Report No. SR92-02-01
Design and Operation of an
Instrumented "Chase Car" for
Characterizing the Driving
Patterns of Light-Duty
Vehicles in Customer Service
prepared for:
U.S. Environmental Protection Agency
Certification Division
February 28,1992
prepared by:
Sierra Research, Inc.
1521 I Street
Sacramento, California 95814
(916) 444-6666
-------
EPA-420-R-92-104
Design and Operation of
an Instrumented "Chase Car"
for Characterizing the Driving Patterns
of Light-Duty Vehicles in Customer Service
prepared for:
U.S. Environmental Protection Agency
Certification Division Division
Under Contract No. 68-C9-0053
Work Assignments 1-03 and 1-04
February 28, 1992
prepared by:
Thomas C. Austin
Robert G. Dulla
Francis J. DiGenova
Sierra Research, Inc.
1521 I Street
Sacramento, CA 95831
(916) 444-6666
-------
Design and Operation of
an Instrumented "Chase Car"
for Characterizing the Driving Patterns
of Light-Duty Vehicles in Customer Service
Table of Contents
page
1. Summary ... . . . . 1
2. Introduction ... . ... 2
3. Advantages and Limitations of Chase Car . . ... 4
4. Chase Car Specifications and Instrumentation .. 9
5 Field Testing 24
6. Route Selection . ... 29
7. Chase Car Operational Procedures . . .35
8 Data Processing ... . . ... 40
-------
List of Figures
Figure page
1. Chevrolet Caprice Chase Car . ... 10
2 Laser Atlanta Hand-Held Laser Gun .. .... .... 11
3. Custom Laser System Mounted in Chase Car 12
4. Frontal View of Chase Car with Laser Rangefinder Installed . 13
5 Laser Rangefinder Distance Measurement Accuracy .... 13
6 Correlation Between Computed and Measured Longitudinal
Acceleration .... 16
7 8mm Camcorder Installation . 16
8. Switchbox Used by Observer ... .. . 17
9. Illustrations of the U S. DOT Level of Service
Classification System .... 18
10. Chase Car Passenger Compartment Showing Portable Computer
Equipped with Data Acquisition System . 20
11 AC/DC Power Supply System . . .... 22
12. Power Supply System Close-Up 2 3
13 Target Vehicle Speed-Time Trace For Downtovn/Greenhaven/
Downtown Road Route .... 24
14. Target Vehicle Speed—Time Trace Compared to Laser-Based
Estimate . 25
15. Target Vehicle Speed-Time Trace Compared to Individual
Laser Data Points 26
16 Target Vehicle Speed-Time Trace Compared to Chase Car
Speed-Time Trace and Laser-Based Target Vehicle Estimate 26
17 Laser Estimated Speed vs Actual Speed of a Target Vehicle 27
18. Diurnal Variatiaon in Travel for Selected Communities 33
19 Recommended Allocation of Baltimore Trips by Purpose
and Time of Day 34
-ii-
-------
List of Tables
Table page
1 Sample Analog Data File Created With Labtech Notebook .. . 21
2. Proposed Diurnal Allocation of Baltimore Trips 34
-iii-
-------
1. Summary
With funding provided by the U.S Environmental Protection Agency and
the California Air Resources Board, Sierra Research has developed an
instrumented "chase car" that, with the use of a behind the grill
mounted laser rangefinder, is capable of recording major speed changes
of vehicles it follows without maintaining a constant following distance
or matching the acceleration characteristics of the vehicle being
followed The chase car is also equipped with systems designed to
provide second-by-second information on its own speed, longitudinal
acceleration, lateral acceleration, roadway grade, engine load, roadway
type, and level of congestion. By following randomly selected vehicles
traveling along road routes known to be representative of a particular
area, the chase car collects information that characterizes the full
range of operating conditions experienced on public roadways
Initial operation of the chase car was in the Greater Metropolitan Los
Angeles area where approximately 100 routes were driven Over 200
routes were then driven in Baltimore, Maryland During 1992, additional
data collection is anticipated in Los Angeles and possibly other areas
Through analysis of the data collected by the chase car and supplemental
data on vehicle activity at "trip ends" and "soak time" between trips,
the current operational characteristics of light-duty vehicles can be
compared to the operational characteristics embodied in the "LA4"
driving cycle, used in the Federal Test Procedure for light-duty
vehicles since the 1972 model year
After the construction of one or more driving cycles that represent the
operating characteristics recorded by the chase car, tests of a
representative sample of vehicles can be used to determine whether such
cycles produce emissions results that are significantly different from
the emissions produced using the LA4 cycle
###
-1-
-------
2. Introduction
In January 1990, the Mobile Source Division of the California Air
Resources Board (CARB) issued a Request For Proposals for
"Characterization of Driving Patterns and Emissions from Light-Duty
Vehicles in California" The RFP called for the development of new
driving cycles to represent light-duty vehicle travel during peak
morning, peak afternoon, and off-peak periods Also during 1990,
amendments were made to the Clean Air Act (§206(h)) requiring the U S
Environmental Protection Agency (EPA) to "review and revise as necessary
the regulations . regarding the testing of motor vehicles and motor
vehicle engines to insure that vehicles are tested under circumstances
which reflect the actual current driving conditions under which motor
vehicles are used, including conditions relating to fuel, temperature,
acceleration, and altitude "
In response to the CARB RFP, Sierra Research, Inc (Sierra) proposed,
and was eventually awarded a contract, to use an instrumented "chase
car" to collect speed vs time data while following randomly selected
vehicles operating in the Greater Metropolitan Los Angeles Area The
original scope of work under Sierra's contract with CARB called for the
use of a chase car instrumented only to record its own speed-time
profile Accurate characterization of the speed-time profile of
vehicles being followed would have been a function of the ability of the
chase car driver to match their driving patterns.
Having been given the responsibility for conducting the review of the
light-duty driving cycle mandated by the recent Clean Air Act
amendments, EPA's Certification Division issued "work assignments" to
Sierra to supplement the chase car effort being performed for CARB
Under Work Assignment 1-03 of Contract No 68-C9-0053, Sierra was
required to perform an evaluation of data collection methods that could
be employed to record vehicle operation in customer service The
evaluation was to consider instrumentation options for vehicles driven
by randomly selected motorists in addition to instrumentation options
for "chase cars" designed to follow other vehicles In addition to the
conceptual evaluation under Work Assignment 1-03, Work Assignment 1-04
directed Sierra to develop and field test a complete vehicle
Development of a methodology for constructing representative driving
cycles from the chase car data was also required
Following this introductory section, Section 3 summarizes the advantages
and limitations of chase cars and sets forth Sierra's recommended
approach for characterizing light-duty vehicle driving patterns through
the use of chase car data supplemented with data collected from "trip
end" surveys, motorist surveys, and instrumented vehicles being driven
in routine customer service Section 4 describes the chase car
instrumentation package developed by Sierra Section 5 presents the
results of field tests conducted using the chase car to follow another
instrumented vehicle in traffic Section 6 covers chase car operational
-2-
-------
procedures, including the protocol used to determine which vehicles will
be followed. Section 7 describes the route selection methodology being
used to determine where the chase car is driven Section 8 describes
the data analysis technique proposed for translating information
collected by the chase car into driving cycles that represent customer
service.
