VEHICLE OPERATIONS SURVEY
VOLUME I
PREPARED FOR
COORDINATING RESEARCH COUNCIL, INC.
30 Rockefeller Plaza
New York, New York 10020
AND
ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF AIR PROGRAMS
MOBILE SOURCE POLLUTION CONTROL PROGRAMS
2565 Plymouth Road
Ann Arbor, Michigan 48105
SCOTT RESEARCH LABORATORIES, INC.
P. O. BOX 2416
SAN BERNARDINO, CALIFORNIA 92406
-------
Final Report
Vehicle Operations Survey
CRC APRAC Project No. CAPE-10-68 (1-70)
Volume I
Prepared for
Coordinating Research Council, Inc.
30 Rockefeller Plaza
New York, New York 10020
and
Environmental Protection Agency
Office of Air Programs
Mobile Source Pollution Control Program
2565 Plymouth Road
Ann Arbor, Michigan 48105
December 17, 1971
by
SCOTT RESEARCH LABORATORIES, INC.
2600 Cajon Boulevard, P.O. Box 2416
San Bernardino, California 92406
-------
SRL 2922 13 1271
TABLE OF CONTENTS
1.0 SUMMARY 1-1
2.0 INTRODUCTION 2-1
2.1 PROGRAM OBJECTIVES 2-1
2.2 BACKGROUND INFORMATION 2-1
2.3 SCOPE OF OPERATIONS 2-2
3.0 PROGRAM DESCRIPTION 3-1
3.1 GENERAL APPROACH 3-1
3.1.1 PREPARATORY EFFORTS 3-1
3.1.2 DATA COLLECTION OPERATIONS 3-2
3.1.3 DATA HANDLING 3-3
3.2 INSTRUMENTATION SYSTEMS 3-3
3.2.1 VEHICLES USED FOR COLLECTING DATA 3-4
3.2.2 TRANSDUCERS 3-5
3.2.3 DIGITAL DATA ACQUISITION SYSTEM 3-8
3.2.4 INSTRUMENT POWER SUBSYSTEM 3-12
3.3 SURVEY ROUTE DESIGN 3-19
3.3.1 ROUTE DESIGN OBJECTIVES 3-19
3.3.2 ROUTE DESIGN FACTORS 3-20
3.3.3 ROUTE-MODEL STRUCTURE 3-20
3.3.4 SURVEY ROUTE DESIGN 3-25
3.4 DATA COLLECTION OPERATIONS 3-33
3.4.1 TRAINING OF PERSONNEL 3-33
3.4.2 DATA COLLECTION SCHEDULE 3-38
3.4.3 OPERATING TECHNIQUE 3-38
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922 13 1271
TABLE OF CONTENTS (CONT.)
3.4.4 VEHICLE OPERATIONS SURVEY ON THE LA-4 ROUTE 3-43
3.5 DATA PROCESSING AND ANALYSIS TECHNIQUES 3-46
3.5.1 DATA PROCESSING TECHNIQUES 3-46
3.5.2 DATA ANALYSIS PROCEDURES 3-84
4.0 RESULTS 4~1
4.1 DATA COMPARISONS 4-1
4.1.1 COMPARISON OF BASIC DESCRIPTORS 4-2
4.1.2 MODE COMPOSITION 4-4
4.1.3 CORRELATION OF MATRIX INFORMATION 4-9
4.1.4 DELTA-SQUARE EVALUATION 4-14
4.1.5 SUPPLEMENTAL INFORMATION 4-14
4.2 LOS ANGELES ROUTE REPLICATE EVALUATION 4-25
4.2.1 NORMALIZED TIME IN MODE VARIABILITY 4-25
4.2.2 NORMALIZED MODE FREQUENCY VARIABILITY 4-27
4.2.3 DELTA-SQUARE COMPARISON OF ROUTE REPLICATES 4-27
I
4.2.4 AVERAGE SPEED VARIABILITY 4-27
4.2.5 CORRELATION BETWEEN ROUTE REPLICATE MATRICES AND
THE LOS ANGELES COMPOSITE FOR NORMALIZED TIME IN
MODE DATA 4-27
4.2.6 CORRELATION BETWEEN ROUTE REPLICATE MATRICES AND
THE LOS ANGELES COMPOSITE MATRIX FOR NORMALIZED
MODE FREQUENCY 4_27
4.2.7 SUMMARY OF ROUTE REPLICATE EVALUATION 4-32
4.3 DETROIT ROAD ROUTE 4_32
4.3.1 DESIGN OF THE ROUTE METROPOLITAN ROAD ROUTE 4-32
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922 13 1271
TABLE OF CONTENTS (CONT.)
4.3.2 OPERATING PROCEDURES 4-35
4.3.3 RESULTS 4-35
4.4 SUMMARY OF RESULTS 4-42
REFERENCES R-l
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
LIST OF TABLES
Page No.
Table 3-1 Typical Route Model Structure for Weekdays 3-23
Table 3-2 Correlation of Trip Length Distribution - Weekday
Routes 3-26
Table 3-3 Correlation of Trip Length Distribution - Weekend
Routes 3-27
Table 3-4 Survey Route Design Characteristics 3-28
Table 3-5 Data Collection Schedule - Vehicle Operations Survey 3-39
Table 3-6 Differences Between LA Composite and LA-4 Survey
Parameters 3-45
Table 4-1 Comparison of Basic Descriptors of Vehicle Operation 4-3
Table 4-2 Distribution of Total Time in Mode Categories 4-6
Table 4-3 Distribution of Mode Frequency in Mode Categories 4-8
Table 4-4 Correlation Coefficients Normalized Time in Mode
Comparison 4-10
Table 4-5 Correlation Coefficients Normalized Frequency of
Occurrence Comparison 4-12
Table 4-6 Delta-Square Values for Each City Versus the
5-City Composite 4-15
Table 4-7 Overall Matrix Similarity A-Square Values 4-16
Table 4-8 Mode Composition for Route Replicates in Los Angeles -
Percent of Time in Mode 4-26
Table 4-9 Mode Composition for Route Replicates in Los Angeles-
Percent of Mode Frequency 4-28
Table 4-10 Delta-Square Values and Average Speeds for Route
Replicates in Los Angeles 4-29
Table 4-11 Correlation of Route Replicates with LA Composite
Matrix: Normalized Time-in-Mode 4-30
Table 4-12 Correlation of Route Replicates with LA Composite:
Normalized Mode Frequency 4-31
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
LIST OF TABLES (CONT.)
Page No.
Table 4-13 Summary of Route Replicate Evaluation 4-33
Table 4-14 Construction Characteristics of Detroit (Ann Arbor)
Road Routes 4-36
Table 4-15 Comparison of Vehicle Operating Patterns on Detroit
Road Routes with the Five-City Composite Results 4-39
Table 4-16 Improvement of Detroit Road Route Characteristics by
Adding Idle Time 4-41
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
LIST OF FIGURES
Page No.
Figure 3-1 Functional Block Diagrams, VOS Mobile Instrument
System 3-6
Figure 3-2 Instrumentation Component Locations, VOS Vehicle 3-9
Figure 3-3 VOS Control Display Panel 3-11
Figure 3-4 Control/Display Panel and Power Control Unit
Installed on the VOS Vehicle Dash 3-13
Figure 3-5 Data Acquisition Unit, Magnetic Tape Recorders,
and 115 VAC Inverter Mounted in VOS Vehicle Rear
Deck (a) Accessibility of Electronic Circuit Cords
(b) Ready for Survey Operations 3-14
Figure 3-6 Rear of Instrument Racks Containing the Data
Acquisition Unit, Magnetic Tape Recorder, and
Inverter (a) Ready for Survey Operations
(b) Access to System Cables and Batteries 3-15
Figure 3-7 VOS Power Control/Monitor Unit 3-17
Figure 3-8 Instrument System Alternator and Regulator Instal-
lation in a VOS Vehicle 3-18
Figure 3-9 Relationships Between Factors Affecting Route Design 3-21
Figure 3-10 General Form of Trip Descriptor Functional Relation-
ships 3-22
Figure 3-11 Distribution of Trips by Time of Day and Purpose of
Trip (Weekdays) 3-29
Figure 3-12 Route Directory Sample Page 3-31
Figure 3-13 Speed-Time Traces From Lead Car and Chase Car 3-35
Figure 3-14 Vehicle Make Code 3-36
Figure 3-15 Weather Codes 3-37
Figure 3-16 LA-4 Road Route Description 3-44
Figure 3-17 Mode-Selection Logic 3-50
Figure 3-18 Segmentation of Speed/Time Data Into Modes Using
Preliminary Mode Logic 3-52
Figure 3-19 Matrix Representation of Vehicle Operation 3-54
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
LIST OF FIGURES (CONT.)
Page No.
Figure 3-20 Master Mode Matrix 3~55
Figure 3-21 Vehicle Operation Mode Matrices 3-57
Figure 3-22 Transition Probability Matrix 3-58
Figure 3-23 Matrix of Average Speed Versus Trip Length and
Time of Day 3-60
Figure 3-24 Acceleration-Rate-Versus-Instantaneous-Speed Matrix 3-62
Figure 3-25 Manifold Vacuum Versus Cruise Speed 3-64
Figure 3-26 Compact Matrix Grid Overlaid on 70 x 70 Mode Matrix
Grid Structure 3-68
Figure 3-27 Modified Logic for Segmenting Speed/Time Data Into
Modes 3-71
Figure 3-28 Segmentation of Speed/Time Data Into Modes Using
Modified Mode Logic 3-72
Figure 3-29 Normalized Time in Mode Matrix 3-75
Figure 3-30 Normalized Mode Frequency Matrix 3-76
Figure 3-31 Time, Distance and Average Speed Versus Time of Day
and Road Type Matrices 3-78
Figure 3-32 VOS Data Conversion and Data Processing Computer
Programs 3-80
Figure 3-33 VOS Auxiliary Data Processing Computer Programs 3-81
Figure 3-34 IBM Los Angeles Data Center S/370 Model 155 Computer
Facilities 3-85
Figure 3-35 Mode Matrix Partitioning For Delta-Square Analysis 3-89
Figure 3-36 Matrix Mode Categories For Linear Regression
Analysis 3-91
Figure 4-1 Relative Correlation of Each City versus the
Composite 4-13
Figure 4-2 Acceleration Rate versus Instantaneous Speed: 5-City
Composite 4-18
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
LIST OF FIGURES (CONT.)
Page No.
Figure 4-3 Distribution Functions of 5-City-Composite Accelera- 4-19
tions at Various Instantaneous Speeds
Figure 4-4 Distribution Function of 5-City-Composite Decelera-
tions at Various Instantaneous Speeds 4-20
Figure 4-5 Average 0-30 mph Acceleration Profile for Each City 4-21
Figure 4-6 Average 30-0 Deceleration Profile for Each City 4-22
Figure 4-7 Intake Manifold Vacuum Versus Cruise Speed 4-24
Figure 4-8 Data Collection Schedule for Detroit (Ann Arbor)
Road Routes 4-37
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
Acknowledgments
Scott Research. Laboratories, Inc. wish, to express appreci-
ation to the CRC-APRAC CAPE-10-68 committee members for their tech-
nical direction in the conduct of thi$ program. Special thanks are
offered to the EPA representative, Dr. Thomas Huls, and the CRC Project
Manager, Alan E. Zengel, for the assistance offered during the various
phases of the effort.
The Vehicle Operation Survey was performed under the
guidance of John Harkins, Laboratory Director. Malcolm Smith acted
as Program Manager and Michael J. Manos functioned as Project Engineer.
Other Scott personnel contributing materially to the successful
completion of this project were: J. Marrin and R. G. Kinne - in-
strumentation and vehicle preparation; G. Huibregtse - data processing
and analysis; and S. Fischer - report preparation.
SCOTT RESEARCH LABORATORIES, INC
-------
1-1
SRL 2922-13-1271
1.0 SUMMARY
This report presents the results of CRC-APRAC Project No.
CAPE-10-68, entitled Vehicle Operations Survey, conducted under joint
sponsorship of the Environmental Protection Agency and the Coordinating
Research Council, Inc. The purpose of the program was to define, de-
termine, and typify automobile driving patterns in terms of operating
modes. Data were collected in five major metropolitan areas and sub-
sequently combined to form an overall composite of urban driving patterns.
Data relating to basic vehicle usage patterns were obtained
from a preliminary program conducted by Systems Development Corporation (SDC).
Results of this initial study were utilized in the design and construc-
tion of traffic survey routes for the metropolitan areas of Los Angeles, ^
Houston, Cincinnati, Chicago, and New York City. The SDC reports were
also useful in determining the parameters of vehicle operation to be sur-
veyed during the subject program.
Three surveyor vehicles were instrumented with digital data
acquisition systems for use in the field. The "chase-car" concept
was utilized, whereby the instrumented vehicles were operated to emulate the
driving patterns of various cars representing the traffic population.
Operating parameters of vehicle speed, time, and manifold vacuum, together
with various route descriptors, were obtained from the chase vehicle and
recorded on magnetic tape for computer batch processing.
The data were processed to identify and summarize the basic vehicle
operating modes: acceleration, deceleration, cruise, and idle. Mode
characteristics such as frequency of occurrence, total duration, average
SCOTT RESEARCH LABORATORIES, INC
-------
1-2
SRL 2922-13-1271
duration, and transition probability were defined in matrix form. Supplementary
information was obtained on average trip speed, acceleration-deceleration
profiles, and manifold vacuum rates at various cruise conditions.
Evaluation measures were defined to determine driving pattern
similarities between cities and variability within a city. Merging of
data from all cities yielded an overall composite of vehicle operation.
A road route representing this five-city composite was constructed in
the metropolitan Detroit area.
Volume I of this report discusses the program objectives, pro-
cedures, and analysis of results. Volume II is an Appendix volume which
provides additional detailed information on instrumentation, survey route
design, data processing, and program results.
SCOTT RESEARCH LABORATORIES, INC
-------
2-1
SRL 2922-13-1271
2.0 INTRODUCTION
This section states the objectives of the Vehicle Operation
Survey program, relates some of the background information pertinent to
the project, and presents the basic scope of operations.
2.1 PROGRAM OBJECTIVES
The primary objective of the Vehicle Operation Survey program
was to describe typical urban driving patterns. In meeting this objective,
completion of the following tasks was requisite:
o Define the parameters of vehicle operation which
best describe driving patterns
o Design representative urban road routes over
which vehicle operating data can be collected
o Instrument vehicles for acquisition
of vehicle operating data and related
information
o Process the raw data to generate matrices which
characterize vehicle operating modes and other
descriptors of driving patterns
o Construct a composite of vehicle operating patterns
based on data from all cities and design a road route
in the greater Detroit area which can be said to
represent the overall composite.
2.2 BACKGROUND INFORMATION
Studies were performed in 1964-1965 to establish vehicle operating
patterns for the Los Angeles area*. Engine rpm and manifold vacuum
data, obtained from a vehicle operating within a 6-mile radius of central
Los Angeles, were used to construct a road route (LA-4) and related
* References 1 and 2.
SCOTT RESEARCH LABORATORIES, INC
-------
2-2
SRL 2922-13-1271
dynamometer cycle (XC-15), both of which typify morning rush-hour driving
behavior in the downtown area.
The first part of the CAPE-10 program was conducted by the
System Development Corporation. They performed surveys in the urban areas
of Los Angeles, Houston, Cincinnati, Chicago, New York City, and Minneapolis-
St. Paul to determine vehicle-usage patterns. Information contained in
their reports* relates vehicle usage to the following trip descriptors:
o Start time
o Trip and segment distance
o Trip and segment average speed
o Number of momentary stops
o Idling time
o Road type
o Purpose of trip.
The second part of the total effort is the subject of this report.
Details are contained herein, describing how the SDC information on vehicle
usage was used as a tool to construct driving survey routes. The data
collected on these routes became the basic input for the determination of
national composite urban driving patterns.
2.3 SCOPE OF OPERATIONS
The Vehicle Operation Survey (VOS) was conducted in two
phases. Phase I efforts were divided among the following tasks:
o Design, fabrication, installation, and checkout
of digital data acquisition systems
o Construction of four representative driving
survey routes in Los Angeles
o Training of personnel and data collection
on the Los Angeles routes
o Writing of computer programs and initial processing
of Los Angeles data.
* References 3 through 9.
SCOTT RESEARCH LABORATORIES. INC
-------
2-3
SRL 2922-13-1271
Phase II activities were guided by the experience obtained during
Phase I. Phase II tasks were:
o Design and construction of driving survey
routes for Houston, Cincinnati, Chicago,
and New York City
o Collection of data in the above four cities
o Modification of computer programs written in Phase I
to optimize processing efficiency and output of
useful information
o Processing of data for all five cities (including
reprocessing of Los Angeles data with revised
programs)
o Design, and evaluation of a road route in Ann
Arbor, Michigan, which was intended to typify the
five-city composite results
o Evaluation of vehicle operating patterns on the
LA-4 road route for comparison with urban Los
Angeles results
o Analysis of data and final report.
Throughout the duration of the data acquisition operations,
two primary survey vehicles (plus one back-up vehicle) were used for
collecting data in each city. The data represent a total of
22,465 miles of freeway, arterial, and capillary roads surveyed
during the spring of 1971. Data collection consisted of operating
(nominally) from 6:00 am to 10:00 pm for 7 consecutive days in each
major urban area. Nearly 31 million pieces of information were re-
corded on 1/2 inch magnetic tape for subsequent computer processing.
The initial program schedule included the surveying of
vehicle operating patterns in the Minneapolis^St. Paul area. This
task was dropped, however, in favor of conducting an evaluation of the LA-4
road route, not originally included in the project plans.
SCOTT RESEARCH LABORATORIES, INC
-------
2-4
SRL 2922-13-1271
No data collected during the. course of program were considered
unusable due to extreme weather effects on traffic flow. Rain was
observed in all cities except Chicago; however, neither the rate nor
extent of the precipitation was sufficient to create atypical driving
patterns. The digital data acquisition systems performed with very few
instances of malfunction. Occasional problems with electro-mechanical
components were easily remedied.
The data processing programs yielded various descriptive measures
of driving patterns in tabular, matrix, and graphical format. No single
measure of vehicle operating patterns can be considered as adequate for
representing the total pattern for a city. Therefore, a set of evaluation
measures was used to analyze and compare data for each city with the other
cities and with the composite.
SCOTT RESEARCH LABORATORIES, INC
-------
3-1
SRL 2922-13-1271
3.0 PROGRAM DESCRIPTION
The following sections will discuss the preparations, equip-
ment, procedures, and data processing programs used to conduct the
project:
3.1 General Approach
3.2 Instrumentation Systems
3.3 Survey Route Design
3.4 Data Collection Operations
3.5 Data Processing and Analysis Techniques
3.1 GENERAL APPROACH
3.1.1 Preparatory Efforts
The groundwork was laid during the previously-conducted SDC pro-
ject for the determination of the vehicle operating parameters to be
considered for the VOS program. Vehicle usage trends were defined
primarily in terms of a vehicle trip and descriptors of the trip. These
descriptors, such as trip length, road type, purpose of trip, day of
week, and time of day, were all considered in the design and construction
of the driving survey routes. Traffic density effects were accounted
for in a data processing routine.
Digital data acquisition systems (DDAS) were Scott designed, and
fabricated under subcontract by Datum, Inc., to measure the significant
parameters of vehicle operation required for describing driving patterns.
Chase vehicle speed, time of day, intake manifold vacuum, and coded inputs
of route descriptors were recorded on magnetic tape. Speed measurements
were obtained from a rotary pulse generator, cable driven from a tee-adaptor
SCOTT RESEARCH LABORATORIES. INC
-------
3-2
SRL 2922-13-1271
on the transmission speedometer gear. A potentiometric gage pressure trans-
ducer and signal conditioner monitored intake manifold vacuum data. Additional
coded inputs denoting day of year, make of vehicle being followed, route
number, trip number, segment number, and road type were accomplished
with thumbwheel selector switches. These data were recorded at a rate
of one sample per second on 1/2-inch magnetic tape.
The electronic systems and digital tape recorder were cabinet-
mounted behind the rear seats of the station wagons used as chase vehicles.
Power for operation was supplied by two 12-volt truck batteries,
maintained by an auxiliary alternator belt-driven off the engine.
3.1.2 Data Collection Operations
Vehicle operating data were collected initially in the urban
Los Angeles area using the chase car concept. With this method the
chase-car driver emulated the driving habits of a large sample of the
vehicle population operating on representative portions of free-
ways, major arteries, and capillary roads.
