SHRP 2 Reliability Project L02
Guide to Establishing Monitoring
Programs for Travel Time
Reliability
PREPUBLICATION DRAFT • NOT EDITED
© 2013 National Academy of Sciences. All rights reserved.
ACKNOWLEDGMENT
This work was sponsored by the Federal Highway Administration in cooperation with the
American Association of State Highway and Transportation Officials. It was conducted in the
second Strategic Highway Research Program, which is administered by the Transportation
Research Board of the National Academies.
NOTICE
The project that is the subject of this document was a part of the second Strategic Highway
Research Program, conducted by the Transportation Research Board with the approval of the
Governing Board of the National Research Council.
The members of the technical committee selected to monitor this project and to review this
document were chosen for their special competencies and with regard for appropriate balance.
The document was reviewed by the technical committee and accepted for publication according
to procedures established and overseen by the Transportation Research Board and approved by
the Governing Board of the National Research Council.
The opinions and conclusions expressed or implied in this document are those of the researchers
who performed the research. They are not necessarily those of the second Strategic Highway
Research Program, the Transportation Research Board, the National Research Council, or the
program sponsors.
The information contained in this document was taken directly from the submission of the
authors. This document has not been edited by the Transportation Research Board.
Authors herein are responsible for the authenticity of their materials and for obtaining written
permissions from publishers or persons who own the copyright to any previously published or
copyrighted material used herein.
The Transportation Research Board of the National Academies, the National Research Council,
and the sponsors of the second Strategic Highway Research Program do not endorse products or
manufacturers. Trade or manufacturers’ names appear herein solely because they are considered
essential to the object of the report.
The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished
scholars engaged in scientific and engineering research, dedicated to the furtherance of science and
technology and to their use for the general welfare. On the authority of the charter granted to it by
Congress in 1863, the Academy has a mandate that requires it to advise the federal government on
scientific and technical matters. Dr. Ralph J. Cicerone is president of the National Academy of Sciences.
The National Academy of Engineering was established in 1964, under the charter of the National
Academy of Sciences, as a parallel organization of outstanding engineers. It is autonomous in its
administration and in the selection of its members, sharing with the National Academy of Sciences the
responsibility for advising the federal government. The National Academy of Engineering also sponsors
engineering programs aimed at meeting national needs, encourages education and research, and
recognizes the superior achievements of engineers. Dr. Charles M. Vest is president of the National
Academy of Engineering.
The Institute of Medicine was established in 1970 by the National Academy of Sciences to secure the
services of eminent members of appropriate professions in the examination of policy matters pertaining to
the health of the public. The Institute acts under the responsibility given to the National Academy of
Sciences by its congressional charter to be an adviser to the federal government and, upon its own
initiative, to identify issues of medical care, research, and education. Dr. Harvey V. Fineberg is president
of the Institute of Medicine.
The National Research Council was organized by the National Academy of Sciences in 1916 to
associate the broad community of science and technology with the Academy’s purposes of furthering
knowledge and advising the federal government. Functioning in accordance with general policies
determined by the Academy, the Council has become the principal operating agency of both the National
Academy of Sciences and the National Academy of Engineering in providing services to the government,
the public, and the scientific and engineering communities. The Council is administered jointly by both
Academies and the Institute of Medicine. Dr. Ralph J. Cicerone and Dr. Charles M. Vest are chair and
vice chair, respectively, of the National Research Council.
The Transportation Research Board is one of six major divisions of the National Research Council.
The mission of the Transportation Research Board is to provide leadership in transportation innovation
and progress through research and information exchange, conducted within a setting that is objective,
interdisciplinary, and multimodal. The Board’s varied activities annually engage about 7,000 engineers,
scientists, and other transportation researchers and practitioners from the public and private sectors and
academia, all of whom contribute their expertise in the public interest. The program is supported by state
transportation departments, federal agencies including the component administrations of the U.S.
Department of Transportation, and other organizations and individuals interested in the development of
transportation. www.TRB.org
www.national-academies.org
Copy No. _____
SHRP 2 Project L02
Establishing Monitoring Programs for Travel Time Reliability
Guidebook – Final Validated
Prepared for:
Strategic Highway Research Program 2
(SHRP 2)
TRANSPORTATION RESEARCH BOARD
NAS-NRC
LIMITED USE DOCUMENT
This document is furnished only for review by members of the SHRP
2 Technical Coordinating Committee and is regarded as fully privileged.
Dissemination of information included herein must be approved by SHRP
2 Program officials.
September 10, 2012
Institute for Transportation Research and Education
In association with:
Iteris/Berkeley Transportation Systems, Inc.
Kittelson & Associates, Inc.
National Institute of Statistical Sciences
University of Utah
Rensselaer Polytechnic Institute
Joseph Schofer of Northwestern University
Asad Khattak of Planitek
Guidebook – Final Validated Draft
SHRP 2 L02
Establishing Monitoring Programs for Travel Time Reliability
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
GLOSSARY OF TERMS AND ACRONYMS
EXECUTIVE SUMMARY ....................................................................................................... ES-1
CHAPTER 1: INTRODUCTION ................................................................................................1-1
The Context within the Strategic Highway Research Program 2
Structure of the Guidebook
How To Use This Guidebook
What Is Travel Time Reliability?
Supply-Side and Demand-Side Perspectives
Trip-Making Concepts
Probability Density Functions
References
1-4
1-5
1-7
1-7
1-10
1-11
1-12
1-15
CHAPTER 2: DATA COLLECTION AND MANAGEMENT ...............................................2-1
Infrastructure-Based Sources
Automated Vehicle Identification Sources
Automated Vehicle Location Sources
Private Sector-Based Sources
Data Management for Recurring Conditions
Data Management for Non-Recurring Events
Summary
References
2-3
2-5
2-8
2-9
2-10
2-12
2-15
2-15
CHAPTER 3: COMPUTATIONAL METHODS ......................................................................3-1
Network Concepts
Trip-Making Concepts
Operating Conditions and Regimes
Imputation
Segment Travel Time Calculations
Route Travel Time Calculations
Influencing Factor Analysis
Summary
References
3-2
3-4
3-6
3-8
3-14
3-29
3-37
3-40
3-40
ii
Institute for Transportation Research and Education
SHRP 2 L02
Guidebook – Final Validated Draft
Establishing Monitoring Programs for Travel Time Reliability
CHAPTER 4: APPLICATIONS .................................................................................................4-1
Case Studies
San Diego, California
Northern Virginia
Sacramento–Lake Tahoe, California
Atlanta, Georgia
New York/New Jersey
Berkeley Highway Lab
Use Cases
See What Factors Affect Reliability (AE1)
Assess the Contributions of the Factors (AE2)
View the Travel Time Reliability Performance of a Subarea (AE3)
Assist Planning and Programming Decisions (AE4)
Determine When a Transit Route is Unreliable (AP5)
Be Alerted When the System has Become Unreliable (MM5)
Summary
4-1
4-3
4-6
4-9
4-13
4-16
4-18
4-20
4-23
4-30
4-35
4-37
4-41
4-43
4-45
CHAPTER 5: SUMMARY ..........................................................................................................5-1
Step 1: Collect and Manage Traffic Data
Step 2: Measure Travel Times
Step 3: Characterize
Step 4: Collect, Manage, and Impute Non-Recurring Event Data
Step 5: Identify the Sources of Congestion and Unreliability
Step 6: Understanding the Impact of the Sources of Unreliability
Step 7: Make Decisions
Conclusion
5-2
5-3
5-7
5-8
5-9
5-10
5-11
5-12