###
-3-
-------
3. Advantages and Limitations of Chase Cars
Two approaches available for characterizing the operation of vehicles in
customer service are 1) to collect data from instrumented vehicles
driven by a variety of motorists, and 2) to collect data using an
instrumented "chase car" that follows randomly selected motorists The
two driving cycles currently used by EPA and CARB for determining
compliance with emissions and fuel economy standards for light-duty
vehicles are the Urban Dynamometer Driving Schedule, commonly referred
to as the "LAV, and the "Highway" cycle Generally speaking, the LA4
cycle was developed based on the use of data from an instrumented
vehicle operating in customer service and the Highway cycle was
developed based on the use of chase car data
The LA4 was developed using an instrumented vehicle driven over a
specific, 13 mile long road route (the LA4 road route) believed to be
representative of travel in Los Angeles in terms of the fraction of time
spent at various combinations of speed and load Different drivers
drove the route in different elapsed times The data selected for cycle
development was based on the one trip over the route that came the
closest to the average time for all drivers. As a result, portions of
the speed-time trace generated by one particular driver, driving one
particular car, during one particular trip over the route were used to
construct the 7 5 mile cycle that has been used to certify light-duty
vehicles since the 1972 model year * The Highway cycle was developed
using an instrumented chase car that followed other vehicles in traffic
during 1,050 miles of operation over non-urban roadways Using pieces
of the speed-time trace generated by the chase car, a 10 2 mile long
cycle was constructed that matched the average speed, stops per mile,
and major speed deviations per mile for all of the data recorded by the
chase car
As discussed below, neither of the techniques previously used for cycle
development provide assurance that the resultant cycle adequately
represents light-duty vehicle operation in customer service
Practical Considerations
Regardless of the technique used to collect data, there are several
practical considerations that need to be addressed The emissions from
R.E Kruse and T A Huls, "Development of the Federal Urban Driving
Schedule," U S Environmental Protection Agency, SAE paper no 730553,
May 1973
T C Austin, et al, "Passenger Car Fuel Economy During Non-Urban
Driving," U S Environmental Protection Agency, SAE paper no 740592,
August 1974
-4-
-------
passenger cars and light trucks are known to be strongly affected by
operating mode In order for a driving cycle to represent a composite
of operation in customer service, it must encompass a wide range of
operating modes and initial conditions
Cold Start and Warm-Up Effects - Given the vehicle-to-vehicle
variability in cold start and warm-up performance, no single trip length
can possibly represent the average emissions for a particular make and
model of vehicle A related factor is that, given the continuous
distribution of soak times between trips, no single soak time can
possibly be expected to represent the initial conditions of the engine
and emissions control system during all starts The test procedure used
in conjunction with the LA4 cycle attempts to represent the range of
initial conditions by weighting together emissions measured from a
"cold" start (minimum 12 hour soak time) and a "hot" start (10 minute
soak time) Using an estimate of 4 7 total trips per day, two of the
trips are assumed to be from a cold start and 2 7 of the trips are
assumed to be from a hot start Implicit in the test procedure is that
all trips are assumed to be 7 5 miles in length, regardless of whether
they are hot start or cold start. Implicit in the test procedure used
with the Highway cycle is that all non-urban trips are assumed to start
with a warm engine
Speed-Time Profile Effects - Vehicle emissions are known to be strongly
affected by variations in acceleration rate and speed Because traffic
congestion, speed limits, traffic signals, and other roadway
characteristics affect the speed-time profile of vehicles travelling a
particular route, it is apparent that the development of a
representative driving cycle must involve the characterization of
vehicle operation over a wide range of roadway conditions that
collectively represent travel occurring in the area of interest Given
the vehicle-to-vehicle variation in performance and driver—to-driver
variation in "aggressiveness", no single vehicle-driver combination can
be expected to represent the distribution of acceleration rates that
occurs in customer service even if travel over a wide range of traffic
conditions is monitored The need to represent the range of speed-time
profiles is important when emissions that occur infrequently are much
higher than emissions that occur under "typical" conditions For
example, consider the hypothetical case of a particular mode of vehicle
operation that occurs only 0 1% of the time If emissions during this
infrequently occurring mode are 1,000 times higher than during other
modes of operation, then the exclusion of this mode would cause average
emissions to be understated by 50% The overall average emissions of
the vehicle would be twice as high as the emissions that occur 99 9% of
the time
(1 x 999) + (1,000 x 0 001) = 2 0
Given the potential significance of infrequent events, it would be
desirable to know the contribution to emissions in customer service of
as many instantaneous operating conditions as possible With such
information, a cycle could be constructed that not only produces
emissions equal to the true average, but also reflects the specific
operational characteristics that cause the average to be what it is If
-5-
-------
the length of the cycle needed to achieve the proper average is
excessive, weighting factors of less than 1 0 could be considered for
those elements of vehicle operation that cause extraordinarily high
emissions
The LA4 driving cycle contains a speed-time profile that represents how
one particular car was driven by one particular driver during one
particular trip As such, the speed-time profile cannot reasonably be
expected to represent the wide range of speed-time profiles occurring in
customer service The Highway driving cycle incorporates data from
several different trips by a vehicle that followed other vehicles m
traffic, however, the segments of the speed-time trace used in the
development of the cycle were selected based on how well they
represented the average speed for the full data set. In addition, the
Highway cycle was specifically developed to represent non-urban vehicle
operation in areas where the 55 mph speed limit is strictly enforced
Its applicability to the current situation must be seriously questioned
The freeway speed limit has been raised to 65 mph in most non-urban
areas and frequent violations of the 55 mph limit in urban areas are
known to occur As is the case with the LA4, the Highway cycle also
cannot reasonably be expected to represent the wide range of speed-time
profiles occurring in customer service
Roadway Grade — Previous efforts to characterize the operation of
vehicles in customer service have concentrated on measurement of speed-
time profiles Implicit in the analysis of speed-time data and its
subsequent translation into a dynamometer driving cycle has been the
assumption that the effects of roadway grade could be ignored In areas
with rolling terrain, this assumption may lead to significant
differences between the emissions emitted during dynamometer testing and
the emissions actually occurring over the road Because of the non-
linear relationship between vehicle emissions and vehicle load, there
could be significant emissions effects of travel over road routes with
periodic grade changes even though the net grade change is zero The
most obvious effect would be with carbon monoxide emissions during stop-
and-go operation in hilly terrain During uphill accelerations,
vehicles would be more likely to go into power enrichment which, in the
case of vehicles equipped with 3-way catalysts, could cause emissions to
increase by a factor of 100 Lighter average loads during downhill
operation would not produce correspondingly large emission reductions
Data Collection Alternatives
The considerations outlined above have several implications It is
apparent that soak time prior to the beginning of a trip is important
and must be known to determine how emissions during the initial phase of
a trip are being affected by the initial conditions of the engine
Because of the cold start/warm-up effects, average emissions per mile
will also be affected by the length of the trip In order to ensure
representative speed-time profiles, it will be necessary to ensure that
the data being collected from trips are representative in terms of level
of traffic congestion, roadway type, and driver behavior
-6-
-------
As set forth below, alternative data collection methods have different
strengths and weaknesses
Diarv Data - Asking motorists to keep track of their travel behavior in
a written log could conceivably generate a substantial amount of data on
soak times, trip lengths, and average speeds at relatively low cost
However, representativeness of the sample would be an obvious concern
Motorists willing to participate in a diary data collection effort may
not be representative of the full range of motorists Accuracy and
consistency of information obtained in such a manner would be
questionable and detailed speed-time profiles could not be obtained
Questionnaires/Surveys - Asking motorists to provide information on
their driving through the use of questionnaires or surveys would be
expected to suffer from many of the same deficiencies of diary data
collection efforts Representativeness of sample would remain a
potential problem Expecting motorists to accurately remember times and
mileages could be a serious problem As with the case of diary data,
detailed speed-time information would not be available However, the
technique has some promise for obtaining short term data on trip
locations and soak times with reasonable accuracy and low cost, provided
a high response rate can be achieved through the use of sufficient
incentives for participation
Instrumented Vehicles - The use of instrumented vehicles provides the
opportunity to obtain excellent resolution on speed-time profiles, as
well as accurate information on soak times between trips However,
representativeness of the sample remains a concern for several reasons
The only feasible means of obtaining data from instrumented vehicles
without the knowledge of the motorist would probably involve the use of
rental cars or loaner vehicles There would be obvious concerns with
the representativeness of the manner and circumstances under which such
vehicles might be driven By restricting the sample to motorist-owned
vehicles, two other factors might be expected to influence the results
First, motorists who would volunteer to have their vehicle instrumented
may not be representative Second, motorists who are knowingly driving
an instrumented vehicle might have their behavior affected (e g ,
reduced tendency to speed)
Chase Cars - Surreptitiously following vehicles in traffic with an
instrumented "chase car" eliminates some of the concerns with
instrumented vehicles while introducing other concerns The main
advantage of the chase car approach is that the speed-time profile of
other vehicles can be approximated without the need for volunteers who
know they are involved in some sort of experiment The accuracy of the
speed-time data collected depends on the sophistication of the
instrumentation package on the chase car and/or the ability of the chase
car driver to generate the same speed-time trace as the vehicle being
followed The representativeness of the data collected by the chase car
also depends on whether the vehicles to follow are selected on a random
basis and whether the time and location of chase car operation
adequately represents travel in the area However, if representative
routes have been identified, chase cars provide the ability to sample a