Shortly thereafter, a similar evaluation of driving patterns
was conducted on the LA-4 road route located in central Los Angeles.
The processed data, describing mode operating patterns for the two
surveys, were compared to determine similarities (or differences) in
vehicle operation over these two routes.
Houston, Cincinnati, Chicago, and New York City urban areas
were subsequently surveyed to determine driving habits in those
metropolitan areas. During the course of operations, the survey vehicles
operated from 6:00 a.m. to 10:00 p.m. each day over a minimum of seven con-
secutive days in each city.
SCOTT RESEARCH LABORATORIES, INC
-------
3-3
SRL 2922-13-1271
3.1.3 Data Handling
Initial data processing requirements, including the writing
and debugging of programs and the processing of Phase I (Los Angeles)
data, were subcontracted to TRW Systems Group. The computer programs
accomplished the following tasks:
o Convert 7-track low-density data tapes to 9-track,
800-BPI intermediate data tapes
o Edit parity errors and illegal characters
o Establish operating modes and collect
acceleration/deceleration profile data
on punch cards
o Generate three-dimensional speed-mode
and supplementary matrices
o Merge and compact matrices; i.e., combine
route replicates and rescale matrix dimensions
o Print out results and perform matrix comparison
evaluations.
Prior to Phase II data processing, the computer program logic was
evaluated and, under a subcontract to IBM, minor modifications were incor-
porated to optimize processing efficiency and obtain additional information.
All of the data collected during the course of the project were processed
using the revised computer programs, including the reprocessing of the Phase I
LA data. Scott personnel utilized the facilities of the Los Angeles IBM
service center for this task, operating on the system 370/155 computer.
3.2 INSTRUMENTATION SYSTEMS
The VOS program required the use of three instrumented vehicles
for acquisition of vehicle operating data in the field. The vehicles were
driven as chase cars, emulating the operating patterns of other cars.
SCOTT RESEARCH LABORATORIES, INC
-------
3-4
SRL 2922-13-1271
Data on vehicle speed, intake manifold vacuum, time-of-day, and several
manually-coded inputs were thus obtained from the instrumented vehicle and
established as representative of the vehicles being followed.
During operation of the instrumented vehicles, data were recorded
at one-second intervals on IBM-compatible digital magnetic tape and simul-
taneously read back and displayed for real-time verification by the in-
strument system operator. All data tapes were subsequently processed
directly by digital computer.
Details describing the selection of chase-car vehicles, the
digital data acquisition systems (DDAS), and related operating systems are
presented in the following sections.
3.2.1 Vehicles Used for Collecting Data
The vehicles selected to transport the DDAS and related operating
systems and to function as the chase car were 1971 Ford Ranch Wagons.
Factors affecting the selection were:
o Available space in the engine compartment for mounting
the auxiliary power supply alternator for the instru-
ment system
o A recessed dash panel on the passenger side, convenient for
mounting the system control/display unit
o A station wagon body style, with increased interior space,
for mounting the DDAS and for transporting replacement
instrument system components and magnetic recording
tape during inter-city travel
o A high-powered (360 HP) engine for carrying the added
weight of the instrumentation and for adequate acceleration
characteristics necessary for duplicating other vehicles'
operating patterns
o Air conditioning to provide a controlled temperature
environment for the instrument system.
SCOTT RESEARCH LABORATORIES, INC
-------
3-5
SRL 2922-13-1271
Air-adjustable shock absorbers were Installed on the rear
axles of each vehicle to improve handling and ensure a level attitude
under the increased loading.
3.2.2 Transducers
Speed and manifold vacuum transducers were installed in each
vehicle. Figure 3-1 shows these transducers and their functional re-
lationship to the other instrument components.
Chase-vehicle speed was measured using a rotary pulse generator
coupled to the transmission speedometer output shaft through a 1:1 tee-
adapter. Use of the tee-adapter also permitted simultaneous operation of
the vehicle speedometer.
The generator output was a series of pulses proportional
to the vehicle distance traveled (10 pulses per mile). These pulses were
supplied to the data acquisition unit where they were accumulated during
a fixed time interval. During calibration of the digital speed channel,
the accumulator time interval was adjusted to provide a count equivalent
to vehicle speed in miles per hour, with a full scale capability of
99.99 mph. The hundredths digit, which was not recorded, was accumulated
to preclude potential problems caused by any whiplash in the speedometer
drive train or by vehicle electrical noise.
Calibration of the digital speed channel was performed using
a calibrated fifth wheel as the comparison standard. These calibrations
were performed on each of the three program vehicles before and after
the Los Angeles survey operation, and after the 5-<-cities survey operation.
Results of these calibrations, showing the deviations of the digital speed
SCOTT RESEARCH LABORATORIES, INC
-------
r
INSTRUMENT POWER SUBSYSTEM
1
30
m
c/>
pi
n
o
90
I
5>
p
To Vehicle Speedometer
12 VDC Alternator/Batteries
115 VAC Inverter
Power Control/Monitor Unit
115 VAC
12 VDC
L.
t
Rotary
Pulse
Generator
Pressure
Transducer
Signal
Conditioner
,._,
1 *
jjj.ijj.iAL, un.1
Data
Acquisition
Unit
A A^yi
15J.lJ.UiN SX311UM
Control/Display
Unit
Digital
Magnetic
Tape
Recorder
1
CO
bo
N)
I
U)
I
U>
I
ON
Figure 3-1. Functional Block Diagrams, VOS
Mobile Instrumentation System
-------
3-7
SRL 2922-13-1271
indications from the fifths-wheel speed indications, are presented in
Appendix A-l.
Engine manifold vacuum was measured using a potentiometric
pressure transducer with a pressure range of -15/0/+5 PSIG. Pressure input
to the transducer was supplied from the chase-vehicle intake manifold through
a snubber valve. This valve was necessary to reduce the high-frequency
content of the manifold-vacuum pressure signal caused by the opening of
each individual intake valve. These high-frequency pressure signals were
found to be damaging to the pressure transducer.
The pressure transducer electrical signal was processed by
a signal conditioner which supplied power and bridge-completion circuitry
for scaling the transducer output to a -300/0/+100 millivolt signal. The
-300/0/+100 scaling was adjusted during calibration of the pressure trans-
ducer to give direct readings in inches of Hg with a resolution of 0.1
in. Hg. This signal was then supplied to the data acquisition unit for
analog-to-digital conversion.
Calibration of the manifold vacuum channel was performed by
supplying an external vacuum source to the transducer and using a mercury
manometer as a comparison standard. Calibrations were performed before
and after the survey operations on each pressure transducer as installed
in its companion vehicle. Results of these calibrations, showing the
deviations of the pressure channel indications from the mercury manometer
measurements, are presented in Appendix A-2.
SCOTT RESEARCH LABORATORIES, INC
-------
3-8
SRL 2922-13-1271
The locations of the speed transducer, tee-adapter, and
manifold vacuum transducer, as installed in the chase vehicles, are shown
pictorially in Figure 3-2.
3.2.3 Digital Data Acquisition System
The DBAS, shown functionally in Figure 3-1, consisted of three
main units: the data acquisition units, the remote control/display unit,
and the digital magnetic tape recorder.
The data acquisition unit, containing all integrated circuit
electronics, performed the following functions:
a. Accumulated the speed pulse data over a precisely-
adjusted time interval to provide a digital count
equivalent to the vehicle speed in MPH, resolved to
0.01 mph
b. Amplified and converted the manifold vacuum analog
signal to a digital format equivalent to the manifold
vacuum in inches-Hg vacuum with a full-scale capability
of 30.0 inches-Hg
c. Generated digital time-of-day data from a 1-MHz clock
oscillator source. The clock was also used to control
internal logic functions
d. Provided the digital manifold vacuum and time-of-day
data to the remote control/display unit for display
e. Received the digitally-coded thumbwheel data from the
remote control/display unit
f. Time-multiplexed the time-of-year, time-of-day, fleet,
speed, manifold vacuum, weather, road type, segment,
trip, and route digital data into a 27-character
binary-coded decimal (BCD) serial format and presented
these data to the magnetic tape recorder write data
input lines once each second for data recording
SCOTT RESEARCH LABORATORIES, INC
-------
trt
I
0
o
90
a. Speed Transducer
b. Speedometer Tee Adaptor
c. Manifold Vacuum Transducer
d. Data Acquisition Unit
e. Control/Display Unit
f. Magnetic Tape Recorder
g. Instrument Alternator
h. Regulator
j 12 VDC Batteries
k. 115 VAC Inverter
m. Power Control Monitor
Figure 3-2 Instrumentation Component Locations, VOS Vehicles
-------
3-10
SRL 2922-13-1271
g. Received the read-back data from the magnetic tape recorder
once each second, de-multiplexed these data, and provided
them to the control/display unit for display
h. Processed and generated all system control signals for
starting and stopping the data acquisition functions.
The remote control/display unit was dash-mounted on the passenger
side of the vehicle front seat for operation by the instrument system
operator. This unit, whose panel is shown pictorially in Figure 3-3,
performed the following functions:
a. Provided NIXIE tube numerical display of the hours,
minutes, and seconds time-of-day data, the manifold
vacuum data, and a selectable 9-digit field of read-
back data. The read-back data were used by the in-
strument system operator for real-time comparison
with the data being written on magnetic tape
b. Provided a three-position rotary switch for selecting the
desired 9 (of 27) digits of read-back data for display
on the read-back NIXIE indicators
c.
d.
e.
f.
Provided digital thumbwheel switches for manually coding
the day, fleet, weather, road, segment, trip, and route
data
Provided a toggle switch for manually overriding the
contents of the fleet thumbwheel switches. This
override presented 3 digits of zeros (000) to the
digital multiplexer and was used when the chase vehicle
was not following another vehicle.
Provided a thumbwheel switch, a run/hold toggle switch
and 6 push-button switches for manually setting the
system clock to any desired time of day.
Provided system status indicators to indicate stopping or
starting of data acquisition, parity-error detection, system
write status, and end-of-tape sensing
Provided push-button switches for starting and stopping
data acquisition, resetting the DBAS, resetting the
parity error indicator, and writing end-of-file
characters on magnetic tape.
SCOTT RESEARCH LABORATORIES, INC
-------
90
C/J
30
n
03
o
i
]
DAY!
i
I
2
]
HOURS
0
i
9
©
0
0
1
FLEE1
fEATH
ER
1
0
n
1
u
FLEET
1
VEATHER
3
1
I
1
EGM
6
S
El
@
I»E
MT
MINU
i
[4
i
>
riME
9
©
(
3,
©
R!
.2
6
"000"
KOAD-
2
[
i
IEGMENT
2
1
VOS CONTROL
2
Tl
1
CC
i
4
©
(FOLD
2
ttmr TitiE rcT ^
7
RUN
1f\
&
MOID
©
s^^
V>
51
9
IIFOLD
2
7
9
VACUUM-
9
I
0
DISPLAY PANEL
O
1
9
5
T^O ^
p^w ;
WRITE
PERMIT EOT
0 0
PARITY
O ©
EOF
(§)
START
O ©
STOP
RESET
©
N3
I
U>
I
UJ
I
Figure 3-3
-------
3-12
SRL 2922-13-1271
The DBAS included a 7-track, 200-BPI, IBM-compatible, synchronous
digital magnetic tape recorder which received, once each second, 27
characters of binary-coded-decimal Q&CD) data in a serial format from the
data acquisition unit multiplexer. The 27-character multiplex was con-
tinuously written on magnetic tape and immediately read back and sent
to the data acquisition unit for processing and display on the remote
control/display unit. The locations of the data acquisition unit, the
control/display unit, and the magnetic tape recorder are shown pictorially
in Figure 3-2.
Figure 3-4 shows the remote control/display unit as it was
installed on the chase-vehicle dash. The unit was provided with an
integral light hood and night light necessary for viewing the panel under
all ambient light conditions. Figure 3-5 shows the data acquisition unit
and the magnetic tape recorder installed in racks in the rear of the
vehicle. Note the immediate accessibility of the circuit cards in the
acquisition unit and recorder. Figure 3-6 shows a rear view of the
instrument racks containing the acquisition unit and the recorder.
3.2.4 Instrument Power Subsystem
Power for operating the data acquisition system and system
transducers was provided by the Instrument Power Subsystem, shown functionally
in Figure 3-1. Primary power for the instrument system was supplied
from an additional 12-VDC 135-AMP alternator mounted in the chase-vehicle
engine compartment and driven by the vehicle engine. This alternator
provided power for operating only the instrument system. The alternator
output was regulated and supplied power to two 12-VDC 135-Amp-H.our
batteries in parallel, providing a stable 12-Volt power source. Output
SCOTT RESEARCH LABORATORIES, INC
-------
3-13
SRL 2922-13-1271
Figure 3-4 Control/Display Panel and Power Control Unit
Installed on the VOS Vehicle dash
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-14
Figure 3-5 Data Acquisition Unit, Magnetic Tape Recorders, and
115 VAC Inverter Mounted in VOS Vehicle Rear
deck. (a) Accessibility of Electronic Circuit Cords
(b) Ready for Survey Operations
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-15
Figure 3-6 Rear of Instrument Racks Containing the Data
Acquisition Unit, Magnetic Tape Recorder, and Inverter
(a) Ready for Survey Operations
(b) Access to System Cables and Batteries
SCOTT RESEARCH LABORATORIES, INC
-------
3-16
SRL 2922-13-1271
from the 12-VDC alternator/battery source was supplied to a DC-to-AC in-
verter which provided 115<-VAC, 500--VA power.
Both the 12-VDC and 115-VAC power were controlled by the Power
Control/Monitor Unit before being supplied to the instrument system. The
Power Control/Display Unit panel is shown pictorially in Figure 3-7. This
panel provided meters for monitoring the system 12-VDC and 115-VAC power
buss voltages and battery-charging current. The panel also contained a
4-position main power rotary switch which controlled the application of
various forms of power to the instrument system during survey operations.
Controls for both electrically and pneumatically calibrating the
manifold vacuum transducer were provided. The necessary adjustment potentio-
meters and switches for electrically balancing and spanning the manifold-
vacuum transducer and signal-conditioning electronics are in the upper
right-hand corner of the panel. A two-position manual valve and a cali-
bration port are installed in the lower portion of the panel for providing
a pressure-calibration source input to the manifold-vacuum transducer.
The locations of the power system alternator, regulator, batteries,
115-VAC inverter, and control/monitor unit, as they are installed in the
chase vehicle, are shown pictorially in Figure 3-2. The instrument system
alternator mounted in the chase-vehicle engine compartment is shown in the
center of Figure 3-8, and its companion regulator is mounted in the lower
left corner of the same figure, The instrument system batteries were mounted
as shown in the lower half of Figure 3-6b and were concealed beneath the vehicle
rear-floor deck as shown in Figure 3^6a. The 115-VAC inverter was installed
SCOTT RESEARCH LABORATORIES. INC
-------
«
SO
PI
C/)
m
I
§
so
§
50
P
P
DC \ VOLTS
BAL
CAL
Eadj
ZERO
OPERATE
AC X VOLTS
OVERNITE
OFF
LUNCH/
DINNER
OPERATE
MAIN POWER
1 AMP
CALIBRATE
OPERATE
VOS POWER CONTROL/MONITOR UNIT
CO
p
N5
V£>
S3
K>
I
I
H-
ro
OJ
I
Figure 3-7
-------
3-18
SRL 2922-13-1271
Figure 3-8 Instrument System Alternator and Regulator Installation
in a VOS Vehicle
SCOTT RESEARCH LABORATORIES, INC
-------
3-19
SRL 2922-13-1271
in the rear instrument rack above the data acquisition unit, as shown in
Figure 3-5. The power control/monitor unit, installed for instrument
operator viewing on the vehicle dash, is shown in the left hand side of
Figure 3-4.
3.3 SURVEY ROUTE DESIGN
This section discusses the approach used in designing survey
routes over which vehicle operating data were collected. The approach
relies on SDC-published data describing vehicle-use patterns for the
various cities.
3.3.1 Route Design Objectives
Four general objectives were prescribed in the formulation of
a survey route. These objectives were:
o The survey routes should be representative of
vehicle usage for a particular city
o Data published in SDC's reports should be
used as design guidelines wherever applicable
o The route should provide an opportunity for
continuous data collection
o For logistic reasons, a survey route should be
driveable in an 8-hour work shift.
SDC data indicated that vehicle usage, characterized by an
evaluation of some 27,000 trips in the five cities, cannot be categorized
into simple, well-described daily-use patterns. The patterns found to be
"typical" actually represented a small percentage of all patterns.
SDC-defined trip descriptors such as start time, trip and segment
distance, elapsed time, average speed, number of stops (idles),
SCOTT RESEARCH LABORATORIES, INC
-------
3-20
SRL 2922-13-1271
road type, and trip purpose were scrutinized to determine existing re-
lationships. By analyzing the relationships between descriptors and
identifying the predominant trends, a method for developing a route
model evolved that had applicability to each city-
3.3.2 Route Design Factors
The following factors (trip descriptors) were considered
in the development of a route model for each city:
o Number of Trips o Purpose of Trips
o Length of Trips o Day of Week
o Road Types o Time of Day
Relationships between these factors are shown in Figure 3-9.
Generally speaking, functional relationships between factors which could
be defined numerically were incorporated directly into the route-model
structure. One of the factors, purpose of trip, could not be mathematically
related to other factors and was accounted for by considering the direction
of travel when mapping out a survey route.
3.3.3 Route-Model Structure
Route models for each city were designed so that the relation-
ships exhibited in Figures 3-10a and 3-10b were comparable between the
original SDC data and the route model. The relationship shown in Figure 3-10c
is accounted for in the data processing program, as discussed later in
this report.
The route model was structured initially by the number of trips
of various lengths. Trips were classified by trip distance and placed
into the categories shown in Table 3-1. Trips over 27 miles in length were
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922-13-1271
3-21
Number
of
Trips
Length Road Purpose Day of Time of
of Trip Type of Trip Week Day
A
Length
of
Trips
Relationship
Relationship
inite or Import
onship
B
A
Road
Type
ant
A
C
C
Purpose
of
Trips
B
B
C
A
Day
of
Week
A
C
C
A
B
Figure 3-9 Relationships Between Factors Affecting Route Design
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922-13-1271
3-22
(a)
(b)
(c)
N = Number of Trips
L = Length of Trips
%N_, = Percent of Trips with Freeway Segment
r
T = Time of Day (From Time = 0000 to 2359)
= Weekdays
= Weekends
Figure 3-10 General Form of Trip Descriptor Functional Relationships
SCOTT RESEARCH LABORATORIES. INC
-------
@
Table 3-1
u. Typical Route Model Structure For Weekdays
I
[T RESEARCH LABORATOR
P
O
Trip
Distance
Interval
(Miles)
0 -
3 -
6 -
9 -
12 -
15 -
18 -
21 -
24 -
3
6
9
12
15
18
21
24 T
27 J
Number
of
Trips
(Total)
1539
695
365
280
194
146
134
82
64
Number
of
Trips
(Factored)
11
5
2
2
1
1
1
1
24
Mean
Length
(Miles)
1.5
4.5
7.5
10.5
13.5
16.5
19.5
24.0
Total
Length
(Miles)
16.5
22.5
15.0
21.0
13.5
16.5
19.5
24.0
148.5
% Trips
With F
Segments
3
16
45
66
76
82
87
94
Number of
Trips With
F Segments
0
1
1
1
1
1
1
1
~7
F
Mileage
Factor
.64
.64
.64
.64
.64
.64
.64
.64
P
to
VD
to
to
Freeway £j
Miles *"
0.0
2.8
4.8
6.7
8.6
10.6
12.5
15.4
61.4
CO
10
-------
3-24
SRL 2922-13-1271
eliminated. This distribution of trips was factored by dividing the
number of trips in each interval by a least common denominator. The
least common denominator was defined to be the number of trips in the
least populated interval or group of intervals. The factored number of
trips in each interval was then multiplied by the mean trip length
for that interval and these products summed to obtain the total length
for the model route.
The percentage of trips with freeway segments within each
distance interval was determined from the SDC data. These percentages
were applied to the number of trips in each interval to obtain the de-
sired number of trips with freeway segments. An overall freeway-mileage
factor, based on the number of miles actually driven on freeways as a
percentage of the total mileage, was then used to determine the desired
length of freeway segments. The total number of trips, total route
length, number of trips with freeway segments, and the freeway mileage
as a percent of the total route mileage were then checked against SDC's
findings. Minor adjustments, made on individual trips to align the overall
route characteristics with SDC's findings, were necessitated by the
rounding-off inherent in the process.
