vehicle; rather, the timing relies simply on the presence of a
vehicle.
Japan has developed a safety plan that combines the use of
roadside infrastructure and in-vehicle sensor technology to assist or automate
a driver’s decisions in numerous potentially hazardous crash conditions. Four AHS pertain to intersection collision
avoidance with other vehicles, as well as pedestrians. These systems are similar to other advanced
warning systems discussed with the two exceptions being the source of data used
by the system and the method used to relay a message to the driver. Components common to all AHS are road
surface detectors; aboveground pedestrian/vehicle detectors; starting point
markers; and two-way roadside/vehicle communication stations. Starting point markers are short-wave band
(13.56 MHz) emitters embedded in the road surface. In conjunction with an on-vehicle receiving antenna and
processing unit, it functions as a passage detector. Data input to the AHS would come from five sources:
· Stored vehicle-specific data (braking ability, vehicle
type, etc.),
· Variable subject vehicle data (speed, position)
generated both in-vehicle and through in-ground detectors,
· Variable principal other vehicle data (speed, position)
generated from roadside detectors,
· Stored road data (shape, intersection dimensions,
lateral and longitudinal grade), and
· Variable road data about the surface condition of the
road (i.e. wet, dry, icy).
Where a warning is given to a motorist, certain critical
variables related to driver behavior must be assumed – perception reaction time
(PRT), normal deceleration, etc. The
AHS uses a driver response to information time of 2.65 seconds. The assumed driver response time to a
warning is 1.0 seconds. The assumed
normal deceleration rate is .3g, and emergency braking is assumed to be
.5g. Also, these values are in
conjunction with a flat, straight, wet road (with a minimum coefficient of
friction value of .5). The times were
chosen to cover 90 percent of motorists (70 percent of elderly drivers). The roadside communication is by two-way
radio transmission operating at 5.8 GHz and covering a transmission zone of 100
meters.
The dilemma zone at an intersection approach is the area
upstream of an intersection where a motorist has an option of either slowing at
an amber signal or speeding up to beat a red signal. After a given minimum green time on an approach, the signal
controller has the option of changing the phasing to amber, based on the
presence, or lack of presence, of vehicles in the dilemma zone. The fewer cars that are caught in the
dilemma zone when the signal changes from green to amber, the fewer chances
there are for a driver to make a decision that could potentially lead to red
light running.
A typical RLRPE system works as follows: When a signal turns red, a nearby camera
activates. A vehicle entering the
intersection after the red signal is then photographed. A time delay, typically 0.3 seconds after
the beginning of the red phase, is usually required before a photograph is
taken. The vehicle is photographed
prior to entering the intersection and subsequently while it is crossing the
intersection. The second picture is
usually on a time delay (0.5 to 0.9 sec) after the first. In both cases, the photograph encompasses
the vehicle license plate and signal phase.
In addition, there is usually a minimum speed (typically 15 – 20 mph)
required for the pictures to be taken.
An RLRPE system targets right-angle (SCP) crashes. RLRPE works as a deterrent through fines or
by the passive warning sign that usually accompanies the cameras.
The RLH system tries to determine probable
beginning-of-phase signal violators and extends the all-red light phase to keep
cross traffic from entering the intersection.
A RLH system detects vehicles in a given zone located prior to an
intersection. When the signal turns
from amber to red, the detectors provide vehicle information to the signal
controller. The controller then makes
the determination to keep the cross traffic phase red for a given amount of
time – the time that the controller computes is required by the violator to
clear the intersection.
In-pavement lighting consists of lights imbedded in a
crosswalk that are activated by pushbutton box, microwave detectors, or
infrared detectors (in the form of bollards).
The in-pavement lighting is used to draw the attention of motorists to
pedestrians in the crosswalk – particularly at night.
Alternatively, infrared detectors are positioned so that
they can “see” a person who is waiting, at the curbside, in front of the
crosswalk. Pedestrian-in-crosswalk
detectors allow for elongation of the cross-traffic red phase. Both curbside and in-crosswalk detectors can
activate strobes on a static crosswalk sign to alert motorists of potential
conflicts- particularly where limitation in motorists’ sight distance
exists.