relatively large number of vehicle miles travelled per day (compared to
instrumented motorist-owned vehicles)
-7-
-------
Provided that an acceptable method is used to randomly select vehicles
to follow, one limitation of the chase car approach is that certain
vehicles are difficult to follow, especially those that are being driven
in an aggressive or erratic manner Another limitation of the chase car
approach is that it is not amenable to determining soak times between
trips or to capturing "trip ends" A concern, but not necessarily a
significant limitation, is that, depending on the technique being used,
following other vehicles could influence their behavior
Recommended Approach
Based on the considerations outlined above, Sierra recommended that both
CARB and EPA use a multifaceted data collection effort to characterize
light-duty vehicle operation in routine customer service To
characterize speed-time profiles on public roadways, Sierra recommended
using a chase car instrumented (with a radar- or laser-based system) to
measure the speed-time profile of other vehicles without following them
in a close or consistent manner. To ensure representative traffic
conditions, it was recommended that road routes for the chase car to
follow be randomly selected from a validated transportation model after
the routes had been "trip—weighted" To overcome the problem with
recording trip end behavior, Sierra recommended obtaining supplemental
data from a trip end survey involving visual observation of trip origins
and destinations To overcome the problem with knowing soak times
between trips, Sierra recommended collecting survey information from a
random sample of motorists stopped at Highway Patrol roadblocks or
interviewed during refueling operations at service stations To obtain
additional information on soak times and to provide a cross check on the
chase car data, Sierra also recommended that some data be obtained from
instrumented vehicles owned by motorists who volunteer to operate their
vehicle with a retrofitted instrumentation package
In response to these recommendations, Radian Corporation was issued a
work assignment by EPA to develop a simple instrumentation package for
retrofit onto vehicles owned by motorists who volunteer to participate
when they are approached at vehicle inspection and maintenance (I/M)
test facilities in Baltimore, Maryland, and Spokane, Washington CARB
issued a Request for Proposals to have data on instrumented vehicles
collected in California as well Sierra pursued the development of a
more sophisticated chase car instrumentation package than had been
contemplated under the original version of the scope of work for CARB
and continued with its efforts to define representative road routes and
operating protocol for chase car data collection in the Los Angeles,
California and Baltimore, Maryland areas
###
-8-
-------
4. Chase Car Specifications and Instrumentation
Under Sierra's original scope of work for CARB, the concept for the
chase car involved using a vehicle with a relatively high power/weight
ratio that would be instrumented to record road speed once per second
while closely following other vehicles in traffic, especially during
major speed deviations With the availability of additional funding for
chase car development by EPA, it became possible to pursue alternative
concepts for capturing the speed-time profile of other vehicles In
addition, a decision was made to pursue the development of an
instrumentation system for determining roadway grade
Vehicle Selection
Once the decision was made to pursue the development of a system to
measure the relative speed of vehicles being followed without
maintenance of a fixed separation distance, the power/weight ratio of
the chase car became somewhat less important Higher priority was
assigned to identifying vehicles with sufficient room for
instrumentation needed to monitor the speed of the vehicle being
followed An inconspicuous appearance remained a high priority
When the size constraints for the instrumentation package were
identified, it was clear that vehicles with the largest possible space
between the radiator/condenser and the grill were going to be desirable
The Chevrolet Caprice and the Lincoln Continental Town Car were
identified as two vehicles with a large space behind the grill Of
these two, the Caprice provided easier access to the behind the grill
space and substantially lower cost at nearly the same power/weight
ratio The Caprice was also considered to be a somewhat less
conspicuous looking car, especially from the front Because the Caprice
is a popular model for police vehicles, it was ordered with aluminum
wheels and white sidewall tires to ensure that it would not be mistaken
for a police vehicle White was selected as the exterior color for two
reasons It is the most common color for a light-duty vehicle (and
therefore somewhat less conspicuous) and it is the most practical color
for minimizing heat build up in the trunk and passenger compartment To
improve handling and performance, the vehicle was ordered with the
trailer-towing package consisting of higher rate springs and a
numerically higher rear end ratio (3 23 1) A picture of the vehicle is
shown in Figure 1
Laser Rangefinder Development
The initial efforts on alternative instrumentation packages involved an
investigation of a forward—looking, "same lane" radar system to record
the relative speed of a vehicle being followed Police-type radars
-9-
-------
Figure 1
Chevrolet Caprice Chase Car
proved to be inadequate because of accuracy problems at speed
differentials below about 10 mph. This limitation is fundamental to
doppler-based radar systems which operate on the principal of a radar
signal's frequency shift as a result of reflection from a target having
a non-zero relative speed. As the relative speed of the target
approaches zero, so does the frequency shift, rendering more difficult
the task of estimating speed. This problem is compounded, in the
application of interest here, because most commercial doppler radar
designs for law enforcement application are not concerned with small
(e.g., <10 mph) relative speed differences.
One system that initially looked the best was a special purpose radar
system called "VORAD" designed for use as a collision avoidance/smart
highway device. The VORAD system is designed to accurately measure
relative speed differences between the instrumented vehicle and the
vehicle being followed in order to ensure that a safe driving distance
is maintained. Although the VORAD system is not yet commercially
available, several prototype systems were already under evaluation.
Concerns about the radar beam being detected by radar detector equipped
vehicles combined with problems in negotiating the timely availability
of a system for Sierra's use ultimately caused efforts to be focussed on
a laser-based system.
A representative of the California Highway Patrol advised Sierra that he
was aware of at least one laser-based speed measurement system currently
-10-
-------
being developed. Various other communications led to the identification
of Laser Atlanta as a company that already had working models of a hand-
held laser rangefinder-based speed measurement system designed to
compete with hand-held police radar units. By measuring time of flight
of an eye safe, infrared laser pulse reflected off a target, the Laser
Atlanta system could measure the range to a target. By sending out
approximately 400 laser light pulses per second and timing the return
reflections, distance to a static target is measured to within about 1
foot on average. If the return signals indicate a variable range, the
instrument automatically switches to speed-measuring mode and attempts
to determine relative speed in mph. Discontinuities in range
measurements are an indication that the laser beam has moved off of the
target and no speed is computed. When there are no discontinuities, the
laser, after about 1—3 seconds (depending on the relative speed),
indicates "lock on" to a target. Figure 2 shows a picture of the hand-
held system.
Figure 2
Laser Atlanta Hand-Held Laser Gun
During an on-site visit to Laser Atlanta, Sierra was given the
opportunity to use the hand-held system to record the speeds of moving
vehicles. The system locked on to vehicles quickly and returned
relative speed readings that appeared to be accurate. Unfortunately,
the system would not function at ranges of less than about 70 feet. In
addition, the system was not designed to compute speed at very low
relative velocities (as would occur if the system were installed in a
moving vehicle following another vehicle).
Although the system designed for police use was not suitable, Laser
Atlanta expressed an interest in developing a modified version of the
-11-
-------
system for use in Sierra's chase car. By isolating the laser gun from
the receiver, Laser Atlanta believed they could accurately measure
shorter time of flight and make the system function down to a range of 2
feet. By increasing the diameter of the laser beam, it would also be
easier to keep the laser on the back end of a vehicle being followed as
the chase car negotiates hills and curves. Sierra signed an agreement
with Laser Atlanta for the development of a vehicle-mountable system
that would measure distance to within 1 foot over a range of 2-200 feet.
Figure 3 shows the custom laser rangefinder developed by Laser Atlanta
installed in the Caprice. The rectangular box above the two lens
encloses a miniature CCD (charge coupled display) television camera,
which provides a high resolution video signal. The slanted glass cover
of the video camera lens provides for a "head up" display of target
relative speed in the video picture, along with an illuminated reticle
which shows where the laser and camera are aimed. The laser light is
infrared and not visible to the naked eye.
Figure 3
Custom Laser System Mounted in Chase Car
Figure 4 shows a front view of the car with the hood closed. Note that
a small section of grill has been cut out to eliminate any obstruction
of the laser gun, receiver, or camera.
Figure 5 shows the results of Laser
measurement accuracy of the system,
original version of the system were
Atlanta's final test of the distance
Several software changes from the
required after Sierra's test results
-12-
-------
Figure 4
Frontal View of Chase Car With Laser Rangefinder Installed
Figure 5
Laser Rangefinder Distance Measurement Accuracy
DATA TAKEN SEPT. 11. 1991 ©12PM
20 40 60 80 10O 120 140
ACTUAL RANGE (FEET)
~ MEASURED DATA UCL/LCL +/-2.5FT
-13-
-------
indicated larger distance measurement error and problems getting the
system to lock on to moving targets As shown in the figure, distance
measurements are accurate to within about 2 5 feet, except for a minor
discontinuity in the 30-40 foot range The cause of the discontinuity
is thought to be some internal interference that has not been precisely
identified As discussed in more detail in Section 5, the data
filtering method being employed to process laser data will minimize the
effects of the discontinuity Even without filtering, the discontinuity
has a modest effect on the calculated speed/time profile of vehicles
being followed For example, when a vehicle accelerates away from the
chase car through the 30-40 range, an instantaneous overestimate of the
distance between the vehicles of 2 feet would cause a true 6 mph/sec
acceleration (a commonly observed rate) to instantaneously appear to be
7 4 mph/sec, followed by an instantaneous 4 6 mph/sec acceleration rate
during the next second With digital filtering of the laser data, the
effect of distance measurement error can be substantially reduced
Chase Car Speed Measurement
A standard GM pulse—generator-type speed sensor on the transmission
produces a signal whose frequency is proportional to speed A lead
spliced into the wire between the sensor and the speedometer is routed
into the passenger compartment and connected to a custom circuit, based
on a frequency to voltage integrated circuit package, that converts the
frequency signal to a D C voltage (0—5 volts) that is proportional to
speed.