This approach yielded unique weekday and weekend route models
by applying the appropriate basic trip information. At this point the
following four route-design factors were accounted for in the structure
of the route model:
o Number of trips
o Length of trips
o Road types
o Day of week (weekday or weekend).
SCOTT RESEARCH LABORATORIES, INC.
-------
3-25
SRL 2922-13-1271
Route-model design parameters were then correlated with the SDC
values to ensure compatibility. The results are shown in Tables 3-2 and
3-3 for weekday and weekend routes, respectively.
The two remaining design factors to be considered, "Time of
Day" and "Purpose of Trips", will be discussed in the following section.
3.3.4 Survey Route Design
The route-model structure shown in Table 3-1 exemplifies the
distribution of trip characteristics associated with a typical weekday
route for most cities. Route models for weekend application were similar
in composition, with a general tendency toward a larger percentage of
shorter trips. However, comparison of the route models for weekday and
weekend application revealed that approximately 80% of the individual trips
were common to both models. This fact facilitated the use of the same road
sections for both weekday and weekend survey routes in any given area.
Trips not common to both routes were located in the trip sequence for ease
of inclusion or exclusion depending on the day of the week in which
operations were being conducted. Table 3-4 presents the basic statistics
on the survey routes designed for each of the cities. Detailed information
on the structure of the individual routes is included in Appendix B.
Predominant in each city was the occurrence of the "home-to-work"
trip during the 0600-0859 time slot and "work-to-home" trip in the early
evening hours, as shown in Figure 3-11. The direction of travel for the
survey vehicles was thus controlled with these factors in mind. Each
base of operations (start point for survey operations) was located
approximately 15-20 miles from the downtown sections. The survey vehicles
SCOTT RESEARCH LABORATORIES, INC
-------
©
SCOTT RESEARCI
g
90
H
O
90
2
Trip
Length,
Miles
0-3
3-6
6-9
9-12
12-15
15-18
18-21
21-24
24-27
Table 3-2
Correlation of Trip Length Distribution
Los Angeles Houston Cincinnati
Actual
1539
695
365
280
194
146
134
82 i
L
64 J
Model
11
5
2
2
1
1
1
1
Actual
743
431
307
215
159
108
52 -,
L
28 J
13
Model
9
5
4
3
2
1
1
0
Actual
1200
613
396
298
162
90
76 )
L
38 J
21
Model
11
5
3
3
1
1
1
0
- Weekday Routes
Chicago
Actual
1824
795
493
315
222
169
108
101 -|
}
67 J
Model
11
5
3
2
1
1
1
1
tn
P
vo
1
u>
New York ^
Actual*
2679
1070
603
352
201
100 T
r
150 J
96 -\
L
100 J
Model j-J
14
5
3
2
1
1
1
0.998
0.996
0.995
0.998
0.999
* Taxi trips factored out of totals. SDC's trips were defined as "key-on to key-off" operation which
resulted in successive fare taxi trips of unusually long mileage. These trips were not comparable to
normal passenger car data and were therefore deleted prior to route model construction.
-------
4c1r&
*&f
a.
x
»J
*3
30
53
>
o
a:
£
0
S3
s
SO
]2
5
o
Trip
Length,
Miles
0-3
3-6
6-9
9-12
12-15
15-18
18-21
21-24
24-27
T
Correlation
Los
Actual
658
270
101
40
38
29
23
15 ^
}
14 J
0
Angeles
Model
14
5
2
1
1
1
1
1
.997
of Trip
Houston
Actual
414
131
75
42
40
27
15 ,
y
9
3
0.
Table 3-3
Length Distribution
Cincinnati
Model Actual Model
17
5
3
2
2
1
1
0
999
680
285
122
75
61
29 ,
V
14 *
8
0
0.999
16
7
3
2
1
1
0
0
- Weekend
Routes
to
P
VO
ISJ
M
1
CO
Chicago New York /.
Actual
950
298
141
95
51
33
30.
}
20J
25
0.
Model Actual*
19 1012
6 309
3 166
2 80
1 38
1 29 -»
12
1
8
0 9 J
999 0.
Model M
i-1
17
5
3
1
1
1
999
* Taxi trips factored out of totals. SDC's trips were defined as "key-on to key-off" operation which
resulted in successive fare taxi trips of unusually long mileage. These trips were not comparable to
normal passenger car data and were therefore deleted prior to route model construction.
-------
Table 3-4
Survey Route Design Characteristics
City
Los Angeles
Houston
Cincinnati
Chicago
N. Y. City
Number
Of Routes
4 Weekday
4 Weekend
2 Weekday
2 Weekend
1 Weekday
1 Weekend
2 Weekday
2 Weekend
2 Weekday
2 Weekend
Average
Route
Mileage
149.82
144.05
162.22
156.35
143.15
129.15
157.70
151.60
143.25
114.55
Average
Freeway
Mileage
62.05
62.05
53.50
49.08
35.40
26.00
42.32
31.88
46.82
34.48
Percent
Freeway
Mileage
41.41
43.08
32.98
31.39
24.79
20.13
26.84
21.03
32.68
30.10
Overall
Percent
Freeway
Mileage
41.85
32.54
23.51
25.22
32.05
CO
p
N3
\O
K)
I
H-
U)
IsJ
-J
CO
NJ
OO
-------
M
SO
o
p
Purpose
of Trip
WORK
BUSINESS
SHOPPING
RECREATION
HOME
% of Total
Trips in
Each Time
Slot
TIME OF DAY
0600 -
0859
80.5%
6.1%
6.7%
93.3
0900 -
1129
21.2%
47.5%
15.3%
8.3%
92.3
1130 -
1329
29.2%
24 . 2%
29 . 2%
12.0%
94.6
1330 -
1629
16.7%
25.8%
22.2%
28.3%
93
1630 -
1829
25%
11.4%
55.7%
92.1
1830 -
2059
30.2%
24.3%
38.7%
93
% of Total
Trips for
Each Purpos
94.5
86.5
95.5
60
82.5
CO
p
VO
ro
u>
OJ
I
NJ
VO
Figure 3-11 Distribution of Trips by Time of Day and Purpose of Trip (Weekdays)
-------
3-30
SRL 2922-13-1271
were
generally directed toward the center of town during the early morning
time slots, and in an outbound direction during the early evening time
slots. It must be recognized that strict adherence to these traffic-
direction objectives was not feasible and probably not desirable, since
the relationships existing between the two factors cannot be absolutely
categorized with respect to direction of travel.
Upon determining all the pertinent trip characteristics
and arranging trips into a sequence, the survey route was placed on a
detailed street map. In general, each route originated in the suburbs,
progressed toward the central city, and returned to the suburbs at least
twice during the course of the route. This double-loop system utilized
different roads on each half of the route and usually oriented the vehicle
in the direction of the majority of traffic during both the morning and
afternoon peak hours. That is, since both directions of travel on a given
route could not be covered simultaneously, it was judged to be of greater
value to obtain operating data on the majority of the traffic.
Each mapped-out route was transcribed into a route directory
%
which provided the surveyors with all the necessary information for driving
the route. Route directories supplied details on the following items:
o Route number
o Starting location
o Specific names of roads
o Direction of turns
o Length of segments
o Distance traveled on each street
o Trip number
o Segment number
o Road type code number.
A sample page from a route directory is shown in Figure 3-12.
SCOTT RESEARCH LABORATORIES, INC
-------
j*SV
%£?
«
2
3
33
M
8
>
SB
£
§
5
3
2
£
2
o
Segment
Road Segm. Trip Length
Type Number Number (Miles) Description
Start at Roscoe Blvd. and Orion
N-l 1 1 1.45 East on Roscoe Boulevard
Left on Columbus Avenue
Right on Parthenia Street
Left on Noble Avenue
Stop at Rayen Street
N-l 1 2 2.50 Straight on Noble Avenue
Right on Nordhof f Street
Left on Terrabella Street
Unit
Length
/\g I _ \
(Miles)
0.40
0.50
0.25
0.30
0.25
0.70
1.55
CO
p
ro
V£>
NJ
1
I"
U)
I-1
NJ
H1
F-2
N-l
Right on G.S. / 5 \ Freeway
8.20
2.60
G.S. / 5 S Freeway
Exit Burbank Boulevard
Right on Burbank Boulevard
Left on Keystone Street
Right on Oak Street
Stop at Naomi
8.20
0.90
1.30
0.40
OJ
Figure 3-12 Route Directory Sample Page
-------
3-32
SRL 2922-13-1271
Consideration was given to traffic density factors during data
processing. The original objective was to weight the vehicle operations
data to account for the number of vehicles in the traffic flow pattern
which were operating in a manner similar to that of the survey vehicle.
After contacting traffic engineers to obtain traffic density data, it
was concluded that the available information was not well suited to our
needs for the following reasons:
o The statistics usually presented traffic density
as the number of vehicles passing a point per
unit time
o The number of vehicles passing a given point per
unit time decreases under very heavy traffic con-
ditions
o Traffic volume data often included traffic in
both directions
o Peak-hour and peak-month traffic volumes are
of most importance to the traffic engineers and
generally were the only form in which such data
were published
o Traffic volume data were often not available for
capillary (low-volume) streets
o Some available data were up to 10 years old.
SDC-published data, in the form of number of trips versus
time of day, were judged to be applicable for the purpose of weighting
the vehicle-operation data. Distributions of trip start times for
fifteen-minute intervals provided weighting factors for weekday and
weekend use for each city. The actual values used for traffic density
weighting are given in Appendix C.
SCOTT RESEARCH LABORATORIES. INC
-------
3-33
SRL 2922-13-1271
3.4 DATA COLLECTION OPERATIONS
This section discusses the procedures and schedule of operations
associated with the data -acquisition phases of the program.
3.4.1 Training of Personnel
A personnel training program was initiated prior to completion
of the instrumentation of the survey vehicles. Instructions in the appli-
cation of the chase-car technique, operation of the DBAS, manual input of data,
tape handling and labeling procedures, and general operating procedures
were presented. Four crews, consisting of two men each, were involved
in the training program.
Three forms of the chase-car concept were evaluated. One
method was based on maintaining a constant distance between the survey
vehicle and the automobile being followed. The second method employed
a constant time delay between the operating events of the two vehicles.
The third method combined the attributes of the first two methods by
maintaining constant distances at cruise conditions while allowing for
a time lag during acceleration and deceleration modes. Selection of the
third method was indicated after comparing the strip-chart recordings from
the two vehicles. The combination method placed less emphasis on driver
reaction time, was less hazardous for the chase vehicle, and produced the
most consistent speed/time traces. The road evaluations revealed another
important factor: the problem of overcompensation on the part of the
chase-car driver, as indicated by erratic speed-time traces. This occurred
when the driver of a chase vehicle concentrated on exactly duplicating
the operation of the car in front. By instructing the drivers just to
follow a vehicle by "floating behind", the speed irregularities were
reduced to a minimum.
SCOTT RESEARCH LABORATORIES, INC
-------
3-34
SRL 2922-13-1271
Following a short practice session during which drivers were
allowed to familiarize themselves with, the chase-car technique, a speed-
time comparison was performed. A lead car and a chase-car, both equipped
with strip-chart recorders, were operated on a local urban road route.
The resultant speed-time traces were superimposed, as illustrated in
Figure 3-13. Comparison of the two speed-time traces indicated that the
technique under consideration was feasible for simulating vehicle operation
in the field.
Operating procedures for the DBAS were explicitly detailed and
hence lengthy. A check sheet was drafted to delineate the procedural
sequence for the following phases of operation:
o Morning start-up
o Short-break stand-by
o Restart after short break
o Long-Break stand-by
o Restart after long break
o Overnight shut down
o Calibration of speed-sensing system
o Calibration of vacuum-sensing system.
A copy of the instrument system check sheet is shown in Appendix D.
During data acquisition periods the vehicle speed, intake
manifold vacuum, and time-clock data were automatically recorded once
per second. This information was supplemented by manually-input data
defining the route descriptors, type of vehicle being followed, and
the existing weather conditions. Numerical codes were developed for
these entries. Vehicles were classified into 14 groups by make of
vehicle, as shown in Figure 3-14. Weather codes were defined with re-
spect to effect on traffic flow, as presented in Figure 3-15.
SCOTT RESEARCH LABORATORIES, INC
-------
I
m
30
O
DO
O
30
O
3D
O
0
Jjr
-
r
-
E
-
f
n n n ^ ° °
J U u u j u ,
S NOUOUOddOD S1OU1NO3
o
^r~
=:
F ~~
35E
r "
e
-._ -
~~r.
- ^
_
r
^I=-
^
- o
"O
-- oO
--O
O
o
- ro
-w
o
o
-~- r
^^-_
EE
:;
y
H
} u
OIHdt
:
M
if
J O
'1 f r) f r\ '
U u '.. U J >.;
«a n
r^=
-
n
c,
1MWH
^.~"
~-^
1
O (
3 ONI
r~
-_-_-
-L-"
_.._.
3H03
-:;-:
::~.:
- -
\
3uig
--:.
:
J J
-:L-
ij
-=
-.--
10 Seconds
Dir
~ Tr
\ )
ecti
of
avel
V
J.
J O
-
1
0
[j u U
-.
--r
..-.:
---
V
O <
--
;~
---
~ "L
-
><
J O
1 ]
-~rr-
---
-__-_
--
-'---
1
I'l
1
"-_
. -- -
: ---
O O
: U
.. L
------
H
x1/
N
.j O
0
^=~
"~
-. --"
---
7-
o
U U 0
~-^
. .....
1
- L":
O '
':-'--.
^.7.1
?_-A
-----
: o
[1
o
0
- o
o
00
- o
"O
o
1
C-
0 I
.--
_jrr:.
^r- :
1
--
-V
j 0
:1.
--^
'-
-
s?
C
0
-_-..
-:~
-.
|
o
009
--"
r'£.~
_--~
'^/ <
--L--
~- -
~E
V C^
0
-:.-.
--/
V
.__ .
o
to
VO
N>
ro
l
Lo
n
Figure 3-13
Speed-Time Traces From Lead Car & Chase Car
Lead Car
Chase Car
-------
3-36
SRL 2922-13-1271
001 CHEVROLET
Chevrolet, Chevelle, Corvair, Chevy II, Nova, Corvette,
Camaro, Vega, El Camino, Monte Carlo
002 FORD , . . ,
Ford Galaxie, Fairlane, Torino, Mustang, Falcon, Maverick
Pinto, Thunderbird, Ranchero
003 PONTIAC
Pontiac, Tempest, Le Mans, GTO, Firebird, Grand Prix
004 PLYMOUTH
Fury, Belvedere, VIP, GTX, Valiant, Barracuda, Swinger
005 BUICK
Buick, Special, Riviera
006 OLDSMOBILE
Oldsmobile, 7-85, Cutlass, Toronado
007 DODGE
Coronet, Monaco, Charger, Dart, Challenger, Duster, Polara
008 MERCURY
Mercury, Montego, Cyclone, Cougar, Lincoln Continental, Comet,
Meteror
009 CHRYSLER
Chrysler, Imperial
010 AMERICAN MOTORS
Ambassador, Rebel, Rambler, Javelin, AMX, Gremlin
Oil CADILLAC
Cadillac, El Dorado
012 OTHER AMERICAN
Checker, Edsel, De Sota, International, Jeep, Kaiser, Nash,
Studebaker, Etc.
013 VOLKSWAGEN
014 OTHER FOREIGN CARS
Toyota, Datsun, Jaguar, MG, Porsche, Etc.
NOTE: The fleet switch is to be in the "000" position when no vehicle
is being followed.
Figure 3-14 Vehicle Make Code
SCOTT RESEARCH LABORATORIES, INC
-------
3 *°
H **
M Visibility: M
V) LA?
n 01 Visibility normal; traffic speed normal N>
a M
5 02 Visibility moderately reduced; traffic speed
g moderately reduced
90 03 Visibility greatly reduced; traffic speed greatly
J2 reduced
Precipitation:
04 Precipitation light (or wet road); traffic speed normal
to
05 Precipitation moderate; traffic speed moderately reduced ^
06 Precipitation heavy; traffic speed greatly reduced
NOTES: No amount of smog will be considered great enough to cause
traffic speed reductions
Precipitation includes rain, snow, sleet and hail.
Figure 3-15 Weather Codes
-------
3-38
SRL 2922-13-1271
3.4.2 Data Collection Schedule
Prior to data collection operations on all driving survey routes,
the survey crews circumnavigated the. routes for a full day. During this
training day the personnel became familiar with the street and freeway
systems, traffic light arrangements, and local driving characteristics.
After the training period, data collection operations commenced, usually
continuing for seven straight calendar days. In this manner it was possible to
obtain vehicle operating data for the five weekdays and two weekend days
without repetition of any one day.
Two-man crews, consisting of a driver and a navigator (system
operator), were used on a two-shift/day basis. This system allowed
collection of data nominally from 6:00 a.m. to 10:00 p.m. daily. Of
the ten people involved in data collection operations, six were qualified
as both driver and navigator (system operator), two were just drivers,
and two were navigators only.
Table 3-5 shows the schedule for data collection operations in
each of the five urban areas and on the LA-4 route. Operations in Los
Angeles commenced five days after the New Year Holiday to avoid the
atypical driving patterns expected at that time of year. Similarly, data
acquisition efforts in Cincinnati were completed before the Easter weekend
to avoid possible holiday traffic influences. No weather conditions severe
enough to create prolonged traffic pattern disruption were encountered during
the entire schedule.
3.4.3 Operating Technique
Operating procedures in the field were systematically controlled
to ensure consistent, high-quality data. The ensuing discussion will
SCOTT RESEARCH LABORATORIES. INC
-------
Table 3-5
Data Collection Schedule - Vehicle Operations Survey
Urban
Area
Los Angeles
Houston
Cincinnati
Chicago
New York City
LA-4
Route Number
and Application
(1)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
98
99
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
W.
0.
D.
D,
D,
D
D
E
E
E
E
D
D
E
E
D
E
D
D
E
E
D
D
E
E
H
H
Reels of
Tape
Recorded
20
20
20
20
8
8
8
8
21
20
8
8
20
8
25
20
8
8
20
23
8
9
17
17
Date of
Survey, 1971
Jan.
Jan.
Jan.
Jan.
Jan.
Jan.
Jan.
Jan.
6-8 and Jan.
6-8 and Jan.
18-22
18-22
9-10
9-10
16-17
16-17
11-12
11-12
P
NJ
VO
10
NJ
I
K)
Mar. 19 and Mar. 22-25
Mar. 19 and Mar. 22-25
Mar. 20-21
Mar. 20-21
April 1-2 and April 5-7
April 3-4
April 13-16 and April 19
April 13-16 and April 19
April 17-18
April 17-18
April 26-30
April 26-30
April 24-25
April 24-25
Feb. 16-18
Feb. 16-18
VD
(1)
W. D. = Weekday W. E. = Weekend
0. H. = Off hours D. H. = Defined Hours
-------
3-40
SRL 2922-13-1271
delineate the basic sequence of operations for a typical work shift, beginning
at approximately 5:45 a.m. Procedures used by the afternoon shift survey
crews were nearly identical, beginning with their start time of 1:45 p.m.
The instrumented chase vehicles were started-up about 15 minutes
prior to the scheduled commencement of data collection. The DDAS power
supply system was energized and proper functioning was verified. The
magnetic heads and tape guide rollers on the recorder were thoroughly
cleaned and a fresh 1200-foot reel of 1/2-inch tape was mounted.
With the electronics warmed-up, the clock was set to the proper
time and started. The speed-measuring circuitry and manifold-vacuum
circuitry were electronically calibrated. With real-time data verification
capabilities designed into the DBAS, it was possible to check for proper
operation of these three circuits by scanning the nixie readouts on the
control/display pannel. All manually-operated thumbwheel selector switches
were then checked for proper functioning and set at the desired positions
for the start of data collection.
The survey crew then proceeded to the survey route starting
point, verifying the proper operation of all systems with the vehicle in
motion. At this time the tape was rewound to the Beginning-of-Tape
marker and data collection operations were started.
The chase-vehicle driver drove the survey vehicle around the
prescribed route in accordance with the directions in the survey-route
directory. The navigator/system operator supplied verbal instructions to
the driver regarding the correct roads to follow and where to make turns.
SCOTT RESEARCH LABORATORIES, INC
-------
3-41
SRL 2922-13-1271
At the appropriate places along the route,the navigator/system operator
reset the trip number, segment number, and road-type selector switches.