Emergency vehicle signal preemption determines if an
emergency vehicle is approaching and automatically gives a green phase to that
vehicle and a red phase to all other traffic.
One product has been successfully installed in the Northwest U.S. using
a directional microphone mounted to a signal head. More prevalent is Opticom™ by3M. Opticom™ uses infrared transmitters mounted in-vehicle with
receivers mounted to signals to grant right of way to approaching vehicles.
With the premise that speed and presence detectors are the
primary ingredients for an ICAS, a general summary is provided of various
available detection technologies for sensing moving and stationary
vehicles. Each detector summary
describes its mode of operation, because there may be a technical limit in its
ability to perform certain functions required of an ICAS. The following types of detectors were
reviewed:
· Microwave and Millimeter-Wave Radar,
· Pulsed-Doppler Ultrasound Detector,
· Active LED Infrared Radar,
· Inductive Loop Detectors,
· Video Image Detection System,
· Passive Infrared (PIR) Detectors,
· Magnetometer,
· Piezo Electric Detectors, and
· Passive Acoustic Detectors.
Detector installation, maintenance and cost issues are also
addressed. In addition, each summary
discusses the advantages and disadvantages of a detector based on their
performance under certain conditions.
Finally, a few vendor specifications for each detector are
included. Appendix A, the Task 1
report, provides this information on the above 9 detector types.
This section summarizes evaluations of various detector
technologies. The evaluations
identified the best performing detectors for the following traffic management
requirements: count in high and low
volumes, speed in high and low volumes, and reliability in inclement
weather. No evaluations of detectors
have been performed with respect to functional requirements of ICAS, however,
the detector requirements between traffic management and ICAS applications are
similar. Therefore, this section
discusses the requirements for ICAS and the applicability of the findings to
infrastructure-based ICAS.
The “Detection Technology for the IVHS” project
obtained and installed current detectors in three states with diverse
climates. Testing was performed in all
types of environments and in both high volume and low volume traffic. The project determined five detector
requirements necessary for ITS traffic management applications. Table
3 below lists the best performing technologies, based
on performance, for each of the five requirements. The technologies below are not listed in order of performance.
Table 3: Most accurate technologies for traffic
management requirements
|
Requirement
|
Most
Accurate Technologies
|
|
Low Volume Count
(presence)
|
Microwave Doppler
Microwave True Presence
Visible VIP
SPVD Magnetometers
Inductive Loop
|
|
High
Volume Count (presence)
|
Microwave Doppler
Microwave True Presence
Visible
VIP
Inductive Loop
|
|
Speed
in Low Volume Flow
|
Microwave Doppler
Magnetometers in series
Inductive Loops in series
|
|
Speed
in High Volume Flow
|
Microwave
Doppler
|
|
Reliability in Inclement Weather
|
Microwave Doppler
Microwave True Presence
SPVD
Magnetometer
Inductive Loop
|
In most of the tests, ground truth for the vehicle count was
obtained with video surveillance. Even
though loops were evaluated, they were also often used to determine true count
against which other detectors were compared, because they are a mature
technology.
The purpose of this subtask was to
identify human factors issues that influence the design of infrastructure-based
collision avoidance technology- particularly drivers’ ability to identify gaps,
speeds and relative distances of other vehicles. Also of concern was the ability of drivers and pedestrians to
understand new or innovative technologies.
A literature review revealed many studies on gap acceptance and
perception-reaction time.
As part of a study to determine the appropriate sight distance
to be used in design, Harwood, Mason and Brydia determined the
critical gap for right-turning passenger cars to be about 6.4 seconds (6). The critical gap for left-turning passenger
cars was calculated to be about 8.1 seconds.
Single unit trucks require critical gaps that are 1 to 2 seconds longer,
and combination trucks require critical gaps that are 1 to 2 seconds longer
than single unit trucks. This study
contrasted with the current AASHTO model, which suggests that sight distance
requirements for left and right turns are essentially equal.