Manifold Air Pressure Measurement
A standard GM manifold air pressure (MAP) sensor produces a D C voltage
(0-5 volts) that is proportional to pressure Output of the sensor is
directed into the passenger compartment through leads spliced into the
connection between the MAP sensor and the vehicles ECU
Road Grade Measurement System
An additional instrumentation feature of the chase car is a Sierra-
designed roadway grade and acceleration measurement system The system
consists of two Lucas NovaSensor unidirectional accelerometers mounted
perpendicular to one another and oriented to record acceleration in the
plane of the vehicle floor pan One accelerometer is aligned with the
longitudinal centerline of the car and the other is aligned in the
lateral direction During operation on level roadways, the lateral
accelerometer indicates when the chase car is turning and the
longitudinal accelerometer produces a signal that is approximately
proportional to the rate of change in speed measured by the pulse
generator on the output shaft of the transmission When the vehicle is
not on a level road, the difference between the rate of change in speed
in the longitudinal direction (measured by the transducer on the
transmission) and the longitudinal acceleration measured by the
longitudinal accelerometer indicates the roadway grade
-14-
-------
The way that the system works can be understood by considering what
happens when the nose of the vehicle is elevated When the vehicle is
not moving, both accelerometers are calibrated to read 0 ±0 01 g on a
level surface If the vehicle was lifted by its front bumper, and hung
vertically, the longitudinal accelerometer would read 1 0 g If the
vehicle was lifted by its rear bumper, and hung vertically, the
longitudinal accelerometer would read -1.0 g When the vehicle is
moving along a non-zero grade roadway at a constant speed, the
longitudinal acceleration indicated by the accelerometer is equal to the
sine of the roadway angle from horizontal (eg, on a 30° angle, the
accelerometer would read 0 5 g) As noted above, when the vehicle is
accelerating or decelerating, the difference between the rate of change
in speed in the longitudinal direction (measured by the first derivative
of the signal from the speed transducer on the transmission) and the
total acceleration measured by the longitudinal accelerometer indicates
acceleration due to the roadway grade, i e, the component in the plane
of the roadway of the acceleration due to gravity
Results reported by the manufacturer of the "G-Analyst" accelerometer
system (a system that was evaluated and rejected) indicate that pitch
response for passenger cars varies from about 1 0—2 5 degrees per g,
depending on the stiffness of the suspension Because the maximum
acceleration/deceleration rates for the chase car seldom exceed 0 5 g, a
maximum pitch response of about 1 degree is expected (In terms of
accelerometer readings, this translates to only about 0 01 g of
acceleration in the plane of the roadway )
Figure 6 illustrates the correlation between the acceleration calculated
from the vehicle speed sensor output compared to the acceleration
measured by the longitudinally oriented accelerometer when the vehicle
is driven on a known approximately level road in downtown Sacramento,
California As the figure indicates, there is some "noise" in the data
due to vibration but there is, nevertheless, excellent correlation
between the two By elevating one end of the vehicle, the
accelerometers were able to predict the grade within 1%.
Visual Observations
The chase car was equipped with two independent systems for recording
information regarding roadway type, traffic congestion, and type of
vehicle being followed Figure 7 shows an 8mm camcorder installed
between the front and rear seats The infrared focussing system used on
this particular model (Sony CCD-F70) was able to focus on traffic in
front of the vehicle instead of the vehicle windshield By using a wide
angle conversion lens (0 7x), the effective focal length of the lens was
changed to 6 mm Although this is still not adequate to cover the full
width of the windshield, it provides a reasonable view of traffic in
adjacent lanes The camcorder is powered by an AC adaptor, plugged into
the power supply system described below
-15-
-------
Figure 6
Correlation Between Computed and Measured Longitudinal Acceleration
0.3
0.2
^ 0.1
G
O
co
ju o
o
o
o
<
C3 —0.1
c
•a
p
t -0-2
.3
-0.3
-0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
On-road Acceleration (g's)
r2=0.987 Y=0.00454 + 1.0030(X) 168 observations
Both sets of data are 3 second averages.
1 1
L_
H
K
B
pi
.. *•
H
¦ ¦¦
i ' i
Figure 7
8mm Camcorder Installation
-16-
-------
Manually recorded observations are obtained through the use of a switch
box with 4 rotary switches and 1 toggle switch to classify and record
conditions as they change. Figure 8 shows a picture of the switchbox in
its initial configuration. Each rotary switch has seven positions and
produces a unique voltage for each switch position. The purpose of each
switch is described below:
• Road Type - used to indicate diamond lane, freeway,
on/off ramp, arterial/collector, local, private and
other.
• Level of Service - used to describe six separate levels
of traffic density (passenger cars/mile/lane): A through
F. An additional switch position is provided to
characterize any "other" conditions (such as extreme
congestion caused by a major traffic accident). Figure
9 shows pictorial representations of each level of
service copied from a U.S. Department of Transporation
publication. (These pictures are mounted on the chase
car dashboard for reference by the observer.)
Figure 8
Switchbox Used by Observer
LA.NI
'"EFiVAr
0*/OfF UP
AflTIfliAi. /
COUf CTOfl
WWVATI
TARGET VEHICLE
HJ-PEA/CA*
^ orvirp CAfl
NOTE i
NOTE 4
MO
NOTE*
OTXEfl
NOME
target
-17-
-------
Figure 9
Illustrations of the U.S. DOT
Level of Service Classification Scheme
-------
• Target Vehicle - used to indicate the type of vehicle
being followed. The seven selections shown in the
figure are high performance car, other car, old car,
light-duty truck 1, light-duty truck 2, other and none
This rotary switch can only produce a signal when the
toggle switch is on, indicating the laser is aimed at a
target vehicle
• Notes - a total of 6 "flags" are available to record the
occurrence of an unusual condition.
Data Acquisition System
As outlined above, there are, in addition to the video recording, ten
different data streams being generated and stored during routine
operation of the chase car
• chase car speed,
• chase car manifold air pressure,
• chase car lateral acceleration,
• chase car longitudinal acceleration,
• roadway type,
• level of service,
• target vehicle type,
• target vehicle range,
• target vehicle speed, and
t a series of "flags"
To sample, digitize and record these data streams at least once each
second, Sierra installed a Metrabyte model DAS-8 data acquisition system
in an IBM-compatible portable computer The system accepts up to 8
analog inputs and one RS-232 input Digitized analog data and RS-232
data outputs are controlled by a program called Labtech Notebook
Because the Metrabyte system requires a computer with an IBM-compatible
expansion slot, the selection of computers was limited Three
alternatives identified by Sierra were the Dell 316LT (80386 processor),
Epson 286LTe, and Packmtell model LA3540 286 Each was equipped with a
20-40 MB internal hard disk and a 1 44 MB diskette drive As shown in
Figure 10, the computers were mounted on a foam pad placed in the center
of the front seat (The monitor shown in the picture displays the image
produced by the laser-mounted camera, allowing the observer to determine
when the laser is on target )
During several weeks of shakedown testing, a variety of power supply
failures, screen failures, and system crashes occurred using the Epson
and the Dell onboard the Caprice After resolution of an initial
problem of the computer not retaining the date and time when shut off,
the Packmtell functioned without failure for over 1,000 miles of on
road operation
For future data collection efforts, Sierra is planning to further modify
the Packmtell computer to increase its reliability Because it is one
-19-
-------
Figure 10
Chase Car Passenger Compartment
Showing Portable Computer Equipped With
Data Acquisition System
of the few portable computers with two expansion slots, it is possible
to install a solid state disk emulator (in addition to the data
acquisition card) that uses battery backed-up memory chips. After
booting the computer from the hard disk and starting up the data
acquisition program, data can be written to the disk emulator and no
reads of or writes to the hard disk will be required while the vehicle
is in motion.
The software is configured to sample and store data from each of the ten
inputs once per second. Two date-coded data files are created with a
time stamp on each entry; one for the analog data and one for the laser
data received over the RS-232 link. Table 1 is an example of what the
data file looks like for the eight channels of analog data. The file
header identifies the type of data being stored in each column.
Although data is stored once per second, the sampling, averaging, and
scaling done prior to data storage depends on the input being monitored.
In the case of the accelerometers, it was determined by trial and error
that an average of ten samples each second was needed to damp out noise
created by the action of the vehicles suspension system as it traverses
roadway irregularities. A scale factor and an offset is also used so
that the data from the accelerometers are stored in "g's". In the case
of the other outputs, the value stored is the last value read, often
-20-
-------
Table 1
Sample Analog Data File Created With Labtech Notebook
"Sierra Research"
"Caprice"
"The time Is 12:07:22.20."
"The date is 09-30-1991."