During the circumnavigation of the route, the driver used the
chase-car technique for following other vehicles. The make of vehicle
being followed was recorded with the appropriate code-number selection
of the "Fleet" thumbwheel switch. When no vehicles were being followed, the
"Fleet" switch was set in the "000" position and the driver continued along
the route at a normal pace. While following other cars, the survey vehicle
was restricted to operations consistent with the driving regulations and
prescribed speed limits for that particular area. While collecting
data on certain lengthy road segments, the driver was instructed to follow
vehicles for no more than approximately 2 minutes and to alternate lanes,
thus sampling a larger variety of the vehicle population.
Approximately every ten minutes the navigator/system operator
routed the display selector switch to each position as a check for
proper recording and readback of all collected data.
Each reel of magnetic tape was sufficient for about 3^ hours of
data collection. At the mid-point of the route, a break was taken. The
tape was stopped at this time, rewound, removed, and properly labeled
with the following items:
o Date o Shift number
o City o Chase vehicle number
o Route number o Tape I. D. Number
o Trip numbers completed o Comments.
SCOTT RESEARCH LABORATORIES, INC
-------
3-42
SRL 2922-13-1271
The same information was recorded in log books at the end of each day by
the survey operations supervisors. The log books thus provided a comprehensive
inventory of all data collected.
Following the break, the tape recorder heads and rollers were
cleaned and a second reel of tape mounted. Data collection then resumed
until the survey route was completed, usually requiring about six full
hours of data collection per shift. The second tape was removed and
labeled and most of the instrumentation sub-systems were shut down. After
completion of the morning shift, the clock system remained active to eliminate
the necessity for resetting by the afternoon shift survey crew. The
afternoon shift began collecting data at approximately 2:00 p.m., using
the same operating procedures. Shift starting times were adjusted as
required to prevent the break periods from occurring at exactly the same
time each day. Under extremely heavy traffic conditions, it was sometimes
necessary to use three reels of tape during one shift of data collection.
A few minor electronic and mechanical problems were observed in
the field. On these infrequent occasions, the back-up survey vehicle was
placed in operation and the problems corrected as rapidly as possible. Among
the survey crew members were a mechanic and an electronic technician whose
skills enabled the survey operations to remain on schedule at all times.
All recorded tapes were logged in by the supervisors and carefully packaged
for air freight delivery to Scott's San Bernardino laboratory every three or
four days.
SCOTT RESEARCH LABORATORIES, INC
-------
3-43
SKL 2922-13-1271
3.4.4 Vehicle Operations Survey on the LA-4 Route
As an auxiliary task to the basic VOS program, vehicle operating
patterns were determined over the LA-4 road route for comparison with the urban
Los Angeles (LA-Composite) driving pattern. The LA-4 road route, designed
in the late 1960's and located within a 6-mile radius of central Los Angeles,
is shown in Figure 3-16. Vehicle operating patterns on the route,
during the weekday hours of 9:00 a.m. to 11:00 a.m. and 1:00 a.m. to 3:00 p.m.,
were intended to typify morning rush-hour driving patterns for central
Los Angeles.
Although the surveys of vehicle operating patterns on the
LA-Composite routes and the LA-4 route were operationally similar in many
respects, there existed certain differences between the two surveys. The
studies were similar with respect to the following items:
o Use of chase-car technique
o Survey-vehicle performance capabilities
o Instrumentation systems
o Operating personnel
o Data format
o Data processing program.
Differences between the two studies were primarily associated
with the overall scope of operations, route structure, and data collection
time periods. Table 3-6 presents the details of these differences. Vehicle
operating data were collected during the "defined hours" of 9:00 a.m. to
11:00 a.m. and 1:00 p.m. to 3:00 p.m., and during the "off-hours" of 7:00 a.m.
to 9:00 a.m. and 3:00 p.m. to 5:00 p.m. (to provide additional information
on the LA-4 route).
SCOTT RESEARCH LABORATORIES, INC
-------
90
W
w
SO
i
2
55
Olympic
s
(D
4J
CO
£
Exposition
Start
5th
San
Pedro
1. Start behind County Air
Pollution Laboratory.
2. Right on Fifth.
3. Right on San Pedro.
4. Left on Third St.
5. Up ramp to Harbor Fy.
South.
6. Exit Harbor Fy. at Expo-
sition.
7. Follow Exposition to
Western.
8. Right on Western.
9. Right on Olympic.
10. Left on Santee.
11. Right on Ninth.
12. Left on San Pedro.
13. End just beyond inter-
section San Pedro and
Fifth Street.
N3
vo
N5
t-0
Figure 3-16 LA-4 Road Route Description
-------
SO
PI
£
Table 3-6
Differences Between LA Composite and LA-4 Survey Parameters
1. Objectives associated with route(s)
2. Route Locations
3. Number of Routes Surveyed
LA Composite
To provide a road course
over which vehicle operating
behavior can be surveyed to
determine overall vehicle
operating characteristics
for metropolitan Los Angeles
Within 25-mile radius of
Central Los Angeles
4 - Weekday
4 - Weekend
LA-4
To provide a road course
that will typify vehicle
operating behavior during
the morning, peak-hour traffic
in Central Los Angeles in
the fall
Within 6-mile radius of
Central Los Angeles
1 - Weekday
to
vo
ro
I
M
NJ
4. Number of Replications per Route
10 - Weekday
4 - Weekend
84
5. Route Length (nominal)
6. Total Survey Mileage (nominal)
7. Survey Time Period
8. Traffic Density Weighting
145 miles
7,850
6:00 A.M. to 10:00 P.M.,
7 Days Per Week
Traffic density values based
on time of day were used to
weight results during computer
processing
13 miles
1,100
Defined Hours Were:
9:00 A.M. to 11:00 A.M. and
1:00 P.M. to 3:00 P.M., Tuesday
Wednesday and Thursday
Considered to be self-weighting
-------
3-46
SRL 2922-13-1271
The basic results of this investigation are presented in Section
4.0 of this report. A complete discussion of the LA-4 survey and compari-
son of results with the LA-Composite survey is contained in Reference 10.
3.5 DATA PROCESSING AND ANALYSIS TECHNIQUES
The large volume of data collected during the VOS Program
necessitated fast, efficient data processing procedures. The basis for these
procedures and descriptions of the various techniques used will be discussed
in this section. The initial approach to data processing will be presented
first, followed by a discussion of the modifications that were made to the
data processing program in Phase II. Descriptions of the final computer
programs as well as the computer facilities, are presented. Since data
from various geographical locations were processed, techniques were developed
to provide indicators of the similarity between the vehicle operating
characteristics for the data collected from these different areas.
3.5.1 Data Processing Techniques
The large volume of data collected necessitated highly
reliable data acquisition equipment in addition to fast, efficient data
processing procedures. The groundwork for successful data processing was
established by using a dependable, accurate Digital Data Acquisition
System (DBAS) and recording the vehicle operating data on high-quality magnetic
tape. Representation of vehicle operating modes (an operating mode is de-
fined here to be a distinct event; i.e. an idle, cruise, acceleration,
or deceleration characterized by an initial speed, final speed, and duration)
in matrix format required a large-core, fast-operating computer. The procedure
SCOTT RESEARCH LABORATORIES, INC
-------
3-47
SRL 2922-1271
to establish the driving modes that represent vehicle operation was the
primary task of the data processing effort and will be described in detail in
this section.
3.5.1.1 Initial Data Processing Approach
The research nature of this program required that flexible
procedures be used in the data processing so that, if evaluation of the
results indicated a modification was needed, this could easily be integrated
into the data processing technique with as little effort as possible.
The preliminary data processing techniques described here were developed
such that, after initial processing of some data and analysis of the
results, required modifications were easily incorporated to make the
results more representative and useful.
As discussed above, the primary objective of the VOS program
was to obtain data which could be used to describe the operating char-
acteristics of vehicles in various large American cities. In order that
these characteristics be representative of the driving from which the
VOS sample was taken, two criteria must be met. First, sufficient
information must be recorded so that the manner in which vehicles are
driven during each operating mode may be determined. This was made
possible by recording vehicle operating data at precise one-second
intervals. Therefore, the operating characteristics may be shown at one-
second intervals during all of the operating modes. This includes cal-
culating the duration of each mode. Secondly, the frequency of occurrence
of the operating modes must be determined. Since each VOS vehicle emulated
SCOTT RESEARCH LABORATORIES, INC
-------
3-48
SRL 2922-13-1271
the operation of only one other vehicle, traffic density weighting factors
developed by SDC were applied to the observed mode frequencies to make the
data representative of all the traffic.
To fulfill the objectives of this program, time data, data per-
taining to vehicle operation, and various auxiliary data were recorded
for later analysis. Time parameters were: day of the year, hour, minute,
and second. Vehicle operating parameters recorded were: speed, in tenths
of a mile per hour; manifold vacuum, in tenths of an inch of mercury
guage; and.a code indicating the type of vehicle being chased. The
auxiliary data consisted of code numbers for: weather conditions, type
of road being driven, and route, trip, and segment numbers for each
city. Each parameter was sampled once per second, and the data so obtained
were recorded in 7-track format at 200 bpi in a 27-character record on
magnetic tape. The order in which the data were recorded and the number
of characters for each parameter are shown in the following table. The
range of possible values for each parameter is also shown.
Number of
Parameter Characters Range
Day of the year 3 0 to 365
Hour 2 0 to 23
Minute 2 0 to 59
Second 2 0 to 59
Fleet Number 3 0 to 999
Speed 3 0 to 70.0 mph
Manifold Vacuum 3 0 to 30.0 inches of
Weather Type 2 0 to 99
Road Type ! Q tQ g
Segment Number 2 0 t Q
Trip Number 2 0 to 99
Route Number 2 0 to 99
Total 27
SCOTT RESEARCH LABORATORIES, INC
-------
3-49
SRL 2922-13-1271
A 27-character record of 9's was used as a special indicator for the end-
of-tape mark on each reel.
The primary VOS data for determining representative vehicle
operation are obtained from the speed and time parameters. Since a
continuous sequence of speeds at one-second intervals is cumbersome to
work with and difficult to analyze, a scheme was developed to partition
the sequence of speeds into a series of three distinct types of vehicle
operating modes. These mode types are: cruises (including idles),
accelerations, and decelerations. The distinguishing characteristics of
each mode are its initial speed, duration in seconds, and final speed.
The initial and final speeds of a mode are used to determine which of the
three mode types is appropriate. The mode-selection logic for segmenting
a sequence of speeds is shown schematically in Figure 3-17.
These mode criteria may be thought of as follows. Assume
the vehicle whose operation is being observed begins cruising at some
speed, s. A 0.5-mph band is established on either side of the initial
cruise speed s. Successive speeds are then analyzed until a speed is
observed to be outside of the 1-mph band. The last speed inside of the
band marks the end of the cruise mode and also the beginning of the next
mode. A cruise mode was also terminated in the original program when
successive speeds differed by more than 0.2 mph. The mode duration is
obtained from the difference in the times corresponding to the final and
initial speeds.
The mode following a cruise was an acceleration if the cruise
ended when a speed exceeded the upper band and was more than 0.2 mph greater
SCOTT RESEARCH LABORATORIES, INC
-------
3-50
SRL 2922-13-1271
S First
speed for
new mode
Next speed
within ± .5 nph
of first speed?
Mode is
cruise
Accumulate time
in cruise mode
Is speed above
(below) 1 mph
cruise band?
Accumulate
frequency
for mode
1
New mode is
acceleration
(deceleration)
[Accumulate time
in acceleration
.(deceleration)
Calculate
acceleration
(deceleration)
rate
Accumulate
frequency for
cruise
Beginning of
new mode and
end cf cruise
T
Figure 3-17 Mode-Selection Logic
SCOTT RESEARCH LABORATORIES, INC
-------
3-51
SKL 2922-13-1271
than the previous speed. The new mode was a deceleration if a speed lower
than the lower band was observed and was also more than 0.2 mph less than
the preceeding speed. For cases where the 0.2-mph criterion was not met,
two successive cruise modes occurred. The final speed of the first cruise
became the initial speed of the second cruise and was, therefore, the
speed about which the one-mph band was established. Once an acceleration
(deceleration) mode had been established, successive speeds were analyzed
to determine if each were greater (less) than the preceeding speed by more
than 0.2 mph. Failure to meet the 0.2-mph criterion ended the acceleration
or deceleration mode. The duration was calculated the same as for the
cruise.
Figure 3-18 shows a sequence of speeds to which the previously
described technique was applied. The initial speed, final speed, and
duration are shown for each mode. Note that the 0.5-mph cruise band for
an initial cruise speed of zero (idle) is meaningful only for positive
speeds. Deficiencies in the initial program logic described above were
identified during the initial processing of LA data and corrected as dis-
cussed below.
Applying these mode criteria to a sequence of speeds enables
us to reduce a cumbersome list of speeds to a sequence of modes uniquely
defined by their initial and final speeds and durations. All that is
lacking is a bookkeeping procedure to allow accumulation of a large volume
of data. The mode-matrix approach to the representation of vehicle operating
data provides the needed procedure, The basis for this technique is a
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-52
3.0
35
6 second
cruise at
0.25 mph
7 second
cruise at
0.25 mph
5 second
accel
from
0. mph to
2 . 5 mph
9 second
cruise at
2. 35 mph
5 second
decel
from
2.2 mph
to 0. mph
NOTE: Shaded area represents ± 0-5 mph cruise band about initial cruise
speed.
Figure 3-18 Segmentation of Speed/Time Data Into Modes
Using Preliminary Mode Logic
SCOTT RESEARCH LABORATORIES, INC
-------
3-53
SRL 2922-13-1271
square grid, as shown in Figure 3-19, with initial speed along the vertical
axis and final speed along the horizontal axis. The fineness of the grid
structure was determined by analysis of our own driving habits and the
practicality of working with matrices of reasonable size. This and the
speed laws of many states led to the initial and final speed axes being
laid out in 70 one-mph grid cells. Each cell thus represents one of the
4900 modes which are possible with this matrix approach. As shown in
Figure 3-19, the cells along the diagonal form the cruise modes, the cells
above this diagonal are the acceleration modes, and the cells below the
diagonal are the deceleration modes.
The master mode matrix, is produced by laying out a time axis
as the third dimension of the basic mode matrix, as shown in Figure 3-20.
The time axis is laid out in 20 cells as follows: the first 15 cells are
each one second wide, the next four cells are each five seconds wide, and
cell number 20 includes time durations of 35 seconds and greater.
As the recorded VOS data were processed into modes, the appropri-
ate traffic density weighting factor was added to the frequency for each
mode in the time cell corresponding to the mode duration. A master mode
matrix was developed for the data collected each time a route was driven.
The number in each cell is the cummulative weighted frequency of
occurrence of that particular mode for the indicated duration. The
master mode matrix thus yields the distribution of time in mode for
each mode. This distribution is obtained from the third dimension
(time axis) for each mode. Plotting the cell frequencies at the cor-
responding times shows this distribution graphically. Following the
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-54
FINAL SPEED, mph
INITIAL
SPEED ,
mph
DE
=.
1
«
101
«
)EE
n
)N
ACCELERATION
MODES
\
\
Figure 3-19 Matrix Representation of Vehicle Operation
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-55
INITIAL
SPEED,
mph
FINAL SPEED, mph
70 x 70 x 20 MATRIX
EACH CELL CONTAINS THE
ACCUMULATED WEIGHTED
MODE FREQUENCY FOR MODES
OF THE CORRESPONDING
DURATION. NOTE THAT THE
THIRD DIMENSION FOR
EACH MODE PROVIDES A
DISTRIBUTION OF TIME IN
MODE.
Figure 3-20 Master Mode Matrix
SCOTT RESEARCH LABORATORIES, INC
-------
3-56
SRL 2922-13-1271
mode analysis of the data for a route replicate, the master mode matrix
is used to calculate the total-time-in-mode matrix and the mode-frequency
matrix.
The total-time-in-mode matrix, formatted as shown in Figure 3-21,
is obtained from the master mode matrix by summing the product of each cell
frequency and the time corresponding to that cell for each mode. Each entry
in the total-time matrix is thus the total weighted number of seconds of
vehicle operation in that particular mode. The mode-frequency matrix,
formatted as shown in Figure 3-21, is obtained from the master mode matrix
by summing the frequencies in the time-axis cells for each mode. Each entry
in the mode-frequency matrix is therefore the total weighted number of times
a particular mode is executed.
Division of each entry in the total-time matrix by the corres-
ponding entry in the mode-frequency matrix produces the average-time-in-mode
matrix. The average-time-in-mode matrix, also formatted as shown in
Figure 3-21, contains the average number of seconds required to execute
each mode.
The mode-frequency matrix may be transformed to restore sequence
(in a probabilistic sense) to the data. Each row of the mode-frequency matrix
represents all modes that have the indicated initial speed. Therefore, by
dividing each non-diagonal row entry by the sum of all non-diagonal frequencies
for the row, we obtain the conditional probability that the next speed
transition will end at the speed indicated by the column heading. This matrix
is called the transition-probability matrix and is formatted as shown in
Figure 3-22. Note that each row of the transition-probability matrix sums
SCOTT RESEARCH LABORATORIES. INC
-------
3-57
SRL 2922-13-1271
FINAL SPEED, mph
INITIAL
SPEED,
mph
3 - 70 x 70 MATRICES
TIME IN MODE:
EACH CELL CONTAINS TOTAL WEIGHTED TIME
IN MODE.
MODE FREQUENCY:
EACH CELL CONTAINS TOTAL WEIGHTED
MODE FREQUENCY.
AVERAGE TIME IN MODE:
EACH CELL CONTAINS THE AVERAGE TIME IN
MODE IN SECONDS.
Figure 3-21 Vehicle Operation Mode Matrices
SCOTT RESEARCH LABORATORIES. INC
-------
3-58
SRL 2922-13-1271
INITIAL
SPEED,
mph
FINAL SPEED, mph
70 x 70 MATRIX
THE CELLS OF EACH ROW ARE THE
CONDITIONAL PROBABILITIES OF THE
NEXT MODE ENDING AT THE PARTICULAR
FINAL SPEED ASSUMING OPERATION AT
THE INITIAL SPEED. (THIS MATRIX
IS NOT DEFINED FOR THE DIAGONAL
CRUISE MODE CELLS)
Figure 3-22 Transition Probability Matrix
SCOTT RESEARCH LABORATORIES, INC
-------
3-59
SRL 2922-13-1271
to unity, since all (different) final speeds are possible for each initial
speed. A sequence of modes may thus be generated with random number sampling
based on the transition probabilities.
These four matrices (total time, mode frequency, average time in
mode, and transition probability) form the basis for describing vehicle
operation and will be referred to as the mode matrices. Together with the
development of the master mode matrix, however, auxiliary data are accumulated
into several supplementary matrices. The information obtained from these
matrices is used in conjunction with the mode matrices to provide an overall
description of representative vehicle operation for the observed city.
Each route in a city was composed of a sequence of typical short
trips whose lengths are representative of the distribution of trip lengths
observed in that city. The trip distance is calculated by numerical
integration of the speed/time data. The time required to drive each trip
is accumulated as the data are being analyzed. The thumbwheel switch
indicating the trip number is sampled once each second and recorded on
the magnetic tape so that the end of a trip is noted each time the trip
number changes. Dividing the trip distance by the time required to drive
the trip yields the average speed for the trip. These data are accumu-
lated by average trip speed, trip distance, and time of day in the 3-
dimensional matrix shown in Figure 3-23. The time-of-day slots are
indicative of the various traffic conditions that exist throughout the
day and are included to provide information on the variability of
average speed for trips of the same distance. These time-of-day slots,
which are different for weekdays and weekends, are shown in the table
below.
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-60
TIME
OF DAY,
hours
TRIP LENGTH, miles
8 x 8 x 30 MATRIX
EACH CELL CONTAINS THE FREQUENCY
OF THE APPROPRIATE AVERAGE SPEED
FOR TRIPS OF EACH LENGTH DURING
THE VARIOUS TIMES OF THE DAY.
Figure 3-23
Matrix of Average Speed Versus
Trip Length and Time of Day
SCOTT RESEARCH LABORATORIES, INC
-------
3-61
SRL 2922-13-1271
WEEKDAYS WEEKENDS
0000 - 0600 Hours 0000 - 0830 Hours
0600 - 0900 Hours 0830 - 1100 Hours
0900 ^ 1130 Hours 1100 - 1300 Hours
1130 - 1330 Hours 1300 - 1800 Hours
1330 - 1630 Hours 1800 - 2100 Hours
1630 - 1830 Hours 2100 - 2400 Hours
1830 - 2100 Hours
2100 - 2400 Hours
The trip-length axis is partitioned into seven cells, each
covering an interval of three miles, and an eighth cell covering six miles, to
accomodate trips from 21 to 27 miles. The average speed axis of this
matrix covers a speed range from 0 to 60 mph, divided into 20 cells of
three mph each. As the data for each trip are analyzed, tally marks are
accumulated in the appropriate cells determined by the time of day,
trip length, and average speed. This matrix of average speeds for various
trips supplements the mode matrices by providing insight into the variability
of average speed with trip length and time of day. The time-of-day factor
will reflect traffic density influences on vehicle operation.