SUPPLEMENT A: MONITORING SYSTEM ARCHITECTURE ....................................... A-1
SUPPLEMENT B: METHODOLOGICAL DETAILS ........................................................... B-1
SUPPLEMENT C: CASE STUDIES
Chapter C1: Introduction
Chapter C2: San Diego, California
Chapter C3: Northern Virginia
Chapter C4: Sacramento–Lake Tahoe, California
Chapter C5: Atlanta, Georgia
Chapter C6: New York/New Jersey
C1-1
C2-1
C3-1
C4-1
C5-1
C6-1
SUPPLEMENT D: USE CASE ANALYSES ............................................................................ D-1
Institute for Transportation Research and Education
iii
Guidebook – Final Validated Draft
SHRP 2 L02
Establishing Monitoring Programs for Travel Time Reliability
LIST OF TABLES
Table 2-1: Example Database Table Relationship
2-11
Table 3-1: Illustration of Regimes
3-8
Table 3-2: Example of Measured Speeds by Lane, San Francisco Bay Area
3-16
Table 3-3: Example of an Incidence Matrix – percentage of vehicles having both a travel rate on
the upstream segment ab in a particular range and a travel rate on the downstream
segment bc in a particular range
3-32
Table 4-1: User Types and Their Classification
4-21
Table 4-2: Use Case Template
4-21
Table 4-3: Use Cases for the Travel Time Reliability Monitoring System
4-22
Table 4-4: See What Factors Affect Unreliability (AE1)
4-24
Table 4-5: Assess the Contributions of the Factors (AE2)
4-30
Table 4-6: Semi-Variances for Each Regime for Three Routes in San Diego
4-34
Table 4-7: Percentages for Semi-Variances for Each Regime for Three Routes in San Diego
4-34
Table 4-8: View the Travel Time Reliability Performance of a Subarea (AE3)
4-35
Table 4-9: Contribution by Regime to Total Semi-Variance for Three Routes in San Diego
4-36
Table 4-10: Assist Planning and Programming Decisions (AE4)
4-37
Table 4-11: Changes in Reliability over Time for a Hypothetical Route
4-38
Table 4-12: Determine When a Route is Unreliable (AP5)
4-41
Table 4-13: Hours of Unreliable Operation by Regime and Facility
4-42
Table 4-14: Percentage of Unreliable Operation by Regime and Facility
4-42
Table 4-15: Be Alerted When the System has Become Unreliable (MM5)
4-43
iv
Institute for Transportation Research and Education
SHRP 2 L02
Guidebook – Final Validated Draft
Establishing Monitoring Programs for Travel Time Reliability
LIST OF FIGURES
Figure ES-1: Seven Major Sources (Factors) of Nonrecurrent Congestion
ES-2
Figure ES-2: Information Flow in a TTRMS
ES-3
Figure ES-3: Block Diagram for a Travel Time Reliability Monitoring System (TTRMS) ES-5
Figure ES-4: Illustration of Variation in Travel Times by Time of Day Across a Year with NonRecurrent Events Excluded
ES-6
Figure ES-5: Illustration of Variation in Travel Times by Time of Day Across a Year with NonRecurrent Events Included
ES-7
Figure ES-6: How Travel Rates Are Affected by Congestion and Non-Recurring Incidents
ES-8
Figure 1-1: Five-Minute Average Travel Times on I-5 in San Diego
1-2
Figure 1-2: Information Flow in a TTRMS
1-2
Figure 1-3: Seven Major Sources (Factors) of Nonrecurrent Congestion
1-5
Figure 1-4: Reliability Monitoring System Overview
1-6
Figure 1-5: Concepts of Desired and Actual Times of Arrival
1-8
Figure 1-6: Disutility Function to Characterize Desired and Actual Times of Arrival
1-9
Figure 1-7: Example of Travel Time Histogram for Various Event Conditions
1-12
Figure 1-8: Example of Probability Density Functions for Various Event Conditions
1-13
Figure 1-9: Example of Cumulative Distribution Functions under Various Regimes (Operating
Conditions)
1-14
Figure 2-1: Types of Data Collection Sources
Figure 2-2: Types of Infrastructure-Based Data Collection Sources
Figure 2-3: Types of Vehicle-Based Data Collection Sources
2-1
2-2
2-3
Figure 3-1: Illustration of Monuments
3-3
Figure 3-2: Illustration of Segments
3-4
Figure 3-3: Illustration of Routes
3-5
Figure 3-4: Illustration of Route Bundles
3-6
Figure 3-5: Imputation of Traffic Data
3-9
Figure 3-6: Imputation Routine Example
3-11
Figure 3-7: Super Segment Examples
3-14
Figure 3-8: Example of Five-Minute g-Factor Estimates at a Detector
3-16
Figure 3-9: Example Estimates of g-Factors at a Detector (I-880 North, Oakland, CA)
3-17
Figure 3-10: Assignment of Segments to Point Detectors
3-18
Figure 3-11: Example Bluetooth Responses for a Single Vehicle
3-19
Figure 3-12: Example Application of AVI Sensor Placement
3-20
Figure 3-13: Raw Observations, US-50, South Lake Tahoe, CA to Placerville, CA
3-21
Figure 3-14: Filtered Observations of Trip Times, Placerville to South Lake Tahoe
3-22
Figure 3-15: Examples of Bluetooth-Based Vehicle Travel Times
3-23
Figure 3-16: Example of Cumulative Density Functions of Bluetooth-Based Vehicle Travel
Times
3-24
Institute for Transportation Research and Education
v
Guidebook – Final Validated Draft
SHRP 2 L02
Establishing Monitoring Programs for Travel Time Reliability
Figure 3-17: Daily Trends in Individual Vehicle Travel Times – I-5 Southbound
3-25
Figure 3-18: Daily Trends in Individual Vehicle Travel Times – I-5 Southbound (subset) 3-25
Figure 3-19: Example Illustrating Trends in the 5th and 75th Percentile Travel Times for
Individual Vehicles
3-26
Figure 3-20: Example of Locations and Headings Reported by AVL-Equipped Vehicles Trips
3-27
Figure 3-21: Example of Map Matching Challenges for AVL Data
3-28
Figure 3-22: Example of Route Section Travel Times
3-30
Figure 3-23: Example of Distribution of Simulated versus Actual Travel Rates for a Route
3-33
Figure 3-24: Example of a Network Simulated Using the Point-Queue-Based Model
3-33
Figure 3-25: Example of Actual versus Simulated Travel Times using the Point-Queue based
Model
3-34
Figure 3-26: Examples of Simple Median Method (FasTrak Data, San Francisco Bay Area,
California)
3-36
Figure 3-27: Example of Cumulative Distribution Functions by Regime and Non-Recurrent
Event Type
3-39
Figure 4-1: Case Study Locations
4-1
Figure 4-2: Study Area for San Diego Case Study
4-3
Figure 4-3: Study Area for Northern Virginia Case Study
4-7
Figure 4-4: Study Area for Sacramento–Lake Tahoe Case Study
4-10
Figure 4-5: Study Area for Atlanta Case Study
4-13
Figure 4-6: Study Area for New York/New Jersey Case Study
4-16
Figure 4-7: Study Area for Berkeley Highway Laboratory Data Investigations
4-19
Figure 4-8: Sub-area being examined for Use Case AE1
4-25
Figure 4-9: Five-Minute Average Weekday Travel Rates for Three Routes in San Diego 4-26
Figure 4-10: Five-Minute Average Weekday Travel Rates Plotted Against VMT/Hour for Three
Routes in San Diego
4-27
Figure 4-11: Semi-Variances by Five-Minute Time Period for the Normal Condition for Three
Routes in San Diego
4-29
Figure 4-12: CDFs by Regime for the Three Routes in San Diego
4-32
Figure 4-13: Subarea Aggregation Approaches
4-35
Figure 4-14: Using TR-CDFs to Analyze Performance Changes
4-39
Figure 4-15: Relative Importance of Different Conditions
4-40
Figure 4-16: Trends in Travel Time Percentiles and Their Standard Deviation
4-44
Figure 5-1: Generalized Process for Developing a Travel Time Reliability Monitoring System
5-1
Figure 5-2: The Role of Data
5-2
Figure 5-3: Measuring Travel Times
5-4
Figure 5-4: Characterizing the Observed Travel Times
5-7
Figure 5-5: The Role of Sources of Unreliability
5-8
Figure 5-6: Identifying Sources of Congestion
5-9
Figure 5-7: Understanding Recurrent and Nonrecurrent Causes of Congestion
5-11
vi
Institute for Transportation Research and Education
SHRP 2 L02
Guidebook – Final Validated Draft
Establishing Monitoring Programs for Travel Time Reliability
Figure 5-8: Making Decisions
Institute for Transportation Research and Education
5-12
vii
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
GLOSSARY
ACRONYMS
APC
ASOS
AVI
AVL
AWOS
CAD
CDF
CV
ESS
ETC
GPS
HAR
ITS
LCS
LPR
MAC
PDF
PeMS
RFID
SHRP2
TMC
TT-CDF
TT-PDF
TTRMS
VMT
Automated Passenger Count
Automated Surface Observing System
Automated Vehicle Identification
Automated Vehicle Location
Automated Weather Observing System
Computer Aided Dispatch
Cumulative Density Function
Connected Vehicle
Environmental Sensor Station
Electronic Toll Collection
Global Positioning Satellite
Highway Advisory Radio
Intelligent Transportation System
Lane Control Status
License Plate Reader
Media Access Control
Probability Density Function
Performance Measurement System
Radio-Frequency Identification
Strategic Highway Research Program 2
Transportation Management Center
Travel Time Cumulative Density Function
Travel Time Probability Density Function
Travel Time Reliability Monitoring System
Vehicle Miles Traveled
28
TERMS
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Market: A set of users in combination with a route bundle
Monument: A reference point used for travel time measurement
Non-Recurring Event: An event that does not occur regularly during a typical time of day,
including traffic incidents, work zones, weather, special events, traffic control devices, and
fluctuations in demand. The effect of non-recurring events can be magnified by inadequate base
capacity.