Staplin found a significant effect in that
increasing subject age resulted in larger gap requirements across all varying
laboratory test methodologies (7).
However, older driver gap judgment distances did not change
significantly for the higher speed approach compared to their gap judgment
distances for the lower speed approach; this is contrary to what should
happen. Acceptable gap distances should
lengthen as the oncoming vehicle speeds increase. The study recommended the following countermeasure strategies: 1)
Cue the older turning driver to the presence of vehicles approaching at
significantly higher-than-expected speeds, and 2) Slow down through traffic and
make through drivers more aware of the potential for a conflict ahead with a
turning vehicle.
Wilson, Sinclair, and Bisson (as reported in TRC 419)
conducted research on brake reactions of 40 alerted and practiced drivers. When confronted with emergency obstructions
in the road, traveling at 60 km/h during nighttime driving on a two-lane
roadway, the average reaction time was 0.96 seconds. The 99th
percent response time was 1.6 seconds.
Brake reaction time included the driver perception time, the driver
reaction time, and the vehicle braking response time. Staplin, Lococo, and Sim (as reported in TRC 419) found no
difference between older and younger drivers when confronted with a single
control movement response. This was
confirmed by Olson and Sivak (as reported in TRC 419), who found that both
young and old age groups produced a perception reaction time of 1.6 seconds
when confronted with an unexpected roadway obstacle.
Evaluations by Fugger, et al. on the behavior and gait
response of pedestrians at signal-controlled intersections were also reviewed
(8). Pedestrians who were looking at
the WALK signal had a perception-reaction time of 0.84 seconds (±
0.51), while those who anticipated the light change had a perception-reaction
time of 0.77 seconds (±0.75). Pedestrians
who were distracted had a perception-reaction time of 1.87 seconds (±1.0). The study also showed pedestrians over 55
years of age to have a longer perception-reaction time.
This brief review provides information on some of the human
factors issues that should be addressed in the selection and design of
infrastructure-based technology. This
review concentrated on literature describing the factors that affect driver and
vehicle gap acceptance and response to emergency events. Further details on these studies, as well as
those regarding dilemma zone, can be found in Appendix A, “The Task 1
report.”
The three Infrastructure Consortium states, Virginia,
California, and Minnesota assisted with the selection of high-priority
intersections for study, as well as information about each of the candidate
intersections. BMI requested that
either each state select 20 intersections, or provide access to their state
crash database so that the intersections could be selected based on crash
history. Candidate intersections that
were diversified in traffic control and general environment, and included some
intersections that had experienced crossing path crashes involving pedestrians,
bicyclists, emergency vehicles, and transit vehicles were requested.
California Department of Transportation (Caltrans) staff
provided a text version of the crash database and the accompanying vehicle
database for both rural District 3 and urban District 4. The database contained location, roadway,
crash, and environment data for crashes occurring between 1997 and 1999.
These two databases were used to select preliminary
candidates for the study. Sixty
intersections total were selected from the two districts. The preliminary candidates were selected
based on the number of total crashes and the number of angle crashes at the
intersection. The inventory information
provided by Caltrans staff was used to select 21 candidate intersections for
the study. The selected intersections
have a variety of traffic control types.
The intersections were also selected because of their reasonable
proximity to one another. The candidate
intersections are briefly described in Appendix B.
Preliminary discussions with Minnesota Department of
Transportation (MnDOT) staff revealed that MnDOT was interested in
concentrating on crashes at intersections on rural, high-speed corridors with
low volume minor road crossings. Based
on the interest area of MnDOT and the requirements of the study, the selected
intersections are stop-controlled and signalized; are located on the state
trunk system; and are all
located within one district (to economize site visits).
MnDOT staff selected 20 candidate intersections for the
study. Each of the 20 intersections was
located along Highway 10, a principal arterial through the state. The 20 candidate intersections are briefly
described in Appendix B.