The acceleration-rate-versus-instantaneous-speed matrix, shown
in Figure 3-24, is another auxiliary data matrix of interest. The speed
axis is partitioned into 70 one-mph cells. The acceleration-rate axis is
composed of 80 one-quarter-mph/second cells between plus and minus 10
mph/second. During the mode analysis of the data for a route replicate,
acceleration rates are calculated from every pair of adjacent speeds for all
SCOTT RESEARCH LABORATORIES, INC
-------
3-62
SRL 2922-13-1271
ACCELERATION
RATE,
mph/second
INSTANTANEOUS SPEED, mph
70 x 60 MATRIX
EACH CELL CONTAINS THE ACCUMULATED
WEIGHTED FREQUENCY FOR THE PARTICULAR
ACCELERATION RATE AND INSTANTANEOUS
SPEED.
Figure 3-24 Acceleration-Rate-Versus-Instantaneous-Speed Matrix
SCOTT RESEARCH LABORATORIES, INC
-------
3-63
SRL 2922-13-1271
acceleration and deceleration modes. The appropriate traffic density
weighting factor is then accumulated in the matrix cell corresponding
to the calculated acceleration rate. The instantaneous speed is the
average of two consecutive speeds.
The acceleration-rate-versus-instantaneous-speed matrix does
not provide any information about the shape of a specific acceleration or
deceleration profile; the data are combined by instantaneous speed, in-
dependently of the initial and final speeds. Second-by-second speed
data were therefore obtained on computer-punched cards to obtain data
from which typical acceleration and deceleration profiles may be con-
structed. Various pairs of initial and final speeds were specified
for which punched-card output was obtained.
Modes were selected on the basis of frequency of occurrence and
to provide a representative sample of the 4900 modes in the mode matrix.
One-half-mph tolerences are placed about the initial and final speeds and
punched-card output is obtained for any selected mode whose initial and
final speeds are within the speed bands specified. These punched cards
may be used to create average characteristics for typical modes to provide
insight into the variation of acceleration and deceleration profiles with
geographical location.
The manifold-vacuum-versus-cruise-speed matrix, shown in Figure 3-25,
is used to accumulate manifold-vacuum data for the various cruise modes. The
manifold-vacuum scale extends from 0 to 30 inches of mercury gauge and is
divided into 30 cells, each covering a range of one inch of mercury. A
tally is accumulated in the appropriate cell for each second of observed
cruise data. The data in this matrix are unweighted with respect to traffic
SCOTT RESEARCH LABORATORIES, INC
-------
3-64
SRL 2922-13-1271
CRUISE SPEED, mph
MANIFOLD
VACUUM,
in. Hg
gauge
70 x 30 MATRIX
EACH CELL CONTAINS THE ACCUMULATED
UNWEIGHTED FREQUENCY FOR CRUISE
OPERATION AT THE CORRESPONDING
MANIFOLD VACUUM.
Figure 3-25 Manifold Vacuum Versus Cruise Speed
SCOTT RESEARCH LABORATORIES, INC
-------
3-65
SKL 2922-13-1271
density because the recorded manifold-vacuum data are peculiar to the VOS
vehicles. Only the cruise^node data are represented in this matrix because
of the relative stability of manifold vacuum for those modes.
The matrices described above provide a sound basis for characterizing
vehicle operation. Since all matrices except the manifold-vacuum matrix
have been weighted by the appropriate traffic density factors, these
matrices provide an accurate representation of typical vehicle operation
in the city for which the route was designed. Each of these matrices
is calculated for every route replicate.
Since a complete set of matrices was determined for the data
observed during each route replicate, a matrix merging technique was de-
veloped to combine any number of the same kind of matrix. In general,
vehicle operation data were observed as each basic route was driven 14 times;
i.e., replications were obtained twice per day for each of the five weekdays
and twice per day for each of the two weekend days. The merging technique
thus provides a complete set of matrices which represent vehicle operation
over each route.
Merging of the mode matrices is limited to the total-time
matrix and the mode-frequency matrix. The composite total-time matrix is
obtained by adding the data from the corresponding cells for all route re-
plicates being merged and dividing by the number of route replicates
being merged. Let a1±j be the entry for the mode having initial speed i
and final speed j for the first route replicate total-tune matrix. Similarly,
let a through a have the same meaning for route-replicate matrices
& nij
2 through n. The corresponding entry in the composite matrix, denoted
by c . . , is defined by
SCOTT RESEARCH LABORATORIES. INC
-------
3-66
SRL 2922-13-1271
k=l
Performing this calculation over all i and j completes the com-
posite total-time matrix. Division of each cell in the composite by the
number of matrices being merged is required by computer storage limitations;
the division keeps each cell entry relatively small. The composite mode-
frequency matrix is obtained from the route-replicate mode-frequency matrices
in the same manner as was the composite total-time matrix. The average-
time-in-mode matrix and the transition-probability matrix are then cal-
culated as previously described. These matrices are recalculated from the
total-time and mode-frequency composites to eliminate inconsistencies in
the composite mode matrices that will arise if the average-time-in-mode
and the transition-probability composites are obtained directly by merging.
Composites for the auxiliary matrices are obtained by summing
the data in corresponding cells for the same type of matrix for each route
replicate being merged. This is performed in the same way as matrix merging
for the total-time matrix. The only composite matrices whose elements are
not divided by the number of matrices merged are the average-speed-versus-
trip-length-and-time-of-day matrix and the manifold-vacuum-versus-cruise-
speed matrix. The division is not necessary because these matrices contain
actual unweighted frequencies that are relatively small, even after merging.
SCOTT RESEARCH LABORATORIES, INC
-------
3-67
SRL 2922-13-1271
Higher levels of matrix merging may also be performed in the
same fashion on several of the route composites to ohtain a city composite.
Lower-level composites, such as weekday and weekend composites, for the
data obtained while observing vehicle operation over a specific route are
also of some interest.
Creation of a sequence of vehicle operating modes from the speed/time
data and subsequent storage and accumulation of these data into a set of mode
and auxiliary matrices have provided a compact, easy-to-manipulate representation
of driving patterns. However, it is clear that even though these matrices,
containing as many as 4900 numbers, are easily handled on a computer,
additional simplification is required for us to digest thoroughly what the
data has to offer. The most straightforward approach to simplification is
by matrix compacting. This procedure consists of restructuring the mode
matrices to reduce the number of modes, while still retaining the basic
characteristics inherent in the data represented by the 70 x 70 mode matrices.
Because of built-in general features in the computer programs, the restructuring
of the initial-speed and final-speed axes of the mode matrices may be
accomplished in a variety of ways. The most practical form, based on matrix
size and convenience of the resulting mode speeds, results when the initial-speed
and final-speed axes are each partitioned into 14 intervals. These are defined
as follows: one interval from 0 to 2.5 mph, 12 intervals of 5-mph width, and
one interval which is 7.5 mph wide, thus providing an overall speed range
from 0 to 70 mph.
Figure 3-26 shows the compact 14 x 14 mode matrix grid over-
laid on the 70 x 70 mode matrix. As with matrix merging, only the
total-time-in-mode matrix and the mode-frequency matrix are compacted; the
SCOTT RESEARCH LABORATORIES, INC
-------
3-68
SRL 2922-13-1271
FINAL SPEED, mph
INITIAL
SPEED,
mph
NOTE: LIGfcT LINES REPRESENT FINE MODE MATRIX GRID STRUCTURE (70 x 70)
EEATi LINES REPRESENT COMPACTED MODE MATRIX GRID STRUCTURE
(14 x 14).
Figure 3-26 Compact Matrix Grid Overlaid On 70 x 70 Mode
Matrix Grid Structure
SCOTT RESEARCH LABORATORIES, INC
-------
3-69
SRL 2922-13-1271
average-time-in-mode and transition-^prohahility matrices are then recal-
culated. Each cell of the compact matrix is obtained by adding the data from
all of the cells in the 70 x 70 matrix that lie within the cell from the
compact matrix (Figure 3-26). With the exception of those cells which lie
around the perimeter of the mode matrix, the data from 25 cells in the
large mode matrix are contained in one cell of the compact matrix. It should
be noted that all data are retained in this compacting procedure and that
only the fineness of the details is modified. Patterns and characteristics
representative of the data are more distinguishable in the compact matrices
and the data are now in a manageable form for easy manipulation and for
reporting.
3.5.1.2 Assessment and Modification of Preliminary Approach
Application of the data processing techniques described in
Section 3.5.1.1 to VOS data collected in Los Angeles provided a basis
for evaluation of the procedures. This checkout was essential, of course;
further, the data processing computer program logic is of such a nature that
a large amount of data is required to execute all of the options in the
programs. Assessment of output based on a realistic sample of data
afforded an opportunity to evaluate the results for inconsistencies and
presented an opportunity to determine if the mode-matrix approach adequately
provided an accurate description of vehicle operation. As a result of that
analysis, various modifications and additions were made to the data processing
techniques. This section contains a summary of the reasons for making
modifications to the data processing techniques as well as a description of
the modifications.
SCOTT RESEARCH LABORATORIES. INC
-------
3-70
SRL 2922-13-1271
The large amount of computer output from processing VOS data
required that the printed labels and general formatting he improved to
make the matrices easier to identify and to make the output more readable.
This reformatting was especially important with respect to the compacted
mode matrices so that the hardcopy output from the computer could be
easily reproduced for presentation in reports.
Analysis of a segment of recorded speed/time data revealed that the
initial one or two seconds of acceleration or deceleration data were sometimes
included with the previous cruise mode. This was eliminated by the revised
mode logic shown in Figure 3-27. As an acceleration or deceleration mode
is detected following a cruise mode, the logic regresses backwards in time
to find the actual beginning of the data that satisfy the acceleration/
deceleration criteria. Time is subtracted from the preceding cruise mode in
the same amount as any time added to the acceleration or deceleration mode.
The effect of this processing logic modification is shown in Figure 3-28,
where the same speed/time data that were analyzed in Figure 3-18 are segmented
into modes using the revised logic. The overall effect is to yield slightly
shorter cruise modes and slightly longer acceleration and deceleration modes.
As noted above, the initial processing logic was such that multiple
cruise modes could occur in sequence for speed data with small oscillations.
This condition exists when a cruise mode is terminated by a speed outside
the one-mph band, but is not sufficiently greater or less than the previous
speed to satisfy the acceleration/deceleration criterion. These consecutive
cruise modes, called step cruises, are actually part of a broader-band cruise
that was segmented by the detail of the processing logic. This condition
causes a serious problem when matrix compacting is attempted.
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922-13-1271
3-71
First
speed for
new mode
Next
speed within
±.5 mph of first
speed?
Mode is
cruise
Accumulate time
in
cruise modes
Is speed above
(below) 1 mph
cruise band?
Accumulate
frequency
for mode
New mode is
acceleration
(deceleration)
(Subtract one \
second from L
duration of j
previous cruise /
No
/
Accumulate time
I in acceleration
I (deceleration)
Calculate
acceleration
(deceleration)
rate
Accumulate
frequency for
cruise
(Proceed back
f accel (decel)
criteria is
satisfied.
r
*h
Beginning of \
new mode and I
end of cruise /
Figure 3-27 Modified Logic for Segmenting Speed/Time Data Into Modes
SCOTT RESEARCH LABORATORIES, INC
-------
3-72
3.0
2.0
45
P.
6
3 1.5
1.0
0.5
s 20
TIME, seconds
25
3U
35
6 second
cruise at
0.25 mph
7 second
cruise at
0.25 mph
5 second
accel
from
7 second
cruise at
2.65 mph
7 second
decel from
2.8 mph to
0. mph to 0. mph
2 . 5 mph
NOTE: Shaded areas represent ±0.5 mph cruise band about initial cruise
speed.
Figure 3-28 Segmentation of Speed/Time Data Into Modes
Using Modified Mode Logic
SCOTT RESEARCH LABORATORIES, INC
-------
3-73
SSL 2922-13-1271
It is noted from Figure 3-26 that merging of the data to form
the cruise modes in the compact matrices requires combining data that were
classified as accelerations, cruises, and decelerations using the detailed
70 x 70 grid structure. All other modes being merged are distinctly
accelerations or decelerations regardless of the grid structure that
is selected for the compact matrices. The existence of step cruises and
the process of combining all three types of modes into the cruise modes
of the compact matrices resulted in excessively large cruise frequencies
and, therefore, very small average times for the cruise modes. This
problem was eliminated by imposing the reasonable logic that each acceler-
ation or deceleration should be followed by one and only one cruise mode.
As a result, the total number of accelerations and decelerations
in a given row must equal the total number of cruises for that row. After
initial processing to yield the mode-frequency matrix, the mode frequency
for each cruise is replaced by the sum of all non-diagonal frequencies in
the row in which the cruise occurs. This program modification had its
greatest impact on the preliminary results in that the average-time-in-mode
matrix became realistic with respect to the cruise modes. It should be
noted that the resulting mode-frequency matrix is composed of 50% cruises
(including idles) and 50% accelerations and decelerations. The modified
transition-probability matrix is thus calculated from the mode-frequency
matrix as before, except that each row is normalized by dividing each non-
diagonal entry by the sum of all non-diagonal frequencies. The
transition probability is undefined for the cruise cells because a cruise
does not constitute a transition.
SCOTT RESEARCH LABORATORIES, INC
-------
3-74
SRL 2922-13-1271
Each set of mode matrices is created from a sample of vehicle
operation data that is of variable size due to the fact that a different
time is required to drive each route replicate. This causes difficulty
in comparing the total-time-in-^mode matrices and the mode-frequency
matrices for different sets of data. This problem was easily resolved by
creating the two normalized mode matrices shown in Figures 3-29 and 3-30.
A normalized matrix is created simply by dividing each cell entry by the
sum of all cells and expressing the result as a percentage. The entire
matrix thus sums to 100 percent, since this represents all of the data.
These two normalized matrices provide a consistent basis for comparing
vehicle operation observed for any sets of data.
Following analysis of the data for a route replicate the normalized
matrices are output on magnetic tape and on the printer together with the other
mode matrices and auxiliary matrices. Four useful summary statistics are
obtained for each of the normalized time-in-mode and normalized mode-frequency
matrices. Summary percentages are calculated for idles, cruises, accelerations,
and decelerations by accumulating the data from the cells for each mode type.
Matrix merging and compacting was an inefficient process using
the initial techniques because computer core-storage limitations required
relatively slow access to disk storage for each operation. Switching to
a larger, more powerful computer allowed these operations to be performed
in core with a great saving of computer time. This modification to the
program did not affect the results, of course.
SCOTT RESEARCH LABORATORIES, INC
-------
3-75
SRL 2922-13-1271
FINAL SPEED, mph
INITIAL
SPEED,
mph
EACH CELL CONTAINS THE PERCENT OF
VEHICLE OPERATING TIME FOR THAT MODE.
NOTE THAT THE ENTIRE MATRIX SUMS TO
100 PERCENT.
Figure 3-29 Normalized Time In Mode Matrix
SCOTT RESEARCH LABORATORIES. INC
-------
3-76
SRL 2922-13-1271
FINAL SPEED, mph
INITIAL
SPEED,
mph
EACH CELL CONTAINS THE PERCENT
OF ALL MODE OCCURRENCES OF THAT
TYPE. NOTE THAT THE ENTIRE MATRIX
SUMS TO 100 PERCENT.
Figure 3-30 Normalized Mode Frequency Matrix
SCOTT RESEARCH LABORATORIES, INC
-------
3-77
SRL 2922-13-1271
Average speed and mileage calculations during the preliminary data
processing were made using the lower Darboux sum for accelerations and the
upper Darboux sum for decelerations. Any inaccuracy resulting from this
technique was eliminated by modifying the program to compute the Riemann sum.
This more accurate computation of average speed and distance data was
necessary because of another matrix that was developed to accommodate time
and distance data. Figure 3-31 shows the 3-dimensional structure of the
matrices in which were accumulated time, distance, and average speed versus
road type and time of day. The time-of-day axis is partitioned into either
6 or 8 time slots, depending on whether the data are for weekends or week-
days, respectively. "Freeway" and "non-freeway" designate the two road
type classifications for which data are obtained. The average-speed data
from this matrix, in conjunction with the summary percentage statistics
from the normalized time and normalized frequency matrices, present a con-
cise description of typical vehicle operation in the city in which the data
were obtained.
3.5.1.3 Computer Processing
The large volume of VOS data required that efficient computer
processing be used to analyze the data. The requirement for good reliable
data was ensured by the accurate, dependable DDAS and the use of high-
quality magnetic tape. Transformation of the logic for developing the mode
matrices into computer language was also important to ensure efficient
and effective use of the capabilities of the computer. In addition, it
was important to select a computer capable of performing all the intricate
functions required to process the VOS data from magnetic tapes and to
output the results on magnetic tape and on hardcopy.
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-78
TIME
OF DAY,
hours
ROAD TYPE
EACH CELL IS THE
DISTANCE TRAVELED ON
THE PARTICULAR ROAD
TYPE DURING THAT TIME
PERIOD.
TIME
OF DAY,
hours
ROAD TYPE
EACH CELL IS THE
NUMBER OF HOURS VEHICLE
OPERATION IS OBSERVED
ON THE PARTICULAR ROAD
TYPE DURING THAT TIME
PERIOD.
TIME
OF DAY,
hours
ROAD TYPE
EACH CELL IS THE
AVERAGE SPEED FOR THE
PARTICULAR ROAD TYPE
DURING THAT TIME
PERIOD.
Figure 3-31 Time, Distance And Average Speed Versus Time
Of Day And Road Type Matrices
SCOTT RESEARCH LABORATORIES, INC,
-------
3-79
SRL 2922-13-1271
The complete VOS data processing program, from initial raw
data input to the final computer output, was modularized into logical
units to guarantee easy computer program writing, debugging, and execution.
Each computer program performs an integral task so that no overlapping
occurs during processing. Partitioning the processing into separate tasks
also provided a firm basis for evaluating the progress of the analysis as
each task was performed. Since each task depends on successful completion
of the preceeding task, it is important to have complete control of all
stages of the data processing.
The following is a list of the computer programs that provide a
complete analysis of VOS data:
o Data format conversion from 7 to 9 track
o Data processing (matrix development)
o Matrix merging and compacting
o Matrix printing
o Matrix comparison.
Flow diagrams showing the relationships between these computer programs
are presented in Figures 3-32 and 3-33. All programs, except that pro-
viding raw data format conversion from seven to nine track, were written
in Fortran language. Fortran is a high-level lenguage that provides
efficient matrix operations as well as being versatile in all aspects of
magnetic tape operations. The data conversion program was written in the
relatively low level assembler language to make use of its capabilities
for handling raw input data at low recording densities.
SCOTT RESEARCH LABORATORIES, INC
-------
§
SO
s
SO
o
/-TRACK
RAW DATA
TAPES
DATA FORMAT
CONVERSION
AND REBLOCK
9-TRACK
INTERMEDIATE
DATA
TAPE
EDIT DATA
AND
REBLOCK
ESTABLISH
MODES
ESTABLISH
MATRICES
9-TRACK
CLEAN
DATA
TAPE
TYPICAL
ACCELS 8
DECELS
ROUTE
REPLICATE
MATRIX
TAPE
70
r1
VO
to
ro
I
H-1
IO
I
CD
O
|
L_
_J
Figure 3-32 VOS Data Conversion and Data Processing Computer Programs
-------
I 1
30
pi
r>
s
99
1
ROUTE
REPLICATE
MATRIX
TAPE
MERGE AND
COMPACT
MATRICES
ROUTE/
CITY
MATRIX
TAPE
MATRIX
PRINT
PROGRAM
MATRIX
PRINTOUT
COMPARISON
OF MATRICES
OUTPUT OF
MEASURES
Figure 3-33 VOS Auxiliary Data Processing Computer Programs
I
VO
N3
I
(-*
ro
u>
oo
-------
3-82
SRL 2922-13-1271
It will be recalled that the raw data recording format is seven
track, with a recording density of 200 bpi using IBM-compatible code. The
record size for the input data is 27 characters. These characteristics
were chosen as most effective for data recording in the field. Highly
sophisticated computers are not utilized to maximum capability when operating
on data of this type, so a data format conversion program was written to
transform the raw data input to a more useful format.