Passage Time: A timestamp assigned to a vehicle when it passes a given monument
Regime: The categories of conditions under which the segment, route, or network is operating at
a given point in time (or from one time to another). It is effectively the “loading condition” on
the system.
Route: A sequence of segments
GL-1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Route Bundle: A set of two or more routes
Sample Space: The set of raw data that pertain to each context for which a probability density
function is being developed, such as those that pertain to a regime (e.g., congested conditions) or
to another logical grouping (e.g., 7:00-9:00 a.m.) Also known as an observation set, observation
time frame, or sample frame.
Segment: A path between two monuments
Travel Rate: Travel time per unit distance
Travel Time: The amount of time spent traveling over a given segment or route
Trip Time: The door-to-door time for a trip
Use Case: Description of a system’s behavior based on user needs
User: People or package making a trip across the network
GL-2
1
EXECUTIVE SUMMARY
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
This guidebook describes how to develop and use a Travel Time Reliability Monitoring
System (TTRMS). It explains why such a system is useful, how it helps agencies do a better job
of managing network performance, and what a traffic management center (TMC) team needs to
do to put a TTRMS in place. The guidebook was prepared under Project L02 within the Strategic
Highway Research Program 2 (SHRP 2) under the management of the Transportation Research
Board.
Travel time reliability is the absence of variability in the travel times. If a system is
reliable, people can get to where they want to go, when they want to be there, all the time. If a
freeway is reliable, for example, then its travel times are the same for the same condition, all year
long. It is similar to a vehicle that always starts when the key is turned on. Of course, in reality
no system or roadway is perfectly reliable, and unexpected events (e.g., incidents) sometimes
significantly reduce the reliability of a system.
This executive summary is organized into the following sections:
Structure of the Guidebook
Why Do We Need a Travel Time Reliability Monitoring System?
What Should a Travel Time Reliability Monitoring System Do?
How Should a Travel Time Reliability Monitoring System be Structured?
An Illustration
Implementation
Conclusion
22
STRUCTURE OF THE GUIDEBOOK
23
24
25
26
27
28
29
30
31
32
33
34
35
The guidebook is organized into the following major parts:
An Executive Summary (this document) that gives agency managers a description of
what a TTRMS is, why it is valuable, and how it can be used. The document is
intended to prepare such individuals for the input they will receive from their staff
and help them make informed decisions about future actions.
A five-chapter guidebook that describes the process of measuring, characterizing,
identifying, and understanding the effects of recurrent congestion and non-recurrent
events that affect travel time reliability.
A series of four supplements that provide more detailed information on the functional
specification of a monitoring system, methodological details, a series of case studies,
and a series of use cases (applications) of the guidebook.
The project also has a separate final report that documents the research that led to the
guidebook.
36
WHY DO WE NEED A TRAVEL TIME RELIABILITY MONITORING SYSTEM?
37
38
39
40
Travel time reliability is effectively the absence of variation in travel times. The absence
of reliability can be largely attributed to seven influencing factors that cause congestion and
unreliable travel times (illustrated in Figure ES-1):
Traffic incidents,
ES-1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Work zones,
Weather,
Special events,
Traffic control devices,
Fluctuations in demand, and
Inadequate base capacity.
Source: Strategic Highway Research Program 2
Figure ES-1: Seven Major Sources (Factors) of Nonrecurrent Congestion
System operators also want to determine what actions they need to take to reduce
variability and enhance reliability. System users, on the other hand, want to avoid variability.
When making trips, they want to select routes that have reliable travel times, and they want to
leave at the right moment so they can arrive on time or minimize the probability of being late.
A TTRMS can help operating agencies, like those with Transportation Management
Centers (TMCs), monitor the performance of their system, understand the impacts of the various
influencing factors, provide credible information to the system users about what travel time
reliability to expect, and decide what actions to take to help improve reliability. This is the
reason why a TTRMS should be created. Most TMCs currently do not have these features or
capabilities.
Creating a TTRMS involves creating a new module that plugs into an existing TMC
platform. The TTRMS relies on the TMC to gather the sensor data, manage the data processing
and storage, and communicate the results of its findings to the outside world. The TTRMS
focuses on using the incoming sensor data, along with supplemental information about the
ES-2
1
2
3
4
5
6
7
influencing factors, and create a credible picture of how well the system is performing at the
present time and in the past.
The payoff is better reliability. The TTRMS will ensure that the TMC team knows what
how reliability will suffer if certain events take place. It will let them understand the impacts of
weather, incidents, special events, etc. It will also help the TMC team manage the network’s
reliability—in real time—by providing up-to-date information about how the variability in
segment and route travel times is either increasing or decreasing in response to actions taken.
8
WHAT SHOULD A TRAVEL TIME RELIABILITY MONITORING SYSTEM DO?
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
To fulfill its mission as a decision support tool, the TTRMS needs to execute four key
steps as illustrated in the conceptual diagram of information flow shown in Figure ES-2.
Figure ES-2: Information Flow in a TTRMS
Firstly, the TTRMS needs to effectively measure travel times. This is a complex
technical topic due to the variability of traveler behavior and the plethora of different
measurement sensors. Correctly measuring travel times along a given route requires a great deal
of systems development effort and statistical knowledge. This guidebook serves as a primer on
how to measure travel times using available technologies and statistical techniques. Measuring
an individual travel time is the foundational unit of analysis for reliability monitoring.
Secondly, the TTRMS needs to clearly characterize the reliability of a given system.
This is the process of taking a set of measured travel times and assembling them into a statistical
model of the behavior of a given segment or route. The statistical paradigm outlined in this
guidebook is that of using probability density functions (PDFs) and cumulative distribution
functions (CDFs) to characterize the performance of a given segment or route, usually specific to
a particular operating regime (a combination of congestion level and non-recurring event). This
ES-3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
guidebook gives specific advice on the statistical decisions required to effectively characterize
the travel times. Characterizing the reliability of a segment or route is fundamental to making
good decisions about what to do to improve the performance of that segment or route.
Thirdly, the TTRMS needs to identify the sources of unreliability. Once the reliability of
a segment or route has been characterized, transportation managers need to understand what
caused the unreliability (and how to “fix” it). This guidebook follows the causal list that FHWA
uses to describe why congestion arises, breaking these sources into the seven major influencing
factors described previously. It discusses how to pull in data for these influencing factors and
effectively fuse them with the travel time data generated in previous steps. Identifying the travel
times impacted by these sources of congestion is required preparation for understanding system
reliability.
Finally, the TTRMS needs to help operators understand the impact of these sources of
unreliability on the system. This final step in turning raw data into actionable decisions requires
both quantitative and qualitative methodologies: operators need clear visualizations of data, as
well as quantifications. This dual approach supports both data discovery and final decisionmaking about a given route. Understanding reliability is the key to good decision-making about
improving system reliability.