"AVG LCMC" "AVG LAT" "SPEED" "MAP" "Sl»1" "SUl»2" "SU#3" "Sl«4" "DURATION" "cuntime" "time"
j'a's" I'a's" "HPH" "" "POSITION" "POSITION" "POSITION" "POSITION" "SECS" "sees" ""
0.0258
-0.0492
-0.1
36.7
2 1
1 0
0.0325
-0.0357
-0.1
35.7
2 1
1 0
0.0303
-0.0369
-0.2
36.0
2 1
1 0
0.0270
-0.0365
-0.0
36.0
Z 1
1 0
0.0292
-0.0367
-0.2
36.5
2 1
1 0
0.0301
-0.0369
-0.1
36.3
2 1
1 0
0.0280
-0.0392
-0.1
37.4
2 1
1 0
0.0305
-0.0336
0.1
38.1
2 1
1 0
0.0305
-0.0416
-0.0
37.0
2 1
1 0
0.0509
-0.0420
-0.2
38.0
2 1
1 0
0 0761
-0.0408
0.1
45.4
2 1
1 0
0.1715
-0.0344
4.5
50.9
2 1
1 0
0.1686
-0.0431
6.1
56.7
2 1
1 0
0.1670
-0.0256
8.1
54.1
2 1
1 0
0.1514
-0.0363
10.5
49.8
2 1
1 0
0.1281
-0.0183
12 3
56.8
2 1
1 0
0.1095
-0.0351
14.2
46.4
2 1
1 0
0.0446
-0.0525
15.2
47.0
2 1
1 0
0.0800
-0.0234
16.5
51.0
2 1
1 0
0.0830
0.0069
17.2
52.1
2 1
1 0
0
1
2
3
4
5
6
7
a
9
10
11
12
13
14
15
16
17
18
19
43645
43646
43647
43648
43649
43650
43651
43652
43653
43654
43655
43656
43657
43658
43659
43660
43661
43662
43663
43664
00:12:
00-12:
00:12:
00:12:
00:12
00:12
00.12
00.12
00:12
00:12
00-12
00:12
00:12
00:12
00:12
00:12
00:12
00:12
00-12
00:12
:07:24.539
:07-25.539
:07-26.539
:07:27.539
07:28.539
:07-29.539
:07-30.539
:07:31.539
:07-32.539
:07:33.539
.07 34.539
:C7 35.539
:07.36.539
:07:37.539
:07:38.539
07:39.539
:07:40.539
1:07.61.539
:07:42.539
:07:43.539
adjusted by a scale factor For example, the speed output is calibrated
to record and display the speed in mph
Power Supply System
During shakedown testing of the instrumentation system, some variations
in sensor readings obtained through the data acquisition system were
observed as the load changed on the vehicle's electrical system It is
suspected that this was caused by variations in the reference voltages
supplied to the data acquisition system by the computer power supply.
Measurements of the voltage avai-lable from the vehicle's electrical
system showed variations within the range' of 12-14 volts depending on
electrical system load and whether the battery was being charged by the
alternator
To provide a stable source of DC power for the computer, a supplemental
power supply system was installed in the vehicle's trunk. A 12 volt,
deep cycle marine battery was installed in the trunk and connected to
the vehicle's alternator through an underhood mounted isolator The
marine battery was configured to power a 1000 watt, frequency
compensated DC-to-AC invertor (Tripp-Lite model PV-1000 FC) Connected
to the invertor was a 12 volt/10 amp regulated power supply (Tripp-Lite
model PR-lOb) Using this system, DC voltage to the computer is
maintained at close to 12 8 volts regardless of the load on the
vehicle's electrical system, and irrespective of whether or not the
vehicle's engine is running In addition, the system is used to provide
a more stable voltage for the laser rangefinder 110 volt output is
used to power the 8mm camcorder Figure 11 shows the orientation of the
-21-
-------
Figure 11
AC/DC Power Supply System
marine battery, invertor, and 12 volt power supply in the trunk of the
Caprice. Figure 12 is a close up of the invertor and the 12 volt power
supply.
By monitoring the voltage of the marine battery at the beginning of
every day, it was determined that the vehicle's alternator is capable of
maintaining a full charge while the invertor and regulated power supply
are used during routine operation.
Miscellaneous
Several miscellaneous features have been incorporated in the chase car
to improve the overall safety and efficiency of its operation. A
transportable cellular phone is installed to enable the crew to maintain
communications with the office. This feature has proven useful in
resolving minor equipment problems or in resolving questions regarding
road routes. To minimize the possibility that the chase car is mistaken
for a law enforcement vehicle, no external antenna is used with the
phone.
To further minimize the possibility that passing motorists notice either
the camcorder or other equipment, all windows of the vehicle behind the
front doors have been tinted to achieve 80% light extinction. This type
of tinting is not uncommon on California vehicles and it makes it very
-22-
-------
Figure 12
Power Supply System
Close-Up
difficult for passing motorists to see into the vehicle from the rear or
to notice the silhouette of the camera from the front.
Equipment carried onboard the vehicle includes miscellaneous hand tools,
electrical repair equipment, a multimeter, a spare computer and data
acquisition system, a spare switchbox and a fire extinguisher.
###
-23-
-------
5. Field Testing
To evaluate the effectiveness of the chase car instrumentation system, a
second vehicle, a Chevrolet Lumina, was equipped with a similar speed,
MAP, and acceleration measurement system, but no laser The Caprice
chase car was then used to follow the Lumina in traffic with the laser
system activated
Figure 13 shows the speed—time trace generated by the Lumina over one of
the standard road routes used by Sierra in Sacramento, a common
commuting route for individuals who work in downtown Sacramento The
route starts at Sierra's office in downtown Sacramento, goes south on
Interstate 5 to the Greenhaven/Pocket residential area, and then returns
to downtown along the same route run in the opposite direction The
particular trip shown in Figure 13 having been made during off-peak
conditions, the first and second halves of the speed-time trace are
almost mirror images of one another.
Figure 13
Target Vehicle Speed-Time Trace
For Downtown/Greer.haven/Downtown Road Route
Time in seconds
Target is an instrumented Chevrolet Lumina.
-24-
-------
Figure 14 shows a portion of the trip during which the chase car was
able to maintain an almost continuous laser lock on the Lumina One of
the speed-time traces was generated by the Lumina and the other is the
speed—time trace of the chase car adjusted to account for the relative
speed difference measured by the laser. Differences in the traces in
the 60 mph range were subsequently determined to be caused by
calibration differences of the speed sensing circuitry (The Lumina
speed measurement was high by approximately 2 mph at 60 mph. Adjusting
for this calibration error, the laser estimated speed of the Lumina is
more accurate than indicated by the raw data.) The trace based on the
laser data was generated using the Savitzky-Golay digital filter
(described in Section 8)
Figure 15 is an enlargement of the portion of the trace shown in Figure
14 The individual data points are estimates of Lumina speed based on
the laser data collected by the chase car and the solid line is the
speed-time trace recorded onboard the Lumina Although estimate errors
greater than 1 mph are apparent at some points, major speed deviations
appear to be estimated with reasonable accuracy Figure 16 shows
Figure 14
Target Vehicle Speed-Time Trace
Compared to Laser-Based Estimate
70
CO
c
¦a
60
50
w
20
10
0
100
200
300
400
Time in seconds
500
600
700
-25-
-------
Figure 15
Target Vehicle Speed-Time Trace
Compared to Individual Laser Data Points
Time in seconds
Figure 16
Target Vehicle Speed-Time Trace
Compared to Chase Car Speed-Time Trace
and Laser-Based Target Vehicle Estimate
Time in seconds
Plain solid lane is on-board measured speed of target car (Lumina)
Line with diamonds is speed of Caprice. Line with triangles is laser estimate of target car speed
-26-
-------
similar data for a higher speed portion of the trace with the Lumina
speed-time trace adjusted slightly to account for the calibration error
of its speedometer In addition, Figure 16 shows the speed-time trace
of the chase car itself, unadjusted by the laser data As the figure
shows, the laser data made it possible to estimate the speed of the
target vehicle within 1—2 mph while the chase car speed deviated from
the target vehicle speed by much more (The difference between the
chase car speed and the target vehicle speed during this portion of the
trip was associated with the driver of the chase car speeding up to
prevent a third vehicle from pulling in behind the target vehicle and
then slowing down to restore a more comfortable following distance.)
Figure 17 shows the overall correlation between the laser-based
estimates of target vehicle speed and the actual target vehicle speed as
measured by the target vehicle speed sensor
Figure 17
Laser Estimated Speed vs Actual Speed
of a Target Vehicle
20 30 40 50 60
On-board measurement of target speed in mph
Target is an instrumented Chevrolet Lumina.
Laser is installed in an instrumented Chevrolet Caprice which is following the Lumina.
The conclusion drawn from the data collected was that the laser-based
system is capable of measuring major speed changes of a target vehicle
with reasonable accuracy, but incapable of accurately representing the
high frequency, minor speed deviations of the target vehicle This
limitation of the system appears to be due to the design limit on
distance measurement resolution associated with the fact that the system
records all laser pulse time-of-flight measurements in discreet "bins"
-27-
-------
that are about 2 feet in width. Notwithstanding this limitation, the
laser—based system does offer the ability to capture major speed
deviations without the need for the chase car to follow the target
vehicle in a close and consistent manner in order to duplicate its
acceleration or deceleration.
###
-28-
-------
6. Route Selection
As of the preparation of this report, funding has been established for
data collection in Los Angeles, California (CARB funding) and Baltimore,
Maryland (EPA funding). The methodology used for the initial selection
of representative road routes in the Los Angeles area was somewhat
different from the improved methodology ultimately used to select
representative routes in Baltimore In both Los Angeles and Baltimore,
routes were selected to represent typical weekday operation. Weekends,
holidays, and days with significant rainfall were avoided. From
personal observations, it is obvious that travel under such conditions
is significantly different Resource and tiine constraints do not make
it possible to obtain a statistically robust data set for each of these
conditions; therefore, the decision was made to concentrate on typical
weekday travel, clearly the highest priority category. In addition,
weekday travel without precipitation is clearly associated with the most
significant air pollution episodes.