The converted data were recorded on magnetic tape at an 800-bpi
density using a 9-track format. The record size was increased for greater
efficiency by reblocking of the data, thus requiring fewer tape accesses
during processing. The data conversion program also provided the first data
editing functions. Data recorded with the wrong parity could not be read
and so were eliminated. Illegal characters (a character other than a digit)
that may appear on the data tape are identified by logic in the data processing
program. These types of errors were rare; fewer than one recording error was
observed per 1200-foot raw-data tape.
The major portion of the data reduction task is accomplished with the
data processing computer program. The data for each route replicate are
analyzed separately. The input data consist of those on the intermediate
800-bpi data tape obtained from the data conversion program and the SDC
traffic density weighting factors for the city in which the data were obtained.
As the speed data are accumulated in the master mode matrix,
another data editing is performed to eliminate non-digital characters and
SCOTT RESEARCH LABORATORIES. INC
-------
3-83
SRL 2922-13-1271
to smooth inconsistencies in the speed data. A speed inconsistency
is defined to be a speed change of more than 10 mph per second. Linear
interpolation is used to determine replacement speeds where these in-
consistencies occur. A clean output data tape, using 9-track format at
1600 bpi, is prepared as the data are being analyzed. The auxiliary
matrices are also accumulated during this part of the processing.
The punched-card output of typical acceleration and deceleration
data is also prepared during data reduction. The mode matrices are then
calculated from the master mode matrix upon completion of the processing of
all data for a route replicate. All matrices, including the master mode
matrix, are output on the route replicate matrix magnetic tape for storage
and future data analysis. As each route replicate is processed, the
clean data and the output matrices are accumulated on the respective output
tapes. All output matrices are uniquely labeled for easy identification
during analysis.
Following reduction of the data for all the route replicates
for a route or a city, the appropriate matrices are merged and compacted
to form the desired composites. Initially, the input data consist of the
route replicate matrix tape; however, as higher-level composites are
developed, any matrix tape may be used as input to this program. The
output matrix tape consists of properly labeled route or city composites.
A different approach to merging was required to obtain the
five-city composite matrices that represent vehicle operation in the urban
areas where data were collected. The normal matrix merging approach was not
SCOTT RESEARCH LABORATORIES, INC
-------
3-84
SRL 2922-13-1271
appropriate because this would reflect a composite weighted by the data
sample size for each city. The five-^city composite was obtained from the
city mode matrices by weighting each according to the number of vehicles
registered in that urban area.
Hardcopy printer output may be obtained for any of the data
matrices from the matrix printing program. Options are included to output
selected sets of matrices in either 70 x 70 or compacted form. There-
fore, any data matrix tape may be used as input to this program. This
completes discussion of the package of computer programs used in the
reduction and presentation of VOS data. The program for comparing and
evaluating the output from two sets of input data will be discussed in
Section 3.5.2.
Computer data processing and analysis were performed at the
IBM Data center in Los Angeles. The computer was an IBM S/370 Model 155
with one megabyte of core storage. A schematic of the facilities is shown
in Figure 3-34.
3.5.2 Data Analysis Procedures
The analytic methods may be viewed as fitting into either of
two categories. The first includes what are called basic vehicle operation
descriptors. The other category encompasses various techniques for
comparing vehicle operation described by different sets of data. This section
describes how these measures and techniques were developed and how the re-
sults may be used.
SCOTT RESEARCH LABORATORIES, INC
-------
LINE
PRINTER
L ^ '
LINE
PRINTER
^
/CRT
CPU
CONSOLE
V DISPLAY
9 TRACK
MAGNETIC
TAPE
AUXILIARY
CARD READER/
PUNCH
A
1 TRACK
MAGNETIC
TAPE
DRIVES
ro
\£>
NJ
I
M
N>
IBM
S/370
MODEL 155
CENTRAL
PROCESSING
UNIT
(CPU)
\
DIRECT f\
ACCESS /
MAGNETIC I
DRUM STORAGE\ ,
/ CPU
/ HARDCOPY
\. CONSOLE
NJTPEWRITTEF
\
OJ
OO
A
Figure 3-34
IBM Los Angeles Data Center S/370 Model 155 Computer Facilities
MAGNETIC
DISK
DRIVES
MAGNETIC
DISK
DRIVES
-------
3-86
SRL 2922-13-1271
3.5.2.1 Basic Vehicle Operation Descriptors
One of the most useful descriptors of vehicle operation is
average speed. The versatile nature of the data processing program provides
a variety of average-speed statistics summarizing different aspects of
the data. The most extensive average-speed data may be obtained from the
matrix of average speed versus trip length and time of day (Figure 3-23).
Data of a more practical nature are summarized in the matrices of time,
distance, and average speed versus time of day and road type (Figure 3-31).
These matrices also provide a speed averaged over all times of day for each
road type for each route replicate. The average and standard deviation
of the route replicate average speeds for an entire city provide a measure
of the variation in vehicle operation due to daily traffic pattern fluc-
tuations.
The time and frequency percentage statistics for vehicle
operation during idles, cruises, accelerations, and decelerations provide
insight into the driving habits of the individuals observed along each
route. These percentage statistics are summarized for all route replicates
in a city in the same way as for average speed.
The average-time-in-mode matrix is an excellent summary of
how rapidly accelerations and decelerations are executed. The shapes of
accelerations and decelerations may be obtained by averaging the profiles
from the punch card output for each of them. Comparison of the average-
time-in-mode profile for each city indicates the variation in driving
habits between cities. The matrix of acceleration rate versus instantaneous
speed provides information on acceleration and braking stresses to which
typical vehicles are subjected under normal operation.
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922-13-1271 3~87
Average profiles of manifold vacuum versus cruise speed may be
obtained to compare the operation of the VOS vehicles in each city surveyed.
These data indicate vehicle operating fluctuations and the topographical
nature of each city. The data may not be used to infer anything about
typical vehicle operation in any of the cities, of course, because the
observed manifold vacuum data are peculiar to the particular VOS vehicle
from which they were obtained.
3.5.2.2 Comparisons of Vehicle Operation
The nine basic vehicle operation statistics, average speed and
the eight mode percentages defined in the previous section, provide a
useful basis for comparing driving habits from two geographical areas.
A method was sought, however, which quantified the driving patterns for
comparison. The search for general comparison techniques was centered
around procedures that may be applied to the compacted normalized total-time
and mode-frequency matrices. Priority was given to approaches that
provided relatively few output measures. Applicability of the analysis to
both the normalized total-time and normalized mode-frequency matrices was
a requirement.
A method of summarizing differences between corresponding
cells from two matrices being compared was the first technique developed.
Summing differences between two sets of numbers that have the same sum
will produce zero unless absolute values of the differences are used.
If the differences are squared, however, division of these squared differencs
by the corresponding cell entries from one of the matrices produces a
measure of the relative difference between the matrices. Reference matrix
cells having a value of zero can not be permitted, of course, so it was
found to be advantageous to combine cells that may be expected to be
SCOTT RESEARCH LABORATORIES, INC
-------
3-88
SRL 2922-13-1271
zero or close to zero. This produced an added feature in that a measure
was defined which was damped to small fluctuations in the two normalized
matrices being compared. The matrix partition shown in Figure 3-35 was
established so that computer processing could be used. The areas were
selected on the basis of two criteria. The value for each of the 13
areas was constrained to be at least 5 percent, so that division by small
numbers was eliminated. The areas also encompass modes of a similar
nature. This analysis procedure produces a statistic, analogous to chi-
square, that is called delta-square.
To illustrate the procedure, consider the calculation of delta-
square for two normalized matrices. Let X. and Y., i=l,...,13, be the
summed values for the thirteen areas shown in Figure 3-35 for each of two
normalized matrices. If the reference matrix contains the Y., then delta-
square is defined by:
(X± - Y±)
A. *
Y.
1
It should be noted that two different values of delta-square will be obtained,
depending on which matrix is selected as the reference. Two identical
matrices will produce a delta-square of zero, while increasing values of
delta-square indicate greater differences between the matrices. The delta-
square statistic provides an overall measure of matrix similarity and may
thus be viewed as a measure of the differences between two driving patterns.
SCOTT RESEARCH LABORATORIES. INC
-------
SRL 2922-13-1271
3-89
FINAL SPEED, mph
0 10 20 30 40 50 60
0 1
10
20
INITIAL
SPEED,
mph
30
40
50
60
10
11
13
T 1 1 T
I i I
L
Figure 3-35 Mode Matrix Partitioning For Delta-Square Analysis
SCOTT RESEARCH LABORATORIES, INC
-------
3-90
SRL 2922-13-1271
Since the delta~square approach, to evaluating matrix similarity
requires the paritioning of each matrix into areas made up of one or more
of the original driving modes, a comparison technique was sought
which would be sensitive to variations in each of the 196 matrix modes.
The procedure of linearly regressing the cells of one matrix on the cells
of another matrix was determined to be an efficient method of making matrix
comparisons. Correlation coefficients were computed for the cruise modes,
acceleration modes, and deceleration modes (Figure 3-36), for two matrices
and for the total matrices (mode-by-mode correlation of all 196 modes).
Each similarity comparison for 2 normalized matrices thus results
in 8 linear regressions: four for the normalized time-in-mode matrix and
four for the normalized mode-frequency-of-occurrence matrix. The regression
analysis provides a correlation coefficient, the slope and intercept of
the regression line, and the standard error of the estimate.
As will be seen in the results section, this linear regression
technique yields great economy in the presentation of the data by re-
presenting the differences between two matrices by a single straight line.
The correlation coefficients, of course, deviate from unity by an amount
proportional to the differences between corresponding matrix cells. A
computer program was written to perform the delta-square and linear
regression analyses.
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
3-91
INITIAL
SPEED,
mph
10
20
30
40
50
60
1 = CRUISE MODES (INCLUDING IDLE): 14 CELLS
2 = ACCELERATION MODES: 91 CELLS
3 = DECELERATION MODES: 91 CELLS
FINAL SPEED, mph
10 20 30 40 50 60
Figure 3-36 Matrix Mode Categories For Linear Regression Analysis
SCOTT RESEARCH LABORATORIE& INC
-------
4-1
SRL 2922-13-1271
4.0 RESULTS
This section discusses the processed data collected during the
course of the program. The complete characterization of vehicle operating
patterns is accomplished by reviewing 17 factors which contribute to a
total assessment of the driving patterns for a route or group of routes.
Basic descriptors of vehicle operation are average speed and mode composi-
tion; i.e., percent of total operating time and percent of mode frequency
of occurrence at idle, cruise, acceleration, and deceleration.
Detailed information on vehicle modal operation is contained in
the compacted speed-mode matrices. These matrices define the following
operating characteristics as a function of initial speed versus final
speed:
o Total time in mode
o Normalized time in mode
o Average time in mode
o Total mode frequency of occurrence
o Normalized mode frequency of occurrence
o Transition probability
4.1 DATA COMPARISONS
Two matrices, normalized time in mode and normalized mode fre-
quency, are considered the primary descriptors of vehicle mode operating
patterns. Comparisons between cities, or between a city and the five-city
composite, are performed by using statistical techniques for determining
the degree of similarity between the respective normalized time and fre-
quency matrices. Appendix E contains detailed matrix information describing
the vehicle operating patterns for each area surveyed during the program.
SCOTT RESEARCH LABORATORIES, INC
-------
4-2
SRL 2922-13-1271
In addition to the aforementioned descriptors, information
on acceleration/deceleration characteristics and intake manifold vacuum
are presented to supplement the characterization of vehicle operating
patterns.
4.1.1 Comparison of Basic Descriptors
Basic descriptors of operating patterns are compared and discussed
in this section. These basic descriptors represent useful summaries of
the detailed data over the total vehicle operating regime. The values for
average speed and the mode composition statistics for the time and frequency
matrices are reflections of the patterns contained in the detailed
speed-mode matrices. The basic descriptors are essentially integrated forms
of particular matrix information and simplify further analysis.
Table 4-1 presents the overall average speed results, with the
freeway/non-freeway average speed values given for each survey. The grand
\
average speed for the VOS program is a weighted average (weighted by vehicle
registration) of the average speeds for the five cities surveyed and is
shown in the second column of Table 4-1. This composite average speed
was 26.0 mph. All further references to composite results will denote data
weighted by vehicle registration.
The overall average speed from the individual urban area
surveys ranged from 21.6 mph in New York City to 29.3 mph in Los Angeles,
representing a variation of -16.9% to +12.7% from the composite value.
The relatively high overall average speed in Los Angeles is believed to
be a function of two primary factors. First, forty percent of the total
miles surveyed in Los Angeles were on freeways. Additionally, non-freeway
average speeds were higher because vehicles are permitted to make right
SCOTT RESEARCH LABORATORIES, INC
-------
X
Table 4-1
Comparison of Basic Descriptors of Vehicle Operation
CO
5
N>
to
^
NJ
o
§ Total Miles
O
Total Freeway Miles
% Freeway Miles
Average Speed:
Freeway , mph
5-Cit
Comp.
29.9
48.1
5-Citj
Comp.^
(3) (4)
N.Y.C. Chicago Cinci. Houston L.A. L.A.-4^ L.A.-4
32.7
46.5
3763.8 4381.0 2008.5 4466.3 7845.5 644.0
1183.8 1046.6 460.1 1390.4 3141.6 157.7
31.4 23.9
40.5 46.5
22.0 32.2 40.0 24.5
54.6 50.3 48.7 45.2
454.3
111.9
24.6
28.3
f
u>
Average Speed:
Non-Freeway , mph
21.6
21.3
17.8 21.3
22.4 23.0 23.2 17.8
15.5
Average Speed:
Overall, mph
25.8
26.0
21.6 24.5
25.9 27.7 29.3 21.0
17.4
(1) Cities weighted equally
(2) Cities weighted by vehicle registration
(3) Results are for defined hours 9:00 a.m. - 11:00 a.m. and 1:00 p.m. - 3:00 p.m.
(4) Results are for off hours 7:00 a.m. - 9:00 a.m. and 3:00 p.m. - 5:00 p.m.
-------
4-4
SRL 2922-13-1271
hand turns "on the red" at most traffic light intersections in Los Angeles,
thereby decreasing idle time. The slow overall average speeds witnessed
in the New York City area are believed to be caused by extreme traffic
density, and to a lesser degree, street maintenance deficiencies.
Average speeds on freeway segments ranged from 40.5 mph in
New York City to 54.6 mph in Cincinnati, with a composite value of 46.5
mph. The freeway system in urban Cincinnati is limited, in extent, and the
high average speed obtained indicates minimal traffic congestion. Non-
freeway average speeds ranged from 17.8 mph in New York City to 23.2
mph in Los Angeles. The composite non-freeway average speed was
21.3 mph.
Vehicle operating data showing the average speed values obtained
over the LA-4 road route are also included in Table 4-1. All three average
speed measurements for the LA-4 defined hours were lower than the corres-
ponding urban Los Angeles values. The location of the LA-4 route (in
central L.A. where traffic density is relatively high), the numerous traffic
lights, and the lower percentage of freeway miles over the LA-4 road route
all contributed to the lower average speed values. Off-hours LA-4
speed values, in comparison to the LA-4 defined-hours data, were slower
yet. A marked reduction in average speed on freeways segments and
minor decreases on non-freeway segments were observed to result from
increased traffic density during the busier hours (off hours).
4.1.2 Mode Composition
4.1.2.1 Distribution of Total Time by Mode
The distribution of total vehicle operating time in the
various modes is reflected in the speed-mode matrices. By adding the
SCOTT RESEARCH LABORATORIES, INC
-------
4-5
SRL 2922-13-1271
times accumulated in the 13 cruise cells of a 14 x 14 normalized time
matrix, a value for the percent of total time spent at cruise conditions
is obtained. Similarly, data contained in the 91 cells representing
various acceleration modes and in the 91 cells for deceleration modes
were summed to yield the percent of time spent in each of those two
general mode cateogries. The remaining cell of the matrix represents the
percent of time spent at idle.
The above summation procedure was performed on the normalized
time-in-mode matrices for each city, for the five-city composite, and
for the LA-4 route. Table 4-2 presents the results of this summary evalu-
ation. The five-city composite value for percent of total time at idle
conditions is 13.06%. Percent of total time at idle ranged from 10.13%
in Los Angeles to 17.45% in New York City. The relatively low value ob-
served in L. A. is probably related to the traffic regulations which
allow right hand turns "on the red" at most traffic lights. The value for
percent idle time witnessed in New York City was markedly greater than that
observed in the other four urban areas, and is attributed primarily to
greater traffic congestion.
Vehicle operation during the defined hours on the LA-4 route
resulted in a percent idle time nearly 3^ percent greater than the corres-
ponding value for urban Los Angeles. During the off hours, the percent
time at idle for the LA-4 off-hours compares closely with that for New
York City. In both cases the major influencing factors are considered to
be high traffic density and frequency of traffic lights.
In general, the values for percent of total time spent at cruise
conditions appears to vary inversely with percent idle time in the five
SCOTT RESEARCH LABORATORIES, INC
-------
Table 4-2
Distribution of Total Time in Mode Categories
60
O
S?
H
g
5-City
% Total Time, Idle 12.87
5-City
Comp.(2) N.Y.C. Chicago Cinci. Houston L_.A.
L.A.-4
. .
U;
13.06
17.45 14.11 11.34 11.30 10.13 13.56
L.A.-4
18.43
...
OJ
i
Total Time, Cruise 31.83
31.50
26.49 30.86 30.72 36.80 34.28 27.25
25.42
Total Time,
Acceleration
29.08
29.16
29.12 28.30 30.89 27.35 29.78 31.73
29.82
Total Time,
Deceleration
26.23
26.30
26.95 26.78 27.06 24.58 25.82 27.49
26.28
(1) Cities weighted equally
(2) Cities weighted by vehicle registration
(3) Results are for defined hours 9:00 a.m. - 11:00 a.m. and 1:00 p.m. - 3:00 p.m.
(4) Results are for off hours 7:00 a.m. - 9:00 a.m. and 3:00 p.m. - 5:00 p.m.
-------
4-7
SRL 2922-13-1271
cities. New York City, which, had the highest idle value, yielded the
lowest percentage of time at cruise, Houston, which had the second lowest per-
cent of time at idle, had the highest percent of time at cruise conditions.
The sum of percent time at idle plus percent time at cruise varied through-
out a range from 42.06% (Cincinnati) to 48.10% (Houston). A similar trend
was observed for LA-4 results during both time periods. However, the sums
for percent idle time plus percent cruise time were somewhat lower at
39.81% and 43.85% for the defined hours and the off hours, respectively.
The percent of total time spent accelerating and decelerating did
not vary greatly between the five cities. The composite value for percent
time accelerating was 29.16%, with a total range of only 3.54%. For
percent time decelerating, the composite value equalled 26.30%, with a
total range of only 2.48%. The percent of total time spent accelerating
and decelerating on the LA-4 route did not indicate any unusual trends.
4.1.2.2 Distribution of Mode Frequency by Mode Categories
Detailed mode frequency of occurrence data are contained in
the 14 x 14 speed-mode matrices. To derive basic summary information,
individual cell contents were combined to generate percent of mode fre-
quency in the four general categories of idle, cruise, acceleration,
and deceleration. The summations were performed on the data contained
in the normalized mode frequency of occurrence matrices, as described
above. The results of this evaluation are shown in Table 4-3.
The summary results from this evaluation are essentially controlled
by the mode-sequence logic. This sequence logic, as discussed in
Section 3.5, constrains the frequency data such that the sum of the
SCOTT RESEARCH LABORATORIES. INC
-------
PI
Ul
PI
60
O
50
s
3
S
O
% Frequency, Idle
Frequency, Cruise
Frequency,
Acceleration
Table 4-3
Distribution of Mode Frequency in Mode Categories
5-City 5-City .
Comp. UJ Comp.^' N.Y.C. Chicago Cinci. Houston L.A.
7.48
42.52
25.34
L.A.-4(3> L.A.-4
(4)
v>
P
VC
I
M
OJ
t-1
N3
7.59
42.41
25.26
7.60 7.81 7.42 6.97 7.60 9.50
42.40 42.19 42.58 43.03 42.40 40.50
25.43 25.38 25.44 25.43 25.04 25.26
10.70
39.30
25.41
oo
% Frequency,
Deceleration
24.66
24.74
24.57 24.62 24.57 24.57 24.96 24.74
24.59
(1) Cities weighted equally
(2) Cities weighted by vehicle registration
(3) Results are for defined hours 9:00 a.m. - 11:00 a.m. and 1:00 p.m. - 3:00 p.m.