Once in place, a TTRMS that accurately and consistently executes these four steps
becomes a powerful tool for traffic management. It is a tool that decision makers can use to
understand how much of their delay is due to unreliability, and how to mitigate that delay. For
example, should a freeway operator deploy more service patrol vehicles (to clear incidents more
quickly) or focus their efforts on coordinating special event traffic (to reduce delay from stadium
access)? A reliability monitoring system, as outlined in this guidebook, can tell an operator
which of these activities is worth the investment, and what the payoff will look like. Building
this system will add a new, powerful, practical traffic management tool to the arsenal of system
operators.
27
28
HOW SHOULD A TRAVEL TIME RELIABILITY MONITORING SYSTEM BE
STRUCTURED?
29
30
31
32
33
34
35
36
37
38
39
40
41
42
The process described above is designed to be integrated within an existing traffic
management system with a structure generalized in Figure ES-3. Inside the large box are the
three major modules of the TTRMS: a data manager, a computation engine, and a report
generator. The data manager assembles incoming information from traffic sensors and other
systems, such as weather data feeds and incident reporting systems, and places them in a
database that is ready for analysis as “cleaned data”. The computation engine works with the
cleaned data to prepare “pictures” of the system’s reliability: when it is reliable, when it is not, to
what extent, under what conditions, etc. In the figure this is illustrated by “regime TT-PDFs,” or
travel time probability density functions organized by regimes (i.e., the congestion level and the
type of non-recurring event, including none, like high congestion and an incident or low
congestion and work zone activity). The report generator responds to inquiries from users
(system managers or travelers) and uses the computation engine to analyze the data and provide
information that can then be presented back to the inquirer or decision maker.
ES-4
1
2
3
4
5
6
7
Each of these modules is discussed and described in the guidebook. In addition, case
studies and use cases illustrate how these modules work together to produce answers to questions
that managers would likely pose. The supplemental material provides further details about how
each of the modules should work, together and separately.
8
AN ILLUSTRATION
9
10
11
12
13
14
15
16
17
18
19
20
21
Figure ES-3: Block Diagram for a Travel Time Reliability Monitoring System (TTRMS)
As an illustration, Figure ES-4 shows the travel times for a specific trip on a freeway
(Interstate 5 in San Diego, California) throughout a typical weekday, excluding any nonrecurrent events such as incidents, weather, unusual demand, or special events. Time of day is
shown along the x-axis, and the travel time in minutes is shown on the y-axis.
It is clear from this figure that the travel times on this roadway segment are not always
the same; the system is unreliable, even without the influence of incidents, weather, etc. Not only
does the travel time vary but the spread in the times varies, slightly during the a.m. peak but
dramatically during the p.m. peak. At about midnight, the minimum and maximum are only 5
minutes different (50 minutes versus 55 minutes) but differ by 50 minutes during the weekday
p.m. peak without the additional influence of non-recurring events (50 minutes versus 100
minutes). This is the effect of recurring congestion.
ES-5
120
Travel Time (min)
100
80
60
40
20
0
0:00
1
2
3
4
5
6
7
8
9
10
11
12
6:00
12:00
Time of Day (hr:min)
18:00
0:00
Figure ES-4: Illustration of Variation in Travel Times by Time of Day Across a Year with NonRecurrent Events Excluded
Figure ES-5 shows the same travel times, now with non-recurrent events included. It is
clear that non-recurring events have an impact, with the spread between minimum and maximum
times widening. A good example is the effect of adverse weather, especially during the peak
period. Traffic incidents also have an effect on travel time reliability, as do special events and
unusually high demand. The TTRMS helps indicate when, why, and by how much travel time
will vary.
ES-6
180
160
Travel Time (min)
140
120
100
80
60
40
20
0
0:00
6:00
No Events
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Incidents
12:00
Time of Day (hr:min)
Demand
18:00
Special Events
0:00
Weather
Figure ES-5: Illustration of Variation in Travel Times by Time of Day Across a Year with NonRecurrent Events Included
Figure ES-6 shows an example from a different data source (Interstate 8 westbound in
San Diego, California) of what to expect as an output from the TTRMS. The figure shows plots
with the distribution of travel rates (seconds per mile) across a three-month period under two
identified operating conditions (uncongested conditions and high recurrent congestion) and
whether an incident is also present. The distributions for this example are shown in a cumulative
fashion (as a CDF); the location of each line shows how many travel rates are that value or
shorter. For example, when traffic incidents occur during heavy (recurrent) congestion, one half
(50%) of the travel rates are up to 65 sec/mile. That is, 50% of the travel rates are this long or
shorter/smaller. The 90th percentile travel rate is 100 seconds per mile. Or put another way, 9 out
of every 10 vehicles is traveling at that rate or faster.
The value comes from comparing one distribution with another. For example, analysts
can compare the distribution for high recurrent congestion and traffic incidents with high
recurrent congestion without incidents. Without incidents, 50% of the vehicles are traveling at
approximately 55 sec/mi instead of 65 sec/mi—considerably faster. And at the 90th percentile,
the difference is even more dramatic, being approximately 60 sec/mi versus 100 sec/mi.
ES-7
1.0
Cumulative Probability
0.9
0.8
0.7
Normal Conditions,
Uncongested
0.6
0.5
Normal Conditions,
High Congestion
0.4
0.3
Incident Conditions,
Uncongested
0.2
Incident Conditions,
High Congestion
0.1
0.0
40
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
60
80
100
Travel Rate (sec/mi)
120
140
Figure ES-6: How Travel Rates Are Affected by Congestion and Non-Recurring Incidents
The cumulative distribution function presented here as an output of the TTRMS are the
technical means by which many useful outcomes can be reached. Some examples include the
following:
Understanding the System: The TTRMS shows that the impact of incidents on
travel rates is dramatic, particularly during periods when the facility is already
congested due to normal traffic conditions (e.g., peak hour travel). This allows an
agency to prioritize funding towards measuring and respond to the types of nonrecurrent events that most significantly affect that agency’s system.
Communicating to the Public: The TTRMS can be used to communicate travel
times or ranges of travel times to the traveling public and to passenger and freight
movers, both pre-trip and en-route. This allows the public to be informed and a
partner in managing the demand on the system.
Measuring the Effects of Actions: The TTRMS can be used to measure the effect of
actions taken to improve reliability. The mitigating actions would be intended to
cause the travel times (or travel rates) during incidents to get much closer to those
when there are no incidents. Moreover, after the mitigating actions were taken, the
TTRMS would be able to show how reliability improved.
The guidebook contains many examples of potential applications of the use of a TTRMS
through a series of use cases. The use cases are organized around the various stakeholders that
use or manage aspects of the surface transportation system, including the following:
Policy and Planning Support: Agency administrators and planners that have
responsibility for and make capital investment decisions about the highway network.
Overall Highway System: Operators of the roadway system (supply), including its
freeways, arterials, collectors, and local streets and drivers of private autos, trucks,
and transit vehicles (demand).
ES-8
1
2
3
4
5
Transit Sub-system: Operators of transit systems that operate on the highway
network, primarily buses and light rail (supply) and riders (demand).
Freight Sub-system: Freight service suppliers (supply) and shippers and receivers
that make use of those services (demand).
IMPLEMENTATION
6
7
8
9
10
11
12
13
14
15
16
17
Much of the work of developing a TTRMS can be carried out by a team that includes
people with expertise in database management, statistics, traffic detection technology, and traffic
engineering. Executive-level guidance to the implementation team should include the following:
18
CONCLUSION
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
In conclusion, a TTRMS will help an agency understand the reliability performance of
their systems and monitor how reliability improves over time. It will help an agency answer a
variety of key questions, including:
What is the desired geographical scope of the system?
Are periodic reports (weekly, monthly, quarterly, or annual) sufficient, or does the system
need to provide continuous real-time reporting capabilities?
Is the reliability information primarily for the agency’s internal use, or is the intent to
provide information to the public in a way that will influence travel decisions?
How much detail is required? Is it sufficient to have a generalized comparison of
reliability on different routes, or is it desirable to know about the reliability of individual
trips on a specific highway segment?
What is the extent of congestion within the system (in space and time)?
What is the effect of non-recurring events (i.e., the seven sources of unreliability) on the
operation of the system?
How are freeways and arterials performing relative to performance targets set by the
agency?
Are capacity investments and other improvements necessary given the current
distribution of travel times?
Are operational improvement actions and capacity investments improving the travel
times and their reliability?