Overview of Initial Los Angeles Methodology
In order tc select the initial routes that the chase c£.r followed,
Sierra obtained the current trip generation matrices from the Urban
Transportation Planning System (UTPS) model employed by the Southern
California Association of Governments (SCAG) to track travel activity in
the Greater Metropolitan Los Angeles Area The current UTPS model for
the Los Angeles area is configured to estimate travel for three separate
periods of operation during a typical week day in 1985 the a m peak
(6 30-8 30 am ), the p m peak (3 30-6.30 p.m. ) and the off-peak (the
other 19 hours of the day). Each period of operation has a separate
trip generation matrix The trip generation matrix specifies the number
of trips that occur both between and within all of the 1,655 Traffic
Analysis Zones (TAZs) that make up the modeling domain of the South
Coast The matrix lists the number of trips that take place between
origin and destination zones (interzonal trips) and within the same zone
(intrazonal trips) All interzonal trips begin and end at a centroid,
the population-weighted center of a TAZ
SCAG developed an estimate of the typical route that would be followed
between each origin-destination pair (OD pair) and included that trip
distance for each OD pair contained in the trip generation matrix
Sierra stratified the OD pairs by trip length before selecting a
representative sample of trips (OD pairs) All of the OD pairs were
ranked by trip length and then sorted into 100 bins with an equal number
of CD pairs The most frequent and second most frequent OD pair within
each bin was selected for each travel period of the day In all, 600 OD
pairs (200 each for each of the three periods of operation) were
selected to represent travel in the Los Angeles area To put this
number in perspective, it should be noted that the three trip generation
matrices contain information on more than 28 million trips
-29-
-------
This methodology ensured that the average length of the selected trips
was equal to the average length of all trips contained in the trip
generation matrix It also ensured that the distribution of selected
trip lengths correctly represented the distribution of trip lengths
contained in the trip generation matrix. To confirm that the
methodology was successful, the average trip length of the selected
trips was computed and compared to the average for the trip generation
matrix for each period of operation. In each case, the average trip
lengths are essentially identical.
SCAG identified the specific links that the selected trips followed in
the computerized network of road links (intersection to intersection)
that have been coded to represent the South Coast road system Sierra
further translated the node numbers detailing each trip into road routes
on a map with street names About 100 of these routes were then run
with the routes selected by time of day based on the profile of a m
peak, p m peak, and off-peak travel for the region
The above-described routes selected from the OD pairs represent
interzonal travel between TAZs on that portion of the Southern
California roadway network that SCAG coded into the UTPS model. Such
travel is over "collectors" and more heavily traveled roadways Travel
over "local" roadways (eg, within subdivisions), the roads that would
be traversed to get from the centroid onto the road network coded into
the model, is not specifically identified by the UTPS model However,
Sierra selected local roads to be followed between the centroid and the
ending link on the road network Starting and ending points for these
trips were selected based on the results of a recent motorist survey
For example, interzonal trips during weekday mornings usually have an
origin in a residential area and a destination in an employer parking
lot. Based on an on-site survey of each TAZ, Sierra selected specific
locations to begin and end the interzonal trips so that a reasonable
distribution of trip end locations was achieved and so that the overall
length of each trip was not significantly affected.
The routes followed by intrazonal trips (i e., within a specific TAZ)
are also not specified by the UTPS model The model assumes that all
intrazonal trip lengths are equal to one half the diameter of the TAZ in
which they occur In addition, the model associates a certain number of
intrazonal trips occurring with each zone that appears to be roughly
proportional to the number of interzonal trips associated with the zone
(We are not clear on precisely how the number of intrazonal trips is
computed.) The methodology used in selecting the routes of these trips
was to stratify all intrazonal trips into bins on the basis of trip
length (i e., one-half TAZ diameter) and to select a random sample of
TAZs from each bin so that the average trip length of the sample matched
the average of all TAZs
Origins and destinations for intrazonal trips will more frequently
involve residence-to-residence travel or travel between residences and
shopping locations As in the case of the interzonal trips, the
specific locations to begin and end each trip were selected so that the
overall distribution of trips by trip end location matches the available
survey data SCAG estimates that intrazonal trips make up roughly 10%
of all trips in the Los Angeles area. Because SCAG uses an estimate
-30-
-------
that the average intrazonal trip length is one-half the diameter of the
TAZ in which it occurs within each selected TAZ, the chase car crew was
directed to drive one-third very short trips (minimum 2 blocks), one-
third trips approximately equal in length to one-half the zone diameter,
and one—third trips nearly as long as the zone diameter. The number of
intrazonal routes run was about 10% of the total intrazonal plus
interzonal trips
Overview of Baltimore Methodology
Sierra obtained the trip generation matrices from the MinUTP
transportation planning model employed by the Baltimore Regional Council
of Governments That model, however, has not been configured to provide
travel estimates for different periods of time (i.e., peak versus off-
peak) . Instead, it provides travel estimates for a 24-hour period
during a typical weekday An additional distinction between the two
communities is that SCAG tracks the daily trip productions and
attractions for five trip purposes, whereas, Baltimore only tracks three
trip purposes
SCAG
- Home-Based Work Trips
- Home-Based Shopping Trips
- Home-Based Other Trips
- Non-Home-Based Work Trips
- Non-Home-Based Non-Work Trips
Baltimore
- Home-Based Work Trips
- Home-Based Non—Work Trips
- Non-Home-Based Trips
The three trip generation matrices provided by SCAG did not distinguish
among trip types That is because the UTPS model cannot distinguish
among trip types after they have been assigned* to the road network
The trips must be assigned in order to determine the distance between
origin and destination zones In contrast, Baltimore provided separate
trip generation matrices for each trip type Evidently, the MinUTP
model is capable of tracking trip types after the assignment process
Despite the differences in the trip generation matrices (i.e., trip type
versus time of day), the initial methodology used to select trips in
Baltimore was essentially the same as the Los Angeles methodology The
The assignment process determines the specific routes that trips will
follow through the road network The methodology used to select routes
is designed to minimize the distance and time required to get from the
origin zone to the destination zone
-31-
-------
trip generation matrix was ranked by trip length and trips were selected
so that the distribution of selected trip lengths represented the
distribution of trip lengths contained within the matrix This approach
ensured that the average trip length of the selected trips equaled the
average of all trips Using that methodology, a total of 600 trips was
selected from the more than 6 million trips contained in the trip
generation matrices
MinUTP determined the network links to be followed on all interzonal
trips Sierra translated the node numbers of those links into routes on
maps with road names Upon completion of this exercise, it was apparent
that an inadequate number of trips to and from the downtown area had
been selected Further analysis uncovered two problems with the
original methodology Because the model starts all trips from the
centroid of a TAZ, the previous methodology introduced a bias in favor
of selecting routes from TAZs of larger geographic area As a result,
too few routes were selected from small TAZs, which are the TAZs that
are the most densely populated In addition, the original route
selection methodology failed to include any routes with low trip
frequency In Baltimore, there is such an extreme difference in
population density between the downtown and outlying areas that the
undersampling of routes beginning or ending in the downtown area was
obvious To correct the problem, routes were randomly selected from a
trip-weighted compilation of routes within the modelling domain
Analysis of the random sample confirmed that it still had the proper
average trip length. In addition, the amount of travel associated with
the densely populated downtown area was m the proper proportion (All
future work in Los Angeles will use routes selected using this same
approach )
Unlike SCAG, the Baltimore Regional Council of Governments agreed to
determine the routes followed for the intrazonal trips that Sierra
selected Unlike interzonal trips, there was no estimate available of
the length of intrazonal trips However, estimates of the average trip
duration were available The specific methodology employed to select
intrazonal routes was to stratify the TAZs into bins on the basis of
average intrazonal trip duration and select a random sample from each
bin so that the average of the sample matched the average of all of the
zones
Finally, after all of the routes were selected and mapped, it was
necessary to determine how to distribute them across the hours of the
day. In the South Coast, the time definitions of the periods of
operation aided that process The a m and p m peaks provided enough
detail on when to travel those routes, and Sierra used trip survey data
provided by SCAG to determine the frequency of off-peak travel during
the remaining 19-hour period A similar approach was needed to
distribute travel on a diurnal basis, by trip type, in the Baltimore
area
Unfortunately, the Baltimore Regional Council of Governments have no
information on diurnal travel activity by trip type The only data
available for Baltimore on diurnal travel activity were based on 1990
traffic counts collected by the Maryland Department of Transportation
Those data, however, only track aggregate travel activity
-32-
-------
Given the fact that no local data were available to diurnally distribute
the trips by purpose, a comparison of the aggregate travel trends
observed in Baltimore, the 1984 NPTS, and several communities across the
U S. was made To provide a uniform basis of comparison, the hourly
data reported by several of the communities was aggregated to the hourly
increments reported by the NPTS and is illustrated in Figure 18 The
figure shows that while there is some variation in aggregate travel
activity across the observed communities, that in general the levels of
activity are relatively consistent More importantly, the levels of
activity reported in the NPTS are generally consistent with those
observed in Baltimore
(/)
Q.