(4) Results are for off hours 7:00 a.m. - 9:00 a.m. and 3:00 p.m. - 5:00 p.m.
-------
4-9
SRL 2922-13-1271
acceleration and deceleration mode frequencies equals the sum of the
idle and cruise mode frequencies.
The five-city composite values for mode frequency of occurrence
are:
o Idle = 7.59%
o Cruise = 42.41%
o Acceleration = 25.26%
o Deceleration = 24.74%
As evidenced by the individual values shown in Table 4-3 for each of the
five cities, there is very little between-city variation in any of the
four categories.
A greater frequency of idle modes was observed on the LA-4
route than for any of the five urban areas. During the off-hours on the
LA-4 route, the percent frequency of idle modes was over 3% greater than the
composite (due to the previously-mentioned influencing factors).
4.1.3 Correlation of Matrix Information
The discussions of the average-speed data and the mode-com-
position statistics provides basic insight into the nature of the results
obtained in each urban area. Another approach to the comparison and eval-
uation of vehicle operating patterns is provided by the correlation
technique discussed in Section 3-5.
4.1.3.1 Correlation of Normalized Time Matrices
Normalized time in mode matrices for each survey area and
for the five-city composite were cross-correlated to evaluate similarities
between survey areas. Values for the correlation coefficients, r, de-
rived from linear regression analysis, are presented in Table 4-4. Each
SCOTT RESEARCH LABORATORIES, INC
-------
30
m
so
n
I
o
90
B
n
LEGEND
CRUISE MODES
ACCEL MODES
DECEL MODES
OVERALL MATRIX
TABLE 4-4
CORRELATION COEFFICENTS
NORMALIZED TIME IN MODE COMPARISON
LA
HOUSTON
CINCI
LA-4
DEFINED
HOURS
.97
.97
,97
.97
.78
.80
.78
.84
LA-4
OFF
HOURS
LA
CHICAGO
NYC
ro
N>
I-1
u>
ro
i
.94
.96
.97
.96
.92
.98
.98
.96
.90
.96
.95
.95
.98
.98
.98
.98
.9/1
.96
.94
.96
.95
.98
.98
.97
.99
.98
.99
.99
.88
.92
.88
.92
.89
, .95
.95
.94
.96
.97
.95
.97
.96
.94
.96
.96
.83
.87
.80
.86
.82
.91
.90
.88
.90
.94
.92
.93
.97
.98
.97
.98
5-CITY
COMP
LA
I
M
O
HOUSTON
CINCI
CHICAGO
-------
4-11
SRL 2922-13-1271
of the five cities is compared with the composite and with each other city.
Normalized time values for the cruise modes (Including the idle modes),
acceleration modes, and deceleration modes were correlated as well as
the overall matrices.
Each of the five urban areas is similar to the five-city
composite, as evaluated with the overall matrix correlation coefficients which
range from 0.959 to 0.989. In the comparison of matrix data between in-
dividual cities, values of r between the overall matrices ranged from
0.860, for LA versus New York City, to 0.976, for Chicago versus New York
City.
Using this technique to compare the data contained in the
normalized time matrices for LA-4 defined hours versus urban LA, an r
value of 0.843 resulted. This indicates that these two sets of data
were the least similar of all the pairs evaluated. LA-4 off-hours data
had a 0.973 correlation coefficient with LA-4 defined-hours data for
normalized time.
4.1.3.2 Correlation of Normalized Mode Frequency Matrices
Correlations were also used to compare the normalized mode-
frequency-of-occurrence matrix data for each of the five cities with the
composite and with each other. Normalized mode-frequency data, being
a more stable measure of driving patterns because they are constrained
by the mode-sequence logic, yielded higher r values in all cases than
did the respective normalized time in mode data. The results are shown
in Table 4-5.
The correlation coefficients for each city versus the composite
are plotted in Figure 4-1. In this illustration, bracketed cities have
SCOTT RESEARCH LABORATORIES, INC
-------
30
PI
30
n
90
§
30
O
LEGEND
CRUISE MODES
ACCEL MODES
DECEL MODES
OVERALL MATRIX
DEFINED
HOURS
TABLE 4-5
CORRELATION COEFFICIENTS
NORMALIZED MODE FREQUENCY OF OCCURRENCE COMPARISON
LA
.98
.96
.96
.99
,82
.81
.79
.90
OFF
HOURS
LA
HOUSTON
CINCI
CHICAGO
NYC
.97
.97
.97
.99
.97
.97
.96
.98
.97
.96
.95
.98
.97
.97
.97
.98
.98
.96
.95
.98
.99
.98
.98
.99
.98
.98
.98
.oq
.91
.91
.90
.96
.93
.92
.93
.97
.92
.93
.93
.96
.97
.96
.96
.98
.89
.89
.88
.95
.88
.88
.90
.95
.89
.89
.90
.95
.98
.97
.97
.99
5-C1TY
COMP
LA
HOUSTON
CINCI
to
\0
U)
I
CHICAGO
-------
SRL 2922-13-1271
4-13
NORMALIZED TIME
IN MODE
.1.000
0.995
Chicago -
. 0.990
Cinci. - 0.985
0.980
NORMALIZED MODE
FREQUENCY
1.000
0.995
0.990
Chicago"!
L.A. J
0.985 -§._New York Citv & Cinci
Houston
0.980
" L.A. _
N.Y.C.-
__ Ho us ton *
0.975
0.970
. 0.965
0.960
. 0.955
0.975 -
0.970 -
0.965 -
0.960 J
0.955 J
VALUES OF
THE CORRELATION
COEFFICIENT
VALUES OF
THE CORRELATION
COEFFICIENT
Figure 4-1 Relative Correlation of
Each City versus the Composite
SCOTT RESEARCH LABORATORIES. INC
-------
4-14
SRL 2922-13-1271
correlations with the composite which are not significantly different
at the 95% confidence level. For example, Cincinnati had a correlation
coefficient for the normalized time matrix equal to 0.985. An interval
about this value from 0.980 to 0.989 represents the range of r values
that would be obtained 95% of the time if Cincinnaiti were surveyed a large
number of times. Since the correlation coefficient for Chicago versus
the composite equals 0.989, and this value is contained within the 95%
interval about Cincinnati, the Chicago and Cincinnati correlations with the
composite are not significantly different.
4.1.4 Delta-Square Evaluation
In addition to the correlation studies, delta-square (modified
chi-square) evaluations were performed on the same matrix data, as dis-
cussed in Section 3.5. Delta-square values can range from 0.0, re-
presenting identical matrices, to very large numbers when the matrices
are very different.
The delta-square evaluation was performed on the normalized
time-in-mode matrices and the normalized mode-frequency matrices for each
city versus the five-city composite and for each city versus each other
city. The results are given in Tables 4-6 and 4-7.
4.1.5 Supplemental Information
Up to this point in the discussion of the results, vehicle
operating patterns have been described using the information contained
in the speed-mode matrices. Additional information was generated to aid
in the understanding of acceleration/deceleration characteristics and
intake manifold vacuum curves.
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
4-15
Table 4-6
Delta-Square Values For Each
City versus the 5-City Composite
City
L.A.
Delta-Square Value
Normalized Time
in Mode
5.8
Normalized Mode
Frequency
2.5
Houston
4.8
1.9
Cinci.
1.6
1.8
Chicago
1.9
1.5
N.Y.C.
8.6
3.6
SCOTT RESEARCH LABORATORIES, INC
-------
so
m
VI
tn
so
n
8
X
o
so
LEGEND
NORMALIZED
TIME COMPARISON
NORMALIZED
MODE FREQUENCY
COMPARISON
LA-4
DEFINED
HOURS
6.2
3.5
31.5
20.0
TABLE 4-7
OVERALL MATRIX SIMILARITY
A- SQUARE VALUES
LA-4
OFF
HOURS
LA
LA
HOUSTON
CINCI
CHICAGO
NYC
5.8
2.5
4.8
1.9
7.6
1.1
1.6
1.8
6.3
2.2
3.8
0.4
1.9
1.5
18.2
9.8
7.9
6.3
5.3
6.1
8.6
3.6
55.9
14.2
23.9
10.6
17.4
9.3
5.5
1.4
5-CITY
COMP
LA
HOUSTON
CINCI
CHICAGO
VD
OJ
I
-------
4-17
SBL 2922-13-1271
4.1.5.1 Acceleration/Deceleration Characteristics
Matrices were generated for each city and the five-city composite
to characterize acceleration and deceleration rates at instantaneous
speeds. Figure 4-2 presents a graph of acceleration and deceleration rates
versus instantaneous speed for the five-city composite data. The average
acceleration rate curve reaches a maximum value of approximately 2.45 mph/
second and the average deceleration curve peaks at approximately 2.25 mph/
second. However, the curves defining the average-plus-two-standard-devia-
tions values of acceleration and deceleration indicate that the deceler-
ations are distributed over a greater range.
The frequencies of accelerations and decelerations at the
higher rates is shown more precisely by the distribution functions in
Figure 4-3 which give the percent of accelerations whose rates exceed
any given rate. Figure 4-4 presents the analogous distribution functions
for decelerations.
Acceleration and deceleration characteristics varied very little
between the five cities. Data were processed to allow drawing profiles
of typical acceleration and deceleration curves (speed versus time).
Figure 4-5 presents typical 0-30 mph acceleration curves for the individual
cities. Typical 30-0 mph deceleration profiles for the five cities are
shown in Figure 4-6.
Of particular interest is the fact that Los Angeles, which had
the highest average speed, indicates the slowest typical acceleration
profile for the 0-30 mph case. Conversely, New York City had the lowest
overall average speed and the highest rate of acceleration profile for
typical 0-30 mph accelerations.
SCOTT RESEARCH LABORATORIES, INC
-------
n
a:
t
BD
o
o
50
-d
c
o
(fl
A
P,
UJ
H
53
O
I
\
\
AVERAGE
ACCELERATION
RATE
\
AVERAGE ACCELERATION RATE
+2 STANDARD DEVIATIONS
INSTANTANEOUS SPEED, mph
10 20 30 40 50 60
70
C'
i1:
[ J
to
AVERAGE
DECELERATION RATE
-4
-6
AVERAGE DECELERATION RATE
+2 STANDARD DEVIATIONS
\
\
Figure 4-2 Acceleration Rate versus Instantaneous
Speed: 5-City Composite
\
-------
4-19
SRL 29:2-13-1271
0,40
COND
PH
LU
0,01
PFRCENT OF ACCELERATIONS AT
A FASTER RATE THAN SHOWN
FIGURE 4-3 DISTRIBUTION FUNCTIONS OF 5-CiTY-
COMPOSITE ACCELERATIONS AT VARIOUS
INSTANTANEOUS SPEEDS
SCOTT RESEARCH LABORATORIES, INC
-------
SRL 2922-13-1271
4-20
10
Q
Z
O
LU
to
X
Q-
Qi
t_J)
LU
(=3
1
i
/
/
IT
If
/
\VJ
wk
1
/
"
1
1
^
4
2
i
^
)
j
i
f
t
s
^
1
f
/
r
/
/
/
A
/
s
^***
'
s
/
f
'
t
/
f
f
/
/
/
'7
r
Jx '
?*^
n
j
7
^
j
f
f*
IflP
>
jr
^
^
j
y
HS:
M'
^
^
,/*
>
v'"
_y
^s
J
/
»-^^
>
/
X"
|
_/
21*
*s
>
jf
_s
*
^ *
^^
,
>
^
/
>
*
^
(
j
'
^
>
V
y
^
-j^
^1
^/
i
V!
.^
f
f
\>
j
x'
?2
*2
f
3H
^
^/"
MpH
_x
^f~r\
n( )
xj*
22-
f
1
.
2.
2
^
^^
MPH
11
"
_
«
-
0,35
0,30
0,25
0,20
0,15
0,10
0,05
100,
10,
1,
0,1
0,01
PERCENT OF DECELERATIONS
AT A FASTER RATE THAN SHOWN
FIGURE
DISTRIBUTION FUNCTIONS OF 5-CiTY
COMPOSITE DECELERATIONS AT VARIOUS
INSTANTANEOUS SPEEDS
SCOTT RESEARCH LABORATORIES, INC
-------
pn
30
n
2
pn
I I I ! < I I ' I I I
l±M- Hi
-i-ri-H-;
Figure 4-5 Average 0-30 mp!
- Acceleration Profile For Each City
: :
ihrirtrH-
'S.
ta
r1
K)
OJ
I
-P-
I
NO
-------
30
m
n
03
o
3D
>
O
3D
O
.±t±H±1 lllhlH t
Figure 4-6 Average 30-0
Deceleration Profile for
Each City
r-o
N)
i
N3
KJ
-------
4-23
SRL 2922-13-1271
4.1.5.2 Intake Manifold Vacuum
The chase vehicles were equipped with pressure transducers for
determining the intake manifold vacuum during survey operations. Although
it is recognized that intake manifold vacuum conditions for the chase ve-
hicle do not properly reflect the manifold vacuum characteristics of the
vehicles being chased, due to major differences in engine size, vehicle
weight, etc., the collected data do provide some useful information on
vehicle operation. By determining the intake manifold vacuum levels at
cruise conditions and comparing these data to road load manifold vacuum
curves, it is possible to derive the approximate amount of time that the sur-
vey vehicle was operated at, or near, normal road load conditions.
Figure 4-7 presents the average of the manifold vacuum levels
recorded at each cruise speed during the entire survey. Plots of
manifold vacuum readings for several cruise speeds indicated the manifold
vacuum data to be distributed approximately normally about the mean
values. The two curves labeled A-2 and A-3 in Figure 4-7 represent ±2
standard deviations about the mean; it can therefore be expected that
approximately 95% of the manifold vacuums at cruise fall between these
upper and lower curves.
One of the chase vehicles was operated at various cruise speeds
on a level road and at three different road grades to establish plots
of cruise manifold vacuum versus percent road grade. The road load manifold
vacuum curves for the vehicle operation on a level grade and on a 4.00%
uphill and downhill grade are shown by the dashed lines in Figure 4-7. The
proximity of the curves in this figure indicate that it would be reasonable
SCOTT RESEARCH LABORATORIES. INC
-------
O
s
M
a
mi.
VOS Survey Data:
A-l = Average Manifold Vacuum
A-2 = Average Vacuum + 20
A-3 = Average Vacuum - 2o
Road Load Test Data:
B-l = Manifold Vacuum on level road
B-2 = Vacuum on 4.00% uphill grade
B-3 = Vacuum on 4.00% downhill grade
I I I I I
III
t/5
P
NJ
NJ
CO
I
I
N>
Figure 4-7
CRUISE SPEED, MPH
Intake Manifold Vacuum Versus Cruise Speed
-------
4-25
SRL 2922-13-1271
to estimate that approximately 85-90% of the time the vehicles cruised
on road surfaces which are within ±4.00% of level.
The data describing the manifold vacuum versus road grade data
were collected in the San Bernardino, California area. All measurements
were made with the vehicle air conditioning system operating. Field survey
data were collected in the five cities with the air conditioning both
operative and inoperative. This factor probably accounts for the small
vertical displacement of the field survey data above the road test readings.
4.2 LOS ANGELES ROUTE REPLICATE EVALUATION
As discussed above, remarkably high matrix correlations between
cities and for each city versus the composite were obtained. It was
decided to investigate further the driving pattern variability within a
city to permit further assessment of the delta-square and correlation values.
The first city visited, Los Angeles, was surveyed for a longer time period
than any of the other four cities. During a two-week time period, four
separate weekday routes and four corresponding weekend routes were
driven. Since each of four routes was driven twice daily for seven full days,
a total of 56 route replicates was obtained. Each route replicate was
compared with the Los Angeles composite to determine the variability of
the data within Los Angeles.
4.2.1 Normalized Time in Mode Variability
Mode composition statistics for the route replicate time-in-mode
matrices are given in Table 4-8. Each weekday route (Routes 1-4) was
replicated 10 times and each weekend route (Routes 5-8) was replicated
four times.
SCOTT RESEARCH LABORATORIES. INC
-------
©
so Mode
M
m
30
£
CD
8
i
s
Route No
1
Route No
Route No
Route No
Weekday
Route No
Route No
Route No
Route No
Table 4-8
Composition For Route Replicates In Los Angeles - Percent Of
Percent Time Percent Time Percent Time
@ Idle @ Cruise @ Acceleration
. 1
. 2
. 3
. 4
Average
. 5
. 6
. 7
. 8
Weekend Average
Grand Average
Mean
8.83
9.41
11.58
10.51
10.08
8.61
9.32
11.12
7.88
9.23
9.84
Std.
Dev.
1.20
1.55
2.60
2.42
1.94
1.67
1.43
1.82
1.24
1.54
1.83
Mean
35
33
33
34
34
35
33
34
37
35
34
.82
.46
.55
.22
.26
.06
.76
.69
.49
.25
.54
Std.
Dev.
1.82
1.95
1.49
1.79
1.76
3.19
1.86
0.98
4.10
2.53
1.98
Mean
29.49
30.92
29.84
28.84
29.77
30.19
31.02
29.42
28.74
29.84
29.79
Std.
Dev.
1.66
1.09
1.81
0.48
1.26
1.96
0.32
1.89
1.94
1.28
1.26
C/3
P
t-O
Time In Mode to
i
i*
Percent Time to
vj
@ Deceleration M
Mean
25.86
26.20
25.03
26.44
25.88
26.14
25.90
24.78
25.90
25.68
25.82
Std.
Dev.
0.
0.
0.
1.
0.
0.
0.
0.
1.
0.
0.
97
56
84
i
14 £
88
79
77
79
03
84
87
-------
4-27
SRL 2922-13-1271
4.2.2 Normalized Mode Frequency Variability
Table 4-9 presents the mode-composition statistics for the
frequency matrices for the route replicates. The variability in the mode-
frequency statistics is consistently less than the variability in percent-
of-time statistics; i.e., mode frequencies are stable from one route to
another, as well as from one city to another, as previously indicated.
4.2.3 Delta-Square Comparison of Route Replicates
The delta-square technique described in Section 3.5 was used to
compare the matrix information generated for each route replicate with the
corresponding Los Angeles composite matrices. Normalized time-in-mode
matrix data and normalized mode-frequency matrix data were evaluated in
this manner. A statistical summary of the results is given in Table 4-10.
4.2.4 Average Speed Variability
Table 4-10 also contains a speed summary showing the variability
in average speeds for the route replicates. Average speeds for weekends
indicated approximately the same degree of variability as average speeds
for weekday route replicates.
4.2.5 Correlation Between Route Replicate Matrices and the Los Angeles
Composite For Normalized Time in Mode Data
Table 4-11 presents the results of correlating each normalized
time-in-mode matrix for a route replicate with the Los Angeles composite
normalized time-in-mode matrix. The correlations computed for all replicates
over the total matrix have an average value of 0.953.
4.2.6 Correlation Between Route Replicate Matrices and the Los Angeles
Composite Matrix for Normalized Mode Frequency
Table 4-12 presents the results of correlating each normalized
mode-frequency matrix for a route replicate with the corresponding Los Angeles
SCOTT RESEARCH LABORATORIES. INC
-------
so
n
1
PI
P
Table 4-9
Mode Composition For Route Replicates In Los Angeles - Percent Of Mode Frequency
Route No. 1
Route No. 2
Route No. 3
Route No. 4
Weekday Average
Route No. 5
Route No. 6
Route No. 7
Route No. 8
Weekend Average
Grand Average
Percent Mode
Frequency
@ Idle
Percent Mode
Frequency
@ Cruise
Percent Mode
Frequency
@ Acceleration
Percent Mode
Frequency
@ Deceleration
GO
?o
r1
U>
Mean
7.36
7.47
7.54
7.70
7.52
8.58
8.28
8.12
6.69
7.92
7.63
Std.
Dev.
0.73
0.56
0.71
1.09
0.77
2.01
0.30
0.65
1.32
1.07
0.86
Mean
42.64
42.53
42.46
42.30
42.48
41.42
41.72
41.88
43.31
42.08
42.37
Std.
Dev.
0.73
0.56
0.71
1.09
0.77
2.01
0.30
0.65
1.32
1.07
0.86
Mean
25.01
25.32
25.24
24.95
25.13
24.76
24.80
24.52
24.42
24.62
24.99
Std.
Dev.