The Virginia Department of Transportation (VDOT) provided
it’s 1999 Intersection Critical Rate Report.
The report provides 1999 crash rates and frequencies in Fairfax County
for crashes within 0.03 miles of an intersection. Specifically, for each intersection, the report provides the
crash rate, critical crash rate, injury crash rate, fatal crash rate, total
crashes, total fatal crashes, total injury crashes, and total property-damage-only
crashes. Additionally the report notes
whether the intersection is signalized or unsignalized, the number of
approaches, and the estimated entering AADT (Annual Average Daily Traffic). Twenty intersections on this list with high
crash rates and high crash frequencies were selected to be used as candidates
for the study. This set of 20
intersections was varied in the number of approach lanes and were both
signalized and stop-controlled. The
candidate intersections are briefly described in Appendix B.
The overall set of 61 candidate intersections were varied in
intersection configuration, traffic control type, and surrounding environment
as shown in Table
4.
Table
4:
Distribution of candidate intersections by general environment and traffic
control
All of the candidate intersections in Virginia were from
Fairfax County, which is generally urban in nature, while Minnesota and
California intersections were located in both urban and rural areas.
All intersection-related crashes were analyzed to identify
the types of crossing path crashes that were occurring and potential causes of
those crashes. Crossing path crashes
are classified into one of the following six categories:
1. Left Turn Across
Path – Opposite Direction (LTAP/OD), and
2. Left Turn Across
Path – Lateral Direction (LTAP/LD), and
3. Left Turn Into
Path (LTIP), and
4. Right Turn Into
Path (RTIP), and
5. Straight Crossing
Path (SCP), and
6. Other/Unknown
The analysis of candidate
intersections included a manual review of three-years of crash reports from
each intersection. When possible, a
field review was also conducted of the intersection. Crash diagrams for each candidate intersection were also prepared
and can be found in Appendix B.
The three most recent years of available crash data was
requested for each of the candidate intersections in the form of hard-copied
police crash reports. The crash reports
from each candidate intersection were manually reviewed to determine the
circumstances involved in each crash.
Crashes that were not intersection related or that occurred at another
intersection (and were erroneously referenced to the candidate intersection)
were sorted out. Crashes that were
related to the operation of the intersection were considered
intersection-related. If the exact
circumstances of the crash were not explicitly stated, the crash was considered
intersection-related.
The subset of intersection-related crashes was further
classified into crossing path crashes and non-crossing path crashes. The general circumstance, report number and
maximum injury sustained for all intersection-related crashes were recorded in
a database. Other detailed information
recorded in the database included the crossing path crash types, traffic
control governing each vehicle, vehicle directions, circumstances involved in
traffic signal violations, and the circumstances involved in poor gap
acceptance. Any factors contributing to
the crash such as alcohol use, excessive speed, or sight obstructions were also
noted in the database. Additionally,
general conditions such as lighting, weather, and roadway conditions at the
time of the crash were recorded.
In the three-year period from 1997-1999, there were 686
crashes at the 21 candidate intersections in California. Of these 686 crashes, 386 (56 percent)
occurred within the intersection, defined by the extension of the crosswalks on
each approach. Hard copies of the crash
reports for these 386 crashes were requested.
The crash reports were used to separate the 386 crashes that occurred within
the intersection into crossing path and non-crossing path crashes. Of the 386 crashes occurring within the
intersection, 312 (81 percent) were crossing path crashes. The frequency of total intersection crashes,
crashes within the intersection (non-crossing path), and crossing path crashes
for each candidate intersection in California are displayed in Appendix B. The distribution of the crossing path crash
types at the 21 candidate intersections in California is provided in Table 54. The SCP
crash was the predominant crash type at the candidate intersections, followed
by the LTAP/OD crash type.