ES-9
1
CHAPTER 1
2
INTRODUCTION
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
This guidebook describes how to develop and use a Travel Time Reliability Monitoring
System (TTRMS). It explains why such a system is useful, how it helps agencies do a better job
of managing network performance, and what a traffic management center (TMC) team needs to
do to put a TTRMS in place.
Travel time reliability is the absence of variability in the travel times. If a system is
reliable, people can get to where they want to go, when they want to be there, all the time. If a
freeway is reliable, for example, then its travel times are the same for the same condition, all year
long. It is similar to a vehicle that always starts when the key is turned on. Of course, in reality
no system or roadway is perfectly reliable, and unexpected events sometimes significantly
reduce the reliability of a system. These unexpected events are due to internal and external
influencing factors like capacity reductions and traffic control devices (internal) and weather,
incidents, special events, work zones, and demand (external).
A TTRMS can help operating agencies, like TMCs, monitor the performance of their
system, understand the impacts of the various influencing factors, provide credible information
to the system users about what travel time reliability to expect, and decide what actions to take to
help improve reliability. This is the reason why a TTRMS should be created. Most TMCs
currently do not have these features or capabilities.
Creating a TTRMS involves creating a new module that plugs into an existing TMC
platform. The TTRMS relies on the TMC to gather the sensor data, manage the data processing
and storage, and communicate the results of its findings to the outside world. The TTRMS
focuses on using the incoming sensor data, along with supplemental information about the
influencing factors, and create a credible picture of how well the system is performing at the
present time and in the past.
The payoff is better reliability. The TTRMS will ensure that the TMC team knows what
how reliability will suffer if certain events take place. It will let them understand the impacts of
weather, incidents, special events, etc. It will also help the TMC team manage the network’s
reliability—in real time—by providing up-to-date information about how the variability in
segment and route travel times is either increasing or decreasing in response to actions taken.
It is clear that this variability exists. Figure 1-1 shows the various 5-minute travel times
observed across 2011 for Interstate 5 in San Diego, California. There are 72,000 data points
plotted in the figure. It is easy to see that during the p.m. peak, the travel times vary widely:
The variation is greater when congestion is higher, that is, when the demand-tocapacity ratio is high;
The variation is much lower off-peak, especially at night and early in the
morning, when congestion is low; and
There were impacts not only from congestion (which is internal to the system) but
from incidents, weather, special events, and unusually high demand.
1-1
180
160
Travel Time (min)
140
120
100
80
60
40
20
0
0:00
6:00
No Events
1
2
3
4
5
6
Incidents
12:00
Time of Day (hr:min)
Demand
18:00
Special Events
0:00
Weather
Figure 1-1: Five-Minute Average Travel Times on I-5 in San Diego
To fulfill its mission as a decision support tool, the TTRMS needs to execute four key
steps as illustrated in the conceptual diagram of information flow shown in Figure 1-2.
1-2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Figure 1-2: Information Flow in a TTRMS
Firstly, the TTRMS needs to effectively measure travel times. This is a complex
technical topic due to the variability of traveler behavior and the plethora of different
measurement sensors. Correctly measuring travel times along a given route requires a great deal
of systems development effort and statistical knowledge. This guidebook serves as a primer on
how to measure travel times effectively using available technologies and statistical techniques.
Measuring an individual travel time is the foundational unit of analysis for reliability monitoring.
Secondly, the TTRMS needs to clearly characterize the reliability of a given system.
This is the process of taking a set of measured travel times and assembling them into a statistical
model of the behavior of a given segment or route. In this guidebook, regimes refer to the
categories of conditions under which the segment, route, or network is operating at a given point
in time (or from one time to another). The statistical paradigm outlined in this guidebook is that
of using probability density functions (PDFs) to characterize the performance of a given segment
or route, usually specific to a particular operating regime (a combination of congestion level and
non-recurring event). This guidebook gives specific advice on the statistical decisions required to
effectively characterize the travel times. Characterizing the reliability of a segment or route is
fundamental to making good decisions about what to do to improve the performance of that
segment or route.
Thirdly, the TTRMS should identify the sources of unreliability. Once the reliability of a
segment or route has been characterized, transportation managers need to understand what
caused the unreliability (and how to “fix” it). This guidebook follows the causal list that FHWA
uses to describe why congestion arises, breaking these sources into the seven major influencing
factors described previously (two internal and five external). It discusses how to pull in data for
these influencing factors and effectively fuse them with the travel time data generated in
1-3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
previous steps. Identifying the travel times impacted by these sources of congestion is required
preparation for understanding system reliability.
Finally, the TTRMS should help operators understand the impact of these sources of
unreliability on the system. This final step in turning raw data into actionable decisions requires
both quantitative and qualitative methodologies: operators need clear visualizations of data, as
well as quantifications. This dual approach supports both data discovery and final decisionmaking about a given route. Understanding reliability is the key to good decision-making about
improving system reliability.
Once in place, a TTRMS that accurately and consistently executes these four steps
becomes a powerful tool for traffic management. It is a tool that decision makers can use to
understand how much of their delay is due to unreliability, and how to mitigate that delay. For
example, should a freeway operator deploy more service patrol vehicles (to clear incidents more
quickly) or focus their efforts on coordinating special event traffic (to reduce delay from stadium
access)? A reliability monitoring system, as outlined in this guidebook, can tell an operator
which of these activities is worth the investment, and what the payoff will look like. Building
this system will add a new, powerful, practical traffic management tool to the arsenal of system
operators.
18
THE CONTEXT WITHIN THE STRATEGIC HIGHWAY RESEARCH PROGRAM 2
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Reliability is one of the four focus areas of the Strategic Highway Research Program
(SHRP) 2, authorized by Congress in 2006. The purpose of the reliability focus area is to “reduce
congestion and improve travel time reliability through incident management, response, and
mitigation” (1). Four themes have been established under this focus area:
Theme 1: Data, Metrics, Analysis, and Decision Support
Theme 2: Institutional Change, Human Behavior, and Resource Needs
Theme 3: Incorporating Reliability into Planning, Programming, and Design
Theme 4: Fostering Innovation to Improve Travel Time Reliability
The project under which this guidebook has been prepared is SHRP2 Project L02,
Establishing Monitoring Programs for Travel Time Reliability. The project supports the first of
these themes by providing guidance to operating agencies about how they can put better
measurement methods into practice and understand the relationship that travel time reliability has
to the seven major sources of nonrecurrent congestion that have been identified (2):
Traffic incidents,
Work zones,
Weather,
Special events,
Traffic control devices,
Fluctuations in demand, and
Inadequate base capacity.
1-4
1
2
3
4
5
6
7
8
A key part of the delivery of travel time reliability to the traveling public is the presence
of a travel time reliability monitoring system that can assist engineers, planners, managers, and
policy makers in making sound traffic management decisions.
9
STRUCTURE OF THE GUIDEBOOK
10
11
12
13
Source: Strategic Highway Research Program 2
Figure 1-3: Seven Major Sources (Factors) of Nonrecurrent Congestion
The guidebook follows the block diagram presented in Figure 1-4 for purposes of
describing the TTRMS. Each module is shown as a box, while the inputs and outputs are shown
as circles. “Other Systems” refers to systems that monitor the influencing factors.
1-5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure 1-4: Reliability Monitoring System Overview
Five chapters are used to define and describe the TTRMS:
Chapter 1, Introduction, presents an overview of travel time reliability.
Chapter 2, Data Collection and Management, discusses the types and application
of various types of sensors, the management of data from those sensors, and the
integration of data from other systems that provide input on sources of
unreliability (e.g., weather, incidents). This represents the left side of the figure in
Figure 1-1 and includes traffic sensors, other systems, and the data manager.
Chapter 3, Computational Methods, discusses how probability density functions
are derived from the variety of data sources. This represents the center part of the
figure in Figure 1-1 and includes the process of generating travel time probability
density functions that can be used to derive a variety of reports to users.
Chapter 4, Applications, presents a discussion of potential use cases for a travel
time monitoring system and illustrates its use with a series of real-world case
studies.
Chapter 5, Summary, illustrates the entire the analytical process of a TTRMS.