"u.
I-
05
+->
O
c
CD
O
*
CD
CL
30
25
20
15
10
5
0
A
EL
to
Figure 18
Diurnal Variation in Travel
for Selected Communities
><
X7I
X/
X/
-------
Figure 19
Recommended Allocation of Baltimore Trips
by Purpose and Time of Day
| Home-based-work
1a.m. 6 a.m. 9 a.m. 1p.m. 4 p.m. 7 p.m. 10 p.m.
to to to to to to to
6 a.m. 9 am. 1p.m. 4 p.m. 7 p.m. 10 p.m. 1a.m.
Hour of Day
hJota on dale, rapcrtsd Ir tin ' 064
National Personal TranspcrtaUar Study
To meet logistical constraints ir, planning and driving routes, some of
the off-peak, hours were aggregated For home-based-work trips, the 1-6
a m off-peak period was retained in order to capture some of the early
morning commute trips The remaining off-peak periods were aggregated
as shown in Table 2
Table 2
Proposed Diurnal Allocation of Baltimore Trips
(Percent of Trips within Trip Category)
Home-Based Work All Other
1
a
m
- 6
a
m.
6
-
6
a
m
- 9
a
m
33
8
9
a
m
- 4
P
m
26
51
4
P
m
- 7
P
m
25
21
7
P
m
- 1
a
m
10
20
100 100
###
-------
7. Chase Car Operational Procedures
The chase car technique being used by Sierra is somewhat different from
the chase car techniques previously used by General Motors and EPA GM
attempted to follow an individual vehicle from the beginning of its trip
to its final destination. EPA attempted to have its drivers flow along
with traffic, passing as many vehicles as passed them Both of the
previously used techniques were subject to criticism. By attempting to
follow another vehicle from trip beginning to trip end, the GM approach
increased the risk that motorists would see that they were being
followed and alter their behavior By design, the approach previously
used by EPA would be insensitive to extremes in driving behavior In
addition, none of the chase cars used in previous studies were capable
of precisely measuring the speed—time profiles of other vehicles on the
road. In addition, these approaches did not account for the possible
influence of road grade.
Because acceleration rates are a significant concern, Sierra is using
GM's approach of following individual vehicles during major speed
changes and accelerations, but then picking up other vehicles to follow
when one vehicle being followed leaves the pre-selected road route or
when there is any indication that motorists sense they are being
followed
Driving Protocol
Trip Beginnings and Ends - Each chase trip begins in or adjacent to the
parking area that is closest to the centroid of the Traffic Analysis
Zone (TAZ) indicated on the map for that particular trip Acceptable
parking areas include private residences, apartment building parking
lots, shopping center parking lots, roadside business parking lots,
service station aprons, and on-street, curbside parking Actual data
recording begins when the chase car first begins moving on a public
street
Each chase trip ends at the parking area, as defined above, that is
closest to the end-point TAZ centroid Data recording ends when the
chase car leaves the public street, or parks along the curb
Standard Technique for Selection of Target Vehicles - Target vehicles
are selected at random from a pool of candidates that are near the
instrumented chase car and travelling in the same direction on the same
route Candidate vehicles include cars and light trucks, except those
pulling trailers and emergency vehicles Motorcycles, buses, and medium
and heavy trucks will be excluded from consideration, as will any
vehicle being driven in an erratic or unpredictable manner, as evidenced
by sudden stops and starts, unsafe speeds, unsafe lane changes, etc
Although it would be desirable to have detailed information on the
speed/time profiles generated by vehicles being driven in an erratic,
-3 5-
-------
unsafe manner, the collection of such data was considered impractical
To get a sense of the possible significance of such vehicles, detailed
review of the video tapes can determine the frequency with which such
vehicles ideally would have been targeted Based on the perceptions of
the chase car crew, such vehicles represented less than 1% of all target
vehicles
The method for selecting the target vehicle at the beginning of a new
roadway link depends on the level of traffic On busy surface streets
with traffic in front of the chase car, the chase car crew finds the
closest white vehicle in front of an imaginary line passing through the
center of the chase car and perpendicular to the direction of travel
For vehicles in the same lane as the chase car, one car length is
subtracted per 10 mph of speed before deciding which white vehicle is
closest After selecting the nearest white vehicle, the chase car moves
into the lane it is travelling in, if not already in that lane The
target vehicle will be the vehicle immediately in front of the chase car
after that maneuver, which vehicle may or may not be the white vehicle
On more lightly travelled surface streets, follow the first candidate
vehicle that the chase car approaches or that passes the chase car
On freeways. begin, if possible, by following the candidate vehicle that
is on the on-ramp immediately in front of the chase car, onto the
freeway and into whatever lane it goes to If there is no candidate
vehicle in front of the chase car on the on-ramp, then the chase car
merges into freeway traffic and selects a target vehicle according to
the standard protocol described above If it becomes necessary to
select another target vehicle, the first candidate vehicle encountered
in the same lane as first used is selected, if it is in laser range If
the first vehicle encountered in the same lane is not a candidate
vehicle, i.e , is not a light-duty vehicle or a light-duty truck, or if
the vehicle is not in laser range, then the target vehicle is selected
using the standard protocol
Although there may be some concern that the focus on white cars will
introduce some sample bias (due to the possible relationship between
driver demographics/aggressiveness and color choice), it is important to
recognize that the technique for selecting target vehicles only uses
white cars to select the lane of traffic to move into, and then only m
certain circumstances White cars are not used to select the lane of
travel on lightly travelled surface streets or on freeway on-ramps or
when the initially selected vehicle has been lost White cars are
primarily used to select the lane of travel on busy surface streets,
where it is less likely that the vehicle in front of the chase car after
the lane change will actually be a white car Subsequent data analysis
will make it possible to determine whether a disproportionate number of
white cars ended up being target vehicles It also should be recognized
that other potentially more random vehicle selection techniques (e g ,
the selection of vehicles based on license plate numbers) proved
impractical in the field
Acquisition of the Target Vehicle — Once the chase car is positioned
behind a candidate vehicle, the vehicle becomes a "target" for the laser
-36-
-------
rangefinder. Target acquisition is indicated by the proper position of
the chase car in the same lane as the target and at a reasonable range
(less than 300 feet) When the target has been acquired, the observer
records the target type using the rotary selector switch and records the
start of chase by turning on the target switch
Following a Target - The chase car driver attempts to remain behind the
target, approximately matching its accelerations and decelerations, to
the extent possible without arousing suspicion by the target driver,
disrupting traffic or creating a safety hazard. The chase car makes
lane changes with the target vehicle only when it appears that the
target vehicle is making a passing maneuver, and not preparing to turn
off the route The distance of the chase car behind the target is to be
not more than about 300 feet and not less than a safe following distance
(approximately one car length for each 10 mph, depending on conditions)
For instance, in free-flowing traffic on wet pavement, maintain a
following distance greater than on dry pavement In very slow stop-and-
go traffic, it is permissible to "tailgate" the target
If, at any time, the target acquisition is lost or is about to be lost
even momentarily, e.g., around a sharp curve or at a sharp grade change,
the observer immediately turns off the target switch, turning it on
again only if it is clear that the target has been reacquired with the
laser range finder
Deselection of a Target - Each selected target is followed as long as
reasonably possible If a target cannot be followed safely through a
lane or speed change, or if it appears to be deviating from the
preplanned route, it is deselected and a new target chosen (As
discussed in Section 8, the potential bias associated with losing
aggressive drivers is addressed during data analysis ) In that case,
the vehicle immediately in front of the target vehicle becomes the new
target if it is a candidate vehicle and if it is in laser range. If
there is no vehicle within laser range when the original target is lost,
then the protocol for initial selection of a target vehicle is employed
The target is also deselected if it stays in a queue of vehicles
apparently waiting to make a turn off of the preplanned route If a
third vehicle comes between the target and the chase car, a new target
is selected using the standard protocol unless the third vehicle moves
away before a new target is identified and acquired, in which case the
original target is reacquired and chase of that target is resumed. The
percentage of target vehicles that are reacquired previous targets will
be estimated during data analysis
If the chase car must change lanes or turn to exit a given roadway in
order to follow a preassigned route, the current target is dropped and a
new target is selected as soon as possible, using the basic target
vehicle selection protocol
If the driver of the target vehicle exhibits erratic behavior, such as
sudden stops or starts, or apparent nervousness about being followed
(e g , by frequent reference to the rear view mirror), this may create a
safety hazard If such erratic behavior or anxiety is detected, the
chase of that vehicle is ended immediately and a new target is selected.
-37-
-------
Chase Car Travel at Other Times - At all times when the chase car is not
following a target, the chase car is driven in a fashion that
approximately matches the general flow of through traffic, i e.,
travelling faster than some vehicles and slower than a similar number of
vehicles (excluding vehicles that are merging right or traveling slowly
in order to exit), consistent with safe driving practices Lane changes
are made by the chase car only as required to acquire a target, to
follow a target, to change routes, to match the general flow of through
traffic, or as required by road or safety considerations.