Table
5: Crossing path crash type distribution at
candidate intersections in California
|
Crossing Path Type
|
Frequency
|
Percent
|
|
LTAP/LD
|
38
|
12%
|
|
LTAP/OD
|
119
|
38%
|
|
LTIP
|
11
|
4%
|
|
RTIP
|
6
|
2%
|
|
SCP
|
136
|
44%
|
|
Other
|
2
|
1%
|
|
TOTAL
|
312
|
100%
|
Reports of crashes at each of the candidate intersections in
Minnesota from 1998-2000 were reviewed.
The crash narrative and sketch along with reporting officer codes of the
contributing factors were relied on to determine the circumstances involving
each crash.
There were 358 intersection crashes at the 20 candidate
intersections in the three-year period.
The manual review established that 276 of those crashes occurred at the
intersection or were intersection-related.
Of those 276 intersection crashes, 130 (47 percent) were crossing path
crashes in the intersection. The
frequency of total intersection crashes, crashes within the intersection
(non-crossing path), and crossing path crashes for each candidate intersection
in Minnesota are displayed in Appendix B.
The distribution of the crossing path crash types at the 20
candidate intersections in Minnesota is provided in Table 6. Similar to
California, SCP crashes were the predominant crash type, followed by the
LTAP/OD crashes.
Table
6: Crossing path crash type distribution at
candidate intersections in Minnesota
|
Crossing Path Type
|
Frequency
|
Percent
|
|
LTAP/LD
|
10
|
8%
|
|
LTAP/OD
|
34
|
26%
|
|
LTIP
|
7
|
5%
|
|
RTIP
|
3
|
2%
|
|
SCP
|
73
|
56%
|
|
Other
|
3
|
2%
|
|
TOTAL
|
130
|
100%
|
Reports of crashes occurring at each of the candidate
intersections in Virginia between 1998 and 2000 were reviewed. The crash narrative along with the reporting
officer codes of the driver’s action in the crash, driver vision obstructions,
conditions of drivers and pedestrians, and vehicle condition were relied on to
determine the circumstances involving each crash. In the three-year period there were 1,005 crashes recorded as
occurring at the 20 candidate intersections.
Virginia defines intersection crashes as any crash occurring in the
intersection or on the approach within 0.03 miles of the intersection. The manual review established that 783 of
these crashes were intersection-related.
Of these 783 intersection-related crashes, 537 (68.5 percent) were
crossing path crashes. The frequency of
total intersection crashes, crashes within the intersection (non-crossing
path), and crossing path crashes for each candidate intersection in Virginia
are displayed in Appendix B. The
distribution of the crossing path crash types at the 20 candidate intersections
in Virginia is provided in Table
7. The LTAP/OD
crash was the predominant crash type, followed by the SCP crash type.
Table
7: Crossing path crash type distribution at
candidate intersections in Virginia
|
Crossing Path Type
|
Frequency
|
Percent
|
|
LTAP/LD
|
70
|
13%
|
|
LTAP/OD
|
244
|
45%
|
|
LTIP
|
5
|
1%
|
|
RTIP
|
29
|
5%
|
|
SCP
|
177
|
33%
|
|
Other
|
12
|
2%
|
|
TOTAL
|
537
|
100%
|
Field reviews were conducted at the candidate intersections.
Each was diagrammed, photographed, and in some cases, videotaped. For each intersection, the field reviewer
noted the traffic control device and its operation, any sight obstructions,
parking restrictions, lane usage, adjacent land use, channelizations, turning
restrictions, adjacent driveways, or any other factors that were relevant to
the operation of the intersection.
Crash diagrams were created for each of the candidate
intersections based on the field reviews, aerial photographs, and crash
reports. The diagrams illustrate the
types of crashes that occurred at the intersection and the maximum injury
sustained in each crash. Only crossing
path crashes were included in the crash diagrams. Based on the crash diagrams, the critical crossing path scenario
and approach were identified for each of the candidate intersections. The crash diagram for each candidate
intersection is included in Appendix F.
The total set of crashes were examined to determine if
specific crash types were related to certain intersection characteristics.
Table 8 includes the distribution of crossing path crash
types by intersection environment and traffic control for all of the candidate
intersections combined.