The guidebook is supplemented by four documents that provide additional detail to
support the development and application of travel time monitoring systems. These documents are
as follows:
Supplement A: Monitoring System Architecture. This document presents examples
of detail data structures for the organization of various data sources. This
document provides supporting detail for Chapter 2 of this guidebook.
Supplement B: Methodological Details. This document presents detailed
discussions of the analytical methods that can be used to calculate travel time
reliability measures from a variety of input sources. This document provides
supporting detail for Chapter 3 of this guidebook.
Supplement C: Case Studies. This document presents a series of detailed case
studies that exercise various aspects of the guidebook, including system
architecture, analysis of recurrent and nonrecurrent sources of congestion, and the
1-6
1
2
3
4
5
6
7
8
application of a variety of use cases. This document provides supporting detail for
Chapter 4 of this guidebook.
Supplement D: Use Case Demonstrations. This document illustrates the
application of a variety of use cases for a travel time reliability monitoring
system. This document provides supporting detail for Chapter 4 of this
guidebook.
HOW TO USE THIS GUIDEBOOK
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Agency staff should read this material with an eye toward understanding what a TTRMS
can do and what features make the most sense for their own situation. For example, from which
sources can an agency collect data? Which steps can an agency employ in processing data? What
performance measures can an agency realistically calculate? How can information be presented
to various end-users? As such, this document can be used as a starting point for an agency to
review current practices, recent reliability research, and developing technologies with the aim of
designing a system that meets its own needs.
The use cases and case studies presented in Chapter 4 helped guide the preparation of the
module descriptions. Use cases are a formal systems engineering construct that transforms user
needs into short, descriptive narratives that describe a system’s behavior. Use cases are used to
capture a system's behavioral requirements by detailing scenario-driven threads through the
functional requirements. The collective use cases define the monitoring system by capturing its
functionalities and applications for various users. Each use case defines the purpose for the
analysis, the steps the user must take to obtain the desired data, the results of the steps, the spatial
aggregation used, and the travel time reliability measures that can be reported. The measures
listed in these use cases are only examples and represent possible ways that reliability
information can be conveyed.
Travel time monitoring increasingly incorporates a blend of infrastructure from the public
and private sectors. As a result, some elements of travel time monitoring systems in current use
have proprietary components. This guidebook is not intended to provide a bid-ready
specification of all aspects of a travel time monitoring system, some aspects of which may be
proprietary. Nevertheless, the reader will be able to better understand the functional specification
of such a system, defined by the required inputs and outputs of such a system.
33
WHAT IS TRAVEL TIME RELIABILITY?
34
35
36
37
38
39
40
41
42
43
This section provides a slightly more technical description of travel time reliability.
While no consensus yet exists, Elefteriadou and Ciu (3) provide a starting basis in their study of
travel time reliability definitions. As they discern, Ebeling (4) provides a useful basic idea,
defining reliability as “the probability that a component or system will perform a required
function for a given period of time when used under stated operating conditions. It is the
probability of a non-failure over time.” It is slightly different from the idea of consistency, which
has to do with the absence of variability.
Brought into the highway network context, the system is reliable (formally speaking) if
each traveler or shipper experiences actual times of arrival (ATA) which match desired times of
arrival (DTA) within some window, as shown in Figure 1-5. In some cases the ATA and DTA
1-7
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
are extremely important; in other cases they are not, depending on trip constraints and penalty at
the destination. For example, a trip to catch a plane is more constrained than a trip to the store.
If the ATA lies outside the DTA window, especially if the ATA is after the DTA, a
reliable trip was not completed. Hence, the transportation system is reliable, technically
speaking, if the ATAs all lie within their DTA window. Otherwise, the system has “failed”. As
Elefteriadou and Ciu (3) point out, such a definition of reliability becomes well defined.
In a more general sense, the reliability of the system can be measured using utility theory,
as described for example by Hansson (5). Utility (of the trip) is maximized if the ATA is inside
the DTA window. Conversely, disutility is greater if the ATA lies outside the DTA window; and
the aggregate disutility for all trips among all users is the “societal cost” of the system’s
unreliability. The function that evaluates the disutility may be symmetric or asymmetric
depending on the situation, as shown in Figure 1-6. Truckers incur significant penalties if they
are either late or early in delivering shipments to the receivers. Individual travelers can be late
for appointments or miss the opportunity to insert additional tasks like stopping for coffee or
sleeping later if they are early.
Figure 1-5: Concepts of Desired and Actual Times of Arrival
1-8
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure 1-6: Disutility Function to Characterize Desired and Actual Times of Arrival
If it were now possible, today, to observe all the trips, with the DTA windows being
known, then the reliability of a given transportation system could actually be assessed. One could
compute the percent of ATAs that fall within their DTA windows. This would be a useful metric
both for the entities making the trips as well as the organizations providing the service (e.g., the
TMC or transit system operator). The aggregate disutility could also be computed by summing
the disutility values for each trip. Obviously, this world does not presently exist.
What can be observed today, at least in part, are travel times on segments and routes in
the network. For example, some TMCs can monitor probes, vehicles equipped with tags in areas
where toll roads exist, and others can generate speed distributions at specific point locations in
the network where sensors (speed traps) are installed. As a result, the TMC can establish desired
travel times (DTT) or, better yet, desired travel rates (DTR) in seconds per mile so that the length
of the facilities does not interfere, with performance levels to be achieved on the segments and
routes, consistent with Ebeling (4). These DTRs can be dependent upon the regime under which
the system is operating (combination of the influencing factors) and they can be adjusted over
time as the network conditions change – demand grows and/or improvements are made.
If single value measures seem easiest to master or communicate, the median travel rate is
a good choice. If more complex and powerful tools are desired, travel rate probability density
functions (TR-PDFs) are a good choice. Research conducted for the development of this
guidebook found that in general, travel times do not follow a normal (Gaussian or bell-shaped)
statistical distribution. As a result the median value (half above and half below) is generally a
better measure of central tendency than the more familiar average (mean) value. Similarly, the
semi-variance was found to be a better measure of variation than the standard deviation.
A segment or route is then performing reliably if its actual travel rate (ATR) lies within
the acceptable DTR window given the regime under which the segment or route is operating.
The TMC team can monitor the number of segments or routes whose ATR lies within the DTR
window; they can see how that number varies based on the network, segment, or route operating
conditions (e.g., an incident during high congestion); and actions can be identified to increase the
number of segments or routes whose ATR is within its DTR window.
1-9
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
This paradigm can also be extended to the system users. Trips can be considered
successful if their actual travel rate (ATR) falls within an allowable DTR window based on the
conditions under which the trip was made. Reliability can be measured by the percentage of trips
whose ATR fall within the allowable DTR window. By extension, the aggregate disutility
experienced by the travelers or shippers can be assessed, in principle, using disutility functions
which compare the ATRs one-at-a-time with their corresponding DTRs and then sums the
results.
Service providers want to see if different ways to operate the system would be likely to
produce better alignment between the ATRs and the DTRs (or if capacity investments are
needed). Naturally, this decision making is aimed at variability reduction and shifts in the median
values either lower or higher so that the requisite confidence interval objectives are met given the
DTT windows.
The decision making by the TMC team is similar to the mean-variance tradeoff analyses
so prevalent in financial planning (see, for example, Maginn et al. (6)). In this instance, the
tradeoff is between minimizing the median travel times, as in building new network links or
adding capacity to reduce the median travel rates, versus taking actions like improving incident
response or managing the impacts of weather better so that the variation in the travel rates is
reduced (getting more of the ATRs within their DTR windows).
19
SUPPLY-SIDE AND DEMAND-SIDE PERSPECTIVES
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Without overly stressing the thought, supply-side and demand-side ideas do have a
bearing on travel time reliability. The supply side focuses on the TMC team and its interest in
managing the segment, route, and network-level reliability so that performance goals are met.
Most likely, these goals are predicated on simple metrics like the median travel rate, rather than
the distribution of individual vehicle travel times. In addition, consistency across time (in a
longitudinal sense) is the primary concern: how does the team make sure that every a.m. peak
has an OK reliability performance?
The demand side focuses on the individual travelers and the travel times they experience
in making trips. Travelers’ decisions regarding modes, routes and departure times are associated
with their own socio-demographics and the context in which they travel. Observations of
individual vehicle (trip) travel times are important, and the travel time density functions of
interest pertain to travel times experienced by individual travelers at specific points in time. That
means the reported percentiles of those travel times (e.g., the 15th, 50th, and 85th percentiles)
pertain to the travel times for those travelers.