0.40
0.54
0.40
0.46
0.45
0.23
0.17
0.69
0.28
0.34
0.42
Mean
24.99
24.68
24.76
25.05
24.87
25.24
25.20
25.48
25.58
25.38
25.01
Std.
Dev.
0.40
0.54 "T
NJ
OO
0.40
0.46
0.45
0.23
0.17
0.69
0.28
0.34
0.42
-------
9B
m
VI
so
§
30
1
m
O
Route No. 1
Route No. 2
Route No. 3
Route No. 4
Weekday Ave.
Route No. 5
Route No. 6
Route No. 7
Route No. 8
Weekend Ave.
Grand Average
iaoie 4-iu
.ues and Average Speeds for Route Replicates
Delta-Square
Normalized Time
in Mode
Mean
5.03
6.26
3.60
4.87
4.94
4.35
2.91
6.70
9.48
5.86
5.20
Range
1.66-8.76
1.15-18.07
1.05-12.37
1.28-14.17
1.05-18.07
2.37-7.85
2.03-3.76
2.28-11.62
5.01-18.64
2.03-18.64
1.05-18.64
Delta-Square
Normalized Mode
Frequency
Mean
2.89
3.19
2.23
3.44
2.94
2.56
1.82
3.59
9.22
4.30
3.32
Range
0.91-5.58
0.22-9.59
0.51-7.60
0.42-14.30
0.22-14.30
0.57-4.30
1.45-2.28
0.56-7.27
6.74-12.38
0.56-12.38
0.22-14.30
ro
NJ
1
in Los Angeles £o
i
i-"
Average
Speed , mph
Mean
29.50
27.88
28.36
29.46
28.80
31.08
30.42
30.33
32.61
31.11
29.46
Std. Dev.
1.63
1.63
1.22
.p-
1.20 S
1.42
0.53
1.00
2.15
1.86
1.38
1.41
-------
tn
I
i
s
p
Route No. 1
Route No. 2
Route No. 3
Route No. 4
Weekday Ave.
Route No. 5
Route No. 6
Route No. 7
Route No. 8
Weekend Ave.
Grand Average
m of Route Replicates
Cruise Modes
Mean
0.932
0.909
0.959
0.933
0.936
0.918
0.940
0.932
0.849
0.916
0.931
Range
0.889-0.
0.754-0.
0.922-0.
0.787-0.
0.754-0.
0.837-0.
0.921-0.
0.886-0.
0.629-0.
0.629-0.
0.629-0.
974
970
988
982
988
951
960
948
931
960
988
Table 4-11
with LA Composite Matrix; Normalized Time-in-Mode
Acceleration Modes Deceleration Modes Total Matrix
Mean
0.917
0.919
0.931
0.915
0.921
0.938
0.944
0.936
0.893
0.930
0.924
Range
0.864-0
0.876-0
0.872-0
0.814-0
0.814-0
0.916-0
0.905-0
0.881-0
0.835-0
0.835-0
0.814-0
.947
.955
.970
.942
.970
.949
.965
.959
.918
.970
.970
Mean
0.922
0.924
0.925
0.901
0.919
0.939
0.940
0.924
0.898
0.927
0.921
Range
0.882-0.
0.874-0.
0.831-0.
0.778-0.
0.778-0.
0.922-0.
0.904-0.
0.854-0.
0.866-0.
0.854-0.
0.778-0.
951
971
967
957
971
961
966
964
937
966
971
Mean
0.955
0.946
0.964
0.954
0.955
0.950
0.960
0.952
0.923
0.948
0.953
Range
0.935-0.
0.892-0.
0.925-0.
0.904-0.
0.892-0.
0.928-0.
0.949-0.
0.911-0.
0.816-0.
0.816-0.
0.816-0.
P
ro
N3
ro
i
NO
1 '
977
976
986 £
o
977
986
962
970
962
962
970
,986
-------
§ Table 4-12
3 Correlation of Route Replicates with LA Composite: Normalized Mode Frequency
30
w Cruise Modes Acceleration Modes Deceleration Modes Total Matrix
*° Mean Range Mean Range Mean Range Mean Range
| Route No. 1 0.948 0.850-0.990 0.898 0.856-0.939 0.900 0.834-0.935 0.972 0.943-0.989
O Route No. 2 0.956 0.872-0.990 0.921 0.859-0.961 0.926 0.864-0.964 0.976 0.947-0.993
Route No. 3 0.967 0.915-0.984 0.935 0.877-0.969 0.932 0.887-0.962 0.980 0.956-0.988
Route No. 4 0.943 0.764-0.985 0.903 0.728-0.938 0.902 0.804-0.950 0.970 0.907-0.987
Weekday Ave. 0.955 0.764-0.990 0.915 0.728-0.969 0.916 0.804-0.964 0.975 0.907-0.993
Route No. 5 0.963 0.941-0.977 0.910 0.868-0.945 0.904 0.831-0.941 0.975 0.962-0.987
Route No. 6 0.976 0.947-0.991 0.933 0.908-0.947 0.927 0.909-0.942 0.982 0.973-0.988
Route No. 7 0.948 0.930-0.963 0.918 0.871-0.942 0.915 0.827-0.940 0.969 0.953-0.978
Route No. 8 0.868 0.748-0.904 0.854 0.768-0.890 0.868 0.804-0.908 0.945 0.895-0.957
Weekend Ave. 0.950 0.748-0.991 0.907 0.768-0.947 0.906 0.804-0.942 0.970 0.895-0.988
Grand Average 0.953 0.748-0.991 0.913 0.728-0.969 0.913 0.804-0.964 0.974 0.895-0.993
U)
I
-------
4-32
SRL 2922-13-1271
composite matrix. The average value of r resulting from this comparison
was 0.974 (total matrix correlation).
4.2.7 Summary of Route Replicate Evaluation
The evaluation of driving patterns for each replicate was
performed on the Los Angeles survey data; however, a preliminary look at
replicate data for the other cities makes reasonable the assumption that
variability within those other cities is of the same order as that for Los
Angeles. Table 4-13 presents a summary of data obtained from the distribution
functions of the evaluation measures for the route replicates.
4.3 DETROIT ROAD ROUTE
The primary objective of the CRC Project No. CAPE-10-68
was to collect and evaluate information on urban traffic patterns and
detailed vehicle modal operating characteristics. Another objective was to
design and evaluate a road route on the metropolitan Detroit area which would
produce vehicle operating data typical of the national average patterns.
Therefore, a basic road route was layed out in the greater Detroit met-
ropolitan area. The route was designed to yield vehicle operating patterns
representative of the five-city composite pattern.
4.3.1 Design of the Detroit Metropolitan Road Route
The road route was actually located in the city of Ann Arbor,
Michigan at a distance of approximately 35 miles from downtown Detroit.
This site was selected as being easily accessible to the EPA vehicle
emission test group which was recently relocated in Ann Arbor near the
University of Michigan Campus.
SCOTT RESEARCH LABORATORIES. INC
-------
4-33
SRL 2922-13-1271
Table 4-13
Summary of Route Replicate Evaluation
Comparison With L.A. Composite
1. Correlation of Normalized
Time in Mode Data
2. Correlation of Normalized
Mode Frequency Data
3. Delta-Square for Normalized
Time-in-Mode Data
4. Delta-Square for Normalized
Mode Frequency Data
5. Average Speed - Weekdays
6. Average Speed - Weekends
Remarks
90% of all replicates resulted in r
values greater than 0.910
90% of all replicates resulted in r
values greater than 0.951
90% of all replicates resulted in
A-squares of less than 11.62
90% of all replicates resulted in
A-squares of less than 7.60
90% of all replicates were within
± 2.77 mph from the mean value
90% of all replicates were within
± 3.18 mph from the mean value
SCOTT RESEARCH LABORATORIES, INC
-------
4-34
SRL 2922-13-1271
4.3.1.1 Route Design Criteria
The construction of the Detroit road route was guided fay the
following basic design objectives:
1. Vehicle operating patterns on the route should be
representative of the five-city composite patterns
2. Repeated operation on the route should produce
minimum variability in vehicle operating patterns
3. The route should require approximately 20-30 minutes
for one lap.
To be representative of the five-city composite driving patterns,
operation on the Detroit road route should result in the following character-
istics :
5-City Composite Results
Average Speed, mph 26.0
Time (? Idle, % 13.06
Time @ Cruise, % 31.50
Time Accelerating,% 29.16
Time Deceleration,% 26.30
Idle Frequency, % 7.59
Cruise Frequency, % 42.41
Acceleration Frequency, % 25.26
Deceleration Frequency, % 24.74
In addition to satisfying the above criteria, it was necessary to specify
an appropriate variety of operating modes so that delta-square comparisons
would yield minimum values and correlation coefficients would be as high
as possible.
The route was designed for operation between the hours of
9:00 a.m. and 4:00 p.m. to minimize the variability associated with peak
morning and afternoon traffic congestion. Additionally, the route
was located so as to avoid the central city and the main campus areas,
SCOTT RESEARCH LABORATORIES, INC
-------
4-35
SRL 2922-13-1271
since they were judged to be sources of both day-to-day and seasonal
variability.
4.3.1.2 Route Construction
Three road routes, designated Route 30, Route 40,
and Route 50, were mapped out for collection of data. The three routes
varied only slightly, the freeway segments of each being identical.
Table 4-14 presents information on the construction of the three routes.
The routes were located on the eastern side of Ann Arbor. Route directories
and maps describing all three routes are included in Appendix B.
4.3.2 Operating procedures
Vehicle operating procedures specified for this task were
basically the same as those used for the collection of data in the 5 cities.
The procedures which were unique to this effort were:
o Collection of data was restricted to the
time period 9:00 a.m. to 4:00 p.m.
o Deletion of the chase car technique during
data collection.
The schedule for collecting data on the routes is shown in Figure 4-8.
As indicated, each route was driven three times daily over 6 days for a
total of 18 replicates per route.
The data were processed using the same computer programs used
previously to allow comparisons with the composite driving pattern
descriptors and also comparisons between routes.
4.3.3 Results
Analysis of the data collected on the three routes indicated
that this task was accomplished successfully, although the driving
patterns attributed to these routes differed from the five-city composite
SCOTT KESFARCH LABORATORIES, INC
-------
§ Table 4-14
m Construction Characteristics of Detroit (Ann Arbor) Road Routes
PI
>
30
n
* Route 30 Route 40 Route 50
g
in
V
VO
hO
Total Length, Miles 9.35 11.16 13.01
| Non-Freeway Miles 7.00 8.81 10.66
B
5 Freeway Miles 2.35 2.35 2.35
P
Freeway Mileage, % 25.2 21.0 18.1
Traffic Lights 44 7 T
UJ
Stop Signs 9 12 10
-------
3
30
n
July 13
Vehicle
Preparation
and
Practice
Driving
r\r\ All
Routes
July 14
Route 30*
Route 40*
Route 50*
July 15
Route 40
Route 50
Route 30
July 16
Route 50
Route 30
Route 40
July 17
Route 30
Route 40
Route 50
July 18
Route 40
Route 50
Route 30
July 19
Route 50
Route 30
Route 40
CO
£
to
vo
U>
* Each route driven three times per day
Figure 4-8 Data Collection Schedule For Detroit (Ann Arbor) Road Routes
-------
4-38
SRL 2922-13-1271
patterns in some respect. Operating characteristics for the three routes
are summarized in Table 4-15.
The average speed on each route compares favorably with the
five-city composite. Statistics depicting the percent of time spent in
the four general mode categories were much lower than desired for percent
idle time and higher than desired for percent cruise time. Percent of
time accelerating was only slightly higher than the composite value and
percent of time decelerating was nearly identical to the desired value.
For these reasons, matrix comparison by correlation techniques yielded
low values of r, and the delta-square values were inordinately
high.
The route statistics describing the percent mode frequency of
occurrence for the same four categories indicated the stability already
noted. Vehicle operation on all three routes yielded values of percent
mode frequency which were comparable to the five-city composite values for
all four mode categories. Route 40 had the least favorable statistics for
mode-frequency data.
In general, the descriptive statistics for the three routes in-
dicated less similarity to the five-city composite than did those for
any other comparison of a city versus the composite or those for compari-
sons between any two cities. This was to be expected because of the
short lengths of the Detroit (Ann Arbor) routes and the fact that the
area being surveyed presented a non-typical survey environment. The factors
believed to be responsible for these results are:
SCOTT RESEARCH LABORATORIES, INC
-------
4-39
SRL 2922-13-1271
Table 4-15
Comparison of Vehicle Operating
Patterns on Detroit Road Routes
with the Five-City Composite Results
Route Route Route 5-City
30 40 50 Comp.
Average Speed, mph 27.2 26.0 26.5 26.0
Time @ Idle, % 3.8 3.2 5.1 13.1
Time @ Cruise, % 36.6 39.0 39.6 31.5
Time Accelerating, % 32.8 32.0 29.4 29.2
Time Decelerating, % 26.8 25.8 25.9 26.3
Time: Correlation 0.61 0.47 0.61 1.00
Time: Delta-Square 24.2 60.7 29.3 0.0
Idle Frequency, % 7.9 5.2 5.7 7.6
Cruise Frequency, % 42.1 44.8 44.3 42.4
Acceleration Frequency, % 25.4 25.5 24.8 25.3
Deceleration Frequency, % 24.6 24.5 25.2 24.7
Frequency: Correlation 0.87 0.62 0.87 1.00
Frequency: Delta Square 8.8 24.3 7.3 0.0
Length, miles 9.35 11.16 13.01
Average time, hrs. 0.343 0.429 0.490
SCOTT RESEARCH LABORATORIES, INC
-------
4-40
SRL 2922-13-1271
o Short length
o Lack of traffic congestion
o Few traffic lights
o Lack of variety of available road types (Ann Arbor's
population is only 98,414 and the area involved is
only 15 square miles).
The first two items, although they contribute to the differences, are
necessary for the development of a repeatable road route.
Analysis of the data representing the three routes indicates
that minor changes would be adequate to improve the statistics for the
driving patterns. While actual changes in the physical structure or
location of certain portions of the routes would also improve the statistics,
additional data collection time would be required to verify this approach.
Instead of changing the composition of the routes, restrictions in driving
behavior can be applied.
To illustrate this approach, two examples are offered. Since the
descriptor of vehicle operating patterns which deviates the greatest
from the five-city composite data is the percent of time at idle, this value
can be altered to approximate the desired value. This can be achieved by
specifying a lengthy initial idle mode (typical of cold start operation)
and by increasing the idle time at other selected places throughout the
routes.
To illustrate how this approach affects the descriptive
statistics, Route 30 and Route 50 were artifically altered by adding 92
and 97 seconds of idle time, respectively. Table 4-16 presents the new
route statistics as determined by this one change. The quality of these
results indicates that both Route 30 and Route 50 now provide relatively
good representation of the average driving patterns. It appears highly
SCOTT RESEARCH LABORATORIES. INC
-------
Table 4-16
Improvement of Detroit Road Route
Characteristics by Adding Idle Time
90
m
v>
P)
§
3D
§
so
H
Average Speed, mph
Time @ Idle, %
Time @ Cruise, %
Time: Accelerating,
Time: Decelerating
Time: Correlation
Time: Delta Square
Route 30
27.2
3.8
36.6
32.8
26.8
0.61
24.2
Synthe-
sized
Route 35*
25.4
10.5
34.1
30.5
24.9
0.79
15.4
Route 50
26.5
5.1
39.6
29.4
25.9
0.61
29.3
Synthe-
sized
Route 55**
25.2
10.0
37.6
27.9
24.5
0.73
24.1
VO
NJ
NJ
I-1
5-City Composite M
26.0 M
13.1
31.5
29.2
26.3
1.0
0.0
.p-
I
* Created by adding 92 seconds of idle time to Route 30
** Created by adding 97 seconds of idle time to Route 55
-------
4-42
SRL 2922-13-1271
probable that additional restrictions on driver behaivor, combined with
minor changes in the physical design, would result in road routes which are
representative of average driving patterns.
4.4 SUMMARY OF RESULTS
The successful completion of the Vehicle Operation Survey has
provided a data bank on typical driving patterns encompassing five major
urban areas. The Digital Data Acquisition Systems, route design techniques,
data collection procedures, and data processing routines all proved to be
highly appropriate for this effort as well as adaptable to other studies of
a similar nature.
More specifically, VOS program experience permits the following
summary:
1. The Digital Data Acquisition Systems described herein
provide an adequate method of obtaining data on
vehicle operating characteristics.
2. The chase-car concept used in this program was
an accurate method for emulating driving habits
of the traffic population.
3. Data collection routes of approximately 150 miles
in length adequately represent vehicle usage
patterns.
4. Vehicle modal operating patterns are conveniently characterized
with the use of matrices of events based on initial speed
versus final speed.
The data collected in the 5 cities were combined (weighted) in
proportion to the number of vehicles registered in the respective areas.
Composite driving patterns obtained in this manner may be summarized as
follows:
SCOTT RESEARCH LABORATORIES. INC
-------
4-43
SRL 2922-13-1271
1. 13.06% of the time that vehicles are in operation
on the road is spent at idle.
2. The time spent at cruise conditions is 31.50%
of the total time. The most frequent cruise
speeds are between 27.5 mph and 37.5 mph. Vehicle
cruising above 45 mph is most likely to occur in the
range of 65-70 mph.
3. Total time spent accelerating is slightly higher
(29-16%) than time spent decelerating (26.30%).
In nearly every mode involving a speed change
the decelerations occur at slightly higher rates
than corresponding accelerations for the same speed
difference.
4. The overall average speed was 25.97 mph. Average
speed on freeway type roads (limited access) was
46.49 mph and vehicle speeds on non-freeway roads
equalled 21.33 mph.
Analysis of the data for the individual cities indicates the following
general trends:
1. Los Angeles exhibited the fastest overall average
speed of 29.34 mph and New York City had the slowest
average speed, 21.64 mph.
2. Los Angeles drivers spent the least amount of time at
idle, 10.13%; New York City had the highest value,
17.45%.
3. Based on the most frequent acceleration and deceleration
modes, 0-30 mph and 30-0 mph respectively, the typical
acceleration and deceleration speed-time profiles are
essentially the same for the five cities.
4. Analysis of the 56 route replicates in Los Angeles in-
dicate less variability than initially expected.
5. Vehicle operating patterns on the LA-4 Route, which
typifies driving characteristics during peak hours
in central Los Angeles, are notably different from the
patterns for the entire metropolitan area, as
expected.
Analysis of the data collected over the routes constructed in
Ann Arbor indicates the following:
SCOTT RESEARCH LABORATORIES. INC
-------
4-44
SRL 2922-13-1271
1. Preliminary efforts to construct a road route
which represents the average driving patterns were
partially successful.
2. Further modifications to the structure of the Detroit
(Ann Arbor) road routes will result in a repeatable
and representative road route.
SCOTT RESEARCH LABORATORIES, INC
-------
R-l
SRL 2922-13-1271
REFERENCES
1. "A Survey of Average Driving Patterns in the Los Angeles Urban
Area;" System Development Corporation, TM(L)4119/000/00, February
28, 1969.
2. "A Survey of Average Driving Patterns in the Houston Urban Area;"
System Development Corporation, TM(L)4119/001/00, August 28, 1969
3. "A Survey of Average Driving Patterns in the Cincinnati Urban
Area;" System Development Corporation, TM(L)4119/002/00, September
30, 1969.
4. "A Survey of Average Driving Patterns in the Chicago Urban Area;"
System Development Corporation, TM(L)4119/003/00, November 26, 1969-
5. "A Survey of Average Driving Patterns in the Minneapolis - St. Paul
Urban Area;" System Development Corporation, TM(L)4119/004/00,
February 13, 1970.
6. "A Survey of Average Driving Patterns in the New York Urban Area;"
System Development Corporation, TM(L)4119/006/00, August 15, 1970.
7. "A Survey of Average Driving Patterns in Six Urban Areas of the
United States: Summary Report'" System Development Corporation,
TM(L)4119/007/00, January 29, 1971.
8. "A Study of Los Angeles Driving As It Relates to Peak Photochemical
Smog Formation;" J. N. Pattison and M. P. Sweeney, Air Pollution
Control Association Presentation, June 1966 .^
9. "Laboratory Simulation of Driving Conditions in the Los Angeles Area;"
G. C. Hass and J. N. Pattison, SAE Paper 660546, August 1966.
10. "Vehicle Operations Survey: Interim Report, LA-4 Road Route Study;"
M. Smith and M. Manos, Scott Research Laboratories, Inc.; SRL 2922-
09-1071, October 29, 1971.
SCOTT RESEARCH LABORATORIES. INC
------- |