In the event that a turn or a freeway exit is missed, data recording is
ended immediately. The chase car is driven back to a point on the route
before the miss occurred, previous driving conditions (e g , lane,
speed) are re-established in the correct direction, and data recording
resumes at the point where the miss occurred
Equipment Operating Procedures
A variety of "check lists" are used to assist in achieving consistent
and efficient data collection during routine operation The check lists
are printed on 5"x7" laminated cards that are held together by a clip
installed on the dashboard of the vehicle The applicable card is
rotated to the first position in the stack
The "Start of Day" check list contains the following items
Clean windshield and laser lens if necessary
Plug in cellular phone and turn on
Install computer and connect cables
Install and aim camcorder, confirm lens set to widest angle
Turn on laser, monitor, and camcorder with invertor switch
Insert fresh 8mm tape with date on label and set camcorder
counter to zero
Start computer, check that at least 5 MB of free space remains
on hard disk
Check clock and date on computer and camcorder
Insert fresh 1 44 MB diskette with date label and check for "no
files" and 1 4 MB free
Start Labtech Notebook
Start car, set switch box to "Note 6", run "GO" to start
recording data and check for reasonable readings from all
sensors and laser.
Escape out of Labtech Notebook.
The "Start of Trip" check list contains the following items
Check fuel level
Make Post-it note with turns
Start camcorder
Check camcorder counter reading and read out loud
Display date on camcorder and read out loud and log
Display time on camcorder and read out loud and log
-38-
-------
Read chase car odometer and log
State location; destination and route number
Confirm laser switches on
Confirm computer is running Labtech Notebook and ready to start
recording data.
Confirm switchbox set correctly for start of run with laser
switch set to "OFF"
Run "GO" and tell driver to drive away when traffic is clear
The "Along the Route" check list contains the following items'
Call out laser ON/OFF
Call out vehicle type
Call out roadway type changes
Call out Level of Service changes
Call out speed limit signs and actual speed.
Call out obvious grades
Activate other "Notes" when unusual events occur (e g , forced
detour off route, forced to make unplanned stop for vehicle
related problem, etc )
Activate "Note 6" when problems are significant enough to abort
run.
The "End of Trip" check list contains the following items
Hit escape key to stop data collection and call out
Read time on camcorder out loud and log
Read chase car odometer and log
State location, destination and route number
Stop camcorder
Pop tape out of camcorder and write run number on label, replace
tape if sufficient time remains for next run
"Quit" Notebook
Note size of last two PRN files
Check remaining space on diskette, install fresh diskette if
necessary
Copy last two PRN files to diskette in a drive
Write run number on diskette
Finally, the "End of Day" check list contains the following items
Remove 8mm tape from camcorder and activate write protect tab
Confirm day's data files are on 1 44 MB diskette(s) and
diskette(s) labeled with date and run numbers, reconcile written
log with number of PRN files written to diskette(s)
Reconcile number of trips on diskette label and number of trips
on 8mm tape label.
Turn off computer
Turn off invertor
Disconnect cables and remove computer and camcorder
Turn off and disconnect phone
-39-
-------
8. Data Processing
By the end of each day in the field, all of the data files are copied to
high density (1.44 MB) diskettes. These diskettes are used to transfer
the trip files to Sierra's in-house VAX computer system The trip files
on the portable computer hard disk are retained until the diskettes have
been successfully copied to the VAX. To further reduce the chances that
any data are lost, the original diskettes are archived and files copied
to the VAX system are backed up onto 8mm tape
Trip Data Analysis
An ever expanding SAS program is being used to screen the data recorded
during each trip for obvious errors (e g , laser range discontinuities,
out of range accelerometer measurements, etc ) and compute descriptive
statistics for both the chase car and target vehicles The program is
also designed to digitally filter range measurements from the laser
rangefinder Using the digitally filtered range data, the program
produces a composite trip, defined in terms of speed and time, of the
chase car supplemented with data collected from target vehicles In
computing composite trip statistics, speed estimated for the target
vehicle is substituted for the chase car speed whenever it is available,
however, the program is structured to ensure that accelerations are not
computed across the transition from chase car to target vehicle or
target vehicle to chase car In addition, stops per mile are computed
only from the chase car to avoid counting the same top twice, or missing
a stop during the transition from the chase car speed data to the target
vehicle speed data
The descriptive statistics produced by the program include the
following
• average speed,
• average speed while moving,
• percent idle time,
• stops per mile,
• percent time on various road types,
•^percent time with various traffic levels of service,
; - /• ^distribution of instantaneous acceleration rates (calculated
each second from successive speed measurements),
• distribution of road grade intervals (based on the difference of
total acceleration, measured by accelerometer 10 times per
second and averaged over 3 seconds, and on-road acceleration,
calculated every second by differentiating speed measurements
and averaging over three seconds), and
• PRE (positive kinetic energy of acceleration per mile)
-40-
-------
Where possible, each of the above-listed statistics is computed
separately for total travel by the chase car, travel by the chase car
when it is not following a target vehicle, travel by the chase car when
it is following a target vehicle, total target vehicle travel, and the
composite of the chase car and the target vehicles
In addition to the above-listed statistics, speed-time traces will be
plotted for each route More detailed analyses, including consideration
of differences between different types of target vehicles, are possible
with minor modifications to the software However, the data collection
effort may have to be expanded for such analyses to be worthwhile
Digital Filtering of Laser Ranee Finder Data - The laser range finder
used for the data collection in Baltimore and the initial data
collection in Los Angeles was designed and configured to provide, at
intervals of one second, measurements of the range or distance in feet
to the target vehicle ahead and relative speed of the target vehicle in
mph If there is no target or the range finder is unable to "lock on" a
reading of either range or speed, for whatever reason, the laser
provides a default output of "9999".
Currently, only range data from the laser, together wLth speed data
collected from the instrumented chase car, are used to analyze
kinematics of the target vehicle In particular, the relative speed of
the target vehicle is computed by smoothing and differentiating the
range data for each second through the use of a Savitzky-Golay digital
filter, and adding to that the speed of the chase vehicle to obtain the
absolute speed of the target vehicle A nine-point digital filter is
used with a cubic/quartic polynomial first derivative As described in
the original paper by Savitzky and Golay", this technique is
analytically equivalent to performing a linear least square best—fit cf
a cubic or quadratic polynomial to nine equally-spaced data points (nine
successive range measurements at one second intervals), differentiating
the polynomial, evaluating it at the center of the nine points, and then
successively advancing the time window one second, as with a moving
average
Optimal configuration of the digital filter in the current application,
has required some experimentation A nine-second analysis has been
found to provide the best compromise between a fLlter which is able to
capture the main features of a hard acceleration or deceleration that
may be as brief as 2-3 seconds, while still providing an objective "best
fit" smoothing of discrete range data In addition, optimization in the
current application has required the recognition of a slight delay
(between one-half and one second) between the time the laser measurement
is made and evaluated and when the data are retrieved over an RS232
serial interface line by Labtech Notebook, an on-board data logging
program running on a laptop computer This delay is accommodated by
"Smoothing and Differentiation of Data by Simplified Least Squares
Procedures", Savitzky, A et al, Analytical Chemistry. Vol 36, No 8,
July 1964, p 1627-1639
-hi-
-------
applying the digital filter to successive ranges of data that are
centered on times that precede the second of interest by one second
Quality Control Checks - Before data from a particular road route are
included in the computation of descriptive statistics, the second-by-
second speed data are checked to identify likely problems Acceleration
or deceleration rates in excess of the capabilities of the chase car are
one of the criteria used to flag potential problems Any occurrence of
low speed (0-10 mph) accelerations in excess of 0.4 g or high speed
(>55mph) acceleration in excess of 0.2 g is used to flag a probable
error (Intermediate acceleration rates are used for intermediate
speeds ) Computed target vehicle acceleration rates are flagged if they
exceed 0.6 g in the 0-30 mph range. Lower acceleration rate thresholds
are used at higher speeds with the lowest threshold being 0.2 g for
accelerations occurring in the 90-100 mph range. Deceleration rates for
both the chase car and target vehicles are flagged if they exceed 1 0 g
Video Tape Analysis
Review of the video tapes will be used to double check apparent errors
identified during second-by-second data analysis of acceleration rates
Review of the tape will indicate whether a data stream with out-of-range
acceleration could conceivably have been valid. In addition, the video
tape will be used to determine the reason for any other questionable
characteristics of the speed/time trace for each trip Depending on the
length of time periods with obvious speed data problems, substitute
speed estimates can be computed from available accelerometer data.
However, it is expected that this approach will be limited to periods of
less than 10 seconds of apparent data errors
Review of the video tapes will also be used to determine how often the
chase car was unable to maintain contact with vehicles being driven in
an unsafe or aggressive manner By reviewing a large subset of the
trips, the percentage of the time that target loss is caused by
aggressive or unsafe behavior on the part of the target vehicle driver
can be estimated During this same review process, the number of times
that previous targets are re-acquired can also be determined
###
-42-
------- |