The important thought to carry away from this discussion has to do with making sure the
context of the analysis is understood. Two analysts could think they are talking about the same
reliability ideas, but in fact, the one is focused on the reliability of median travel times for the
same five-minute time periods across a year—the duty cycle for the network—while the other is
focused on the travel times experienced by individual vehicles in those same five-minute time
periods across the year, or for a single five-minute time period (or an hour) for a given network
on the same day, which is more of a cross-sectional data analysis thought. Neither analyst has the
“right” or “wrong” perspective; both are very valuable. However, before they go too far in their
discussion, they need to make sure they are both thinking about the problem in the same way, or
that they understand how their perspectives are different and relate to one another.
1-10
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
TRIP-MAKING CONCEPTS
Several trip-making concepts are important in understanding the material presented in
this guidebook. The first and most important concept is that trips made by persons and packages
are the fundamental events on which travel time reliability is focused. Personal travel time
reliability has to do with individual person trips, and freight travel time reliability has to do with
individual package shipments.
The second concept is that travel times and trip times are different. A trip refers to an
origin, a destination, and implicitly a trip purpose. Trips can be chained trips involving
intermediate stops, and thus trip times can be different from travel times. Trip times are the endto-end times involved—from origin to destination—and allow for intermediate stops and sidetrips. Travel times have such extra times removed and refer to the amount of time spent
traveling.
However, this distinction also pertains to observations of times for segments and routes.
The trip time is the elapsed time between when one sensor identified a vehicle and a second
sensor at some other location identified the same vehicle at a later time. Trip times can be travel
times, but not necessarily. The trip time could include stops and/or detours, since the specific
route of the vehicle between the two sensors might not be known. Travel time refers to the actual
driving time between the sensors. Trip times need to be filtered in order to obtain accurate travel
times.
The third concept is that of a travel rate. This is travel time per unit distance, as in
minutes per mile. While travel times help travelers understand how long it will take them to
accomplish their trips—and help system managers determine what travel times to display on
variable message signs, as an example—travel rates help system managers compare the
performance of one segment with another so that strategies can be developed for making capital
investments that help to improve the performance of the segments and the network as a whole.
Travel rate probability density functions (PDFs) are distinguished from travel time PDFs by
notation (TR-PDFs versus TT-PDFs).
The fourth concept—an acceptance of the current technological frontier—is that
individual person and package trip and travel times are presently difficult to observe. The best
options presently available are automatic vehicle identification (AVI) and automatic vehicle
location (AVL) based sensors. When those technologies become more prevalent, it will be easier
to collect data on individual trips.
The final concept is that it is acceptable to use aggregate measures as surrogates for the
more detailed picture of individual trips. The aggregates are derived from the more detailed
information (e.g., as in the average speeds provided by single-loop and double-loop detectors).
Also, single-occupant vehicles are a significant portion of the traffic stream, especially during
congested conditions, so insights about individual person trips can be seen in the vehicle data.
The main point is that travel time reliability and the ideas about it presented in this
document are intended to address questions about individual person and package trips, even
though surrogate data are used. This means the methodology is designed to provide insights
about the travel time reliability that might be experienced by persons and packages by extension
of the insights derived from analyses of more aggregate measures.
1-11
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
PROBABILITY DENSITY FUNCTIONS
Much of the material presented in this guidebook emphasizes the creation and analysis of
probability density functions (PDFs) to describe the distribution of travel times and travel rates
and to compare the performance of one facility with another or one operating condition with
another.
These distributions are often presented three ways. The first is as a histogram, where bar
heights are used to represent the relative frequency with which specific conditions pertain. Figure
1-7 shows a histogram of travel times for Interstate 8 in San Diego during the a.m. peak for
various operating conditions: when everything was normal, when the system was being affected
by an incident, and when the system was being affected by weather. No special events were
observed in the selected data set.
Figure 1-7: Example of Travel Time Histogram for Various Event Conditions
The second way in which these distributions are presented is via a probability density
function (PDF). Figure 1-8 shows the corresponding PDFs for the three conditions. The PDF is a
1-12
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
density function that is based on the values in the histograms. In the PDF, as in the histograms,
one can clearly see that some travel times are more common than others, and that the distribution
of the travel times is different for the various operating conditions. Note that these histograms of
San Diego data do not follow the familiar bell-shaped curve. The data has a significant “skew”
even after it has been categorized.
Figure 1-8: Example of Probability Density Functions for Various Event Conditions
The third way in which these distributions are presented is in a cumulative density
function (CDF). The CDF is based on the PDF in that the value shown in the CDF at any point in
the graph is the integral of the PDF up to that point (i.e., the area enclosed within the PDF above
the horizontal axis. A property of the PDF is that its area sums to 1.0 which means the CDF
ultimately rises to a maximum of 1.0. Figure 1-9 shows the CDFs for the various regimes
associated with the performance of a different facility in San Diego (Interstate 5, the same
facility illustrated previously in Figure 1-1). As with the PDF, one can clearly see differences in
the distribution of the travel rates (as compared to times) and that the distribution of rates for
some regimes is much different than for others. It is through the use of these tools—the
histogram, PDF, and CDF—that the case studies and use cases reach conclusions about the
influence of various factors on the travel times and travel rates.
Related to these thoughts about distributions and portraying them, the distributions can
be, and often are, multimodal (in a statistical sense, meaning more than one point at which the
1-13
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
probability density function reaches a maximum). This is because the data may reflect more than
one operating condition or regime across the span of time being studied or the users may have
experienced different treatments during the analysis timeframe. An illustration of the former
arises when the average travel times for a given 5-minute time slice are studied across a year. It
is likely that different operating conditions existed on different days (because of weather and
incidents let alone different demand levels), so multimodality should be expected. An illustration
of the latter is when travelers do not all experience the same control treatment. For arterial
networks, this could arise when some users are able to progress between traffic signals without
stopping while others are not. For freeway networks, it can be because some vehicles are delayed
by ramp metering controls while others are not; or when some vehicles experience delays from
paying tolls (e.g., cash) while others do not (e.g., toll tags).
Figure 1-9: Example of Cumulative Distribution Functions under Various Regimes
(Operating Conditions)
Since the word “mode” is used in other ways in transportation, the word regime is used
instead of mode to describe these various operating conditions (or sub-conditions). Moreover,
common traffic engineering names are used to describe these modes like “congested”,
“uncongested”, “transition”, “incident”, “weather”, etc. The regimes help enhance the quality of
the PDFs. It keeps them from being noisy, and it helps maximize the incremental value derived
from the data acquired every day.
The last concept—and an important one—is that all the reliability metrics of interest can
be derived from these probability density functions (PDFs). The PDFs completely describe the
travel times or travel rates (travel times per unit distance). Hence, the typical metrics of interest
for characterizing reliability—planning index, buffer index, average, median, 95th percentile, or
others—can be computed based on the PDFs. As a result, these PDFs, supplemented by ancillary
data about the environment that does (or will exist) in the timeframe of the analysis (e.g.,
weather, incidents), represent sufficient information to answer the questions about measuring
reliability.
1-14
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
REFERENCES
1. Transportation Research Board. Providing Reliable Travel Times on the Highway
System. Reliability Focus Area Overview, Strategic Highway Research Program 2. June
17, 2009. http://onlinepubs.trb.org/onlinepubs/shrp2/RRPJune2009.pdf. Accessed
September 2, 2012.
2. Cambridge Systematics, Inc., and Texas Transportation Institute. Traffic Congestion and
Reliability: Trends and Advanced Strategies for Congestion Mitigation. Federal Highway
Administration, September 2005.
3. Elefteriadou, L., and X. Ciu, “Review of Definitions of Travel Time Reliability,”
Proceedings, 86th Annual Meeting of the Transportation Research Board, Transportation
Research Board of the National Academies, Washington, D.C., 2007.
4. Ebeling, C. E. Introduction to Reliability and Maintainability Engineering. McGraw-Hill,
1997.
5. Hansson, S.O. Decision Theory: A Brief Introduction, Department of Philosophy and the
History of Technology. Royal Institute of Technology, Stockholm, Sweden, 2005.
6. Maginn, J. L., D. L. Tuttle, D. W. McLeavey, and J. E. Pinto. Managing Investment
Portfolios: A Dynamic Process. CFA Institute Investment Series, John Wiley & Sons, 3rd
Edition, 2007.
1-15
1
CHAPTER 2
2
DATA COLLECTION AND MANAGEMENT
3
4
5
6
7
8
9
10
11
12
13
14
As a first step in reliability monitoring data collection, agencies need to thoroughly
evaluate the existing data sources in their region and determine how they can be leveraged to
support travel time computations. Following this step, agencies can take appropriate measures to
determine how these sources can be integrated into the reliability monitoring system and identify
where existing infrastructure should be supplemented with additional sensors or data sources.
This section describes the potential sources for traffic data that can support the
computation of route-level travel times and the potential sources of non-traffic data related to the
sources of congestion.
Figure 2-1 summarizes the data flow from various data sources into route-level travel
time PDFs. This chapter discusses the lower portions of the figure, namely the traffic data
sources and data types. The processing steps for converting the various data sources into routelevel travel time PDFs is presented in Chapter 3.
15
16
17
18
19
20
21
22
23
24
25
Figure 2-1: Types of Data Collection Sources
As indicated in Figure 2-1, there are two broad categories of traffic data sources—
infrastructure-based sources (devices mounted along the roadside) and vehicle-based sources—
and these can be further separated into four broad categories of traffic data sources:
1. Infrastructure-based detectors that can sense volume, occupancy and other measures,
but not speeds.
2. Infrastructure-based detectors that can sense speeds as well as other measures.
3. Automated vehicle identification (AVI) systems.
4. Automated vehicle location (AVL) systems.
2-1
1
2
3
4
5
6
7
8
9
10
11
Public agencies typically own and operate the infrastructure-based detectors and the AVI
systems (used for tolls) while private, third-party sources often own and operate the AVL
systems (although transit agencies are increasingly using AVL systems). This chapter outlines
the differences in capabilities between these traffic data collection types and discusses the
benefits and drawbacks of available technologies. Figure 2-2 and Figure 2-3 respectively show
the categories and types of infrastructure-based and vehicle-based sensors described in this
chapter.
Figure 2-2: Types of Infrastructure-Based Data Collection Sources
2-2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Figure 2-3: Types of Vehicle-Based Data Collection Sources
INFRASTRUCTURE-BASED SOURCES
Infrastructure-based detectors, which include loop and radar detectors, are already a
common component of traffic management systems in many regions. Some can measure vehicle
speeds directly while others use post-processing algorithms to estimate speeds based on counts
and occupancy. The ones that can directly measure speeds are more valuable. These sensors have
a limited ability to monitor travel time reliability. While very prevalent, the drawback of these
technologies is that they only provide data values at one fixed location along the roadway. As
discussed in the framework section, this means that they can only report spot speeds.
Consequently, they cannot directly inform on an individual vehicle’s route or time of travel
between two points. As a result, the data they transmit requires some processing and
extrapolation before travel times can be calculated. This also means that the accuracy of the
travel time measures they produce is a function of how frequently detectors are spaced along the
roadway. If existing deployments have detectors spaced at a frequency of one-half mile or less,
they are suggested for inclusion in a reliability monitoring system. If detectors are placed less
frequently on key routes, agencies may want to consider either installing more detectors or
supplementing the existing detection with AVI sensors.
Data from infrastructure-based sources needs to be reviewed for quality before
calculating travel time reliability metrics. An example of this is presented in Chapter 4 in the
Northern Virginia case study, where a process is illustrated for identifying detectors that are
stuck or not collecting data for inductive loops and radar-based sensors. The Northern Virginia
case study also describes challenges and mitigations for integrating the data feed from
infrastructure-based detectors with a reliability monitoring system.
2-3
1
2
The following types of technologies are considered infrastructure-based sources.
Loop Detectors
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Loop detectors are located in-pavement on many roadway facilities. They have
historically been the most common traffic monitoring tool due to their relatively low installation
cost and high performance. Coverage, however, varies greatly among cities and states. In many
urban locations, they are common on freeway facilities. Many arterials also use loop detectors to
control actuated and adaptive traffic signals. However, it should be noted that loop detectors used
in traffic-responsive signal systems are usually not well-adapted to providing the data required to
support reliability monitoring. Firstly, due to memory constraints in the traffic signal controller,
data are usually not collected for each individual lane and are usually discarded rather than being
sent to a control center for archiving and analysis. Even in locations where data are sent to a
control center, they are usually aggregated up to levels that are too high (for example, hourly
flow and occupancy values) to enable accurate travel time calculations. However, in some cases
it will be possible for agencies to modify the existing signal system sensors to collect additional
data and transmit them to a centralized location to support reliability monitoring.
Loop detectors typically measure traffic volumes and occupancies and send data to a
centralized location every 20 to 60 seconds. From these data, spot speeds can be calculated with
a reasonable accuracy, at least in freeway applications, and can be used to extrapolate travel
times. Loop detectors in a dual configuration can directly report speed values. Loop detectors
have two significant drawbacks: their intrusive installation, and their significant maintenance
requirements stemming from their vulnerability to damage or destruction.
22
Wireless Magnetometer Detectors
23
24
25
26
27
28
29
30
31
32
33
34
Like loop detectors, wireless magnetometer detectors are located in the roadway surface.
However, wireless magnetometer detectors can be installed simply by drilling a hole into the
pavement, eliminating the need for cutting pavement during installation and reducing
maintenance requirements. These sensors use radio signals to communicate with access points
located on the roadside, usually on poles, preventing the need to hardwire a detector to a
controller cabinet. Like loop detectors, they report volume and occupancy data with a granularity
that depends on the sensor’s setting. Sensors in a dual configuration can also directly report
speed values. The data accuracy of wireless magnetometer detectors is similar to that of
inductive loops (1). As such, where agencies would like to install additional in-road
infrastructure detectors, wireless magnetometer sensors are a good alternative to loop detectors.
Recent developments have also adapted some wireless magnetometer detectors to re-identify
vehicles at a second detector, giving them AVI capabilities that are described later in this section.
35
Video Image Processors
36
37
38
39
40
41
Many agencies have begun installing video image processors, especially on arterial
facilities, as an alternative for loop detection. Video image processing can retrieve volume,
occupancy, and speed data from cameras on the roadway. This technology usually requires users
to manually set up detection zones on a computer that are in the field-of-view of each camera,
meaning that it is important that the cameras not be moved and the detection zones be set up
correctly. Some specialized systems can also re-identify vehicles detected at two separate
2-4
1
2
3
4
5
cameras (2), giving them AVI capabilities discussed later in this section. This technology could
be a viable method for travel time reliability monitoring where agencies already have video
cameras installed. Video-based detection generally works best in locations with relatively mild
weather, as they are sensitive to challenging weather conditions such as snow, fog, or
temperature change.
6
Radar Detectors
7
8
9
10
11
To overcome the intrusive installation and maintenance of loop detectors, many regions
have deployed microwave radar detectors, which are placed overhead or roadside and measure
volume and speed data. One drawback to radar detectors is that they can lose their speed
calibrations. Radar detectors are a viable option for agencies that want to increase the frequency
of data collection infrastructure along a roadway without installing additional loop detectors.
12
Other Infrastructure-Based Sources
13
14
15
16
17
There are a number of additional overhead vehicle detection technologies that have
capabilities similar to microwave radar detectors. These can be considered on a site-specific
basis or used for travel time reliability monitoring where they have already been deployed. These
technologies include passive infrared sensors, ultrasonic sensors, passive acoustic array sensors,
and combinations thereof. (2)
18
AUTOMATED VEHICLE IDENTIFICATION SOURCES
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Automated Vehicle Identification (AVI) data collection sources, which include Bluetooth
readers and License Plate Readers, detect a passing vehicle at one sensor, then re-identify the
vehicle at a second sensor, allowing the vehicle’s trip time between two points to be directly
computed. The drawback of AVI technologies is that while they provide the trip time between
two points, they cannot inform on the route taken by individual vehicles, or whether the trip
included any stops. Since there are often multiple ways to travel between two points, especially
in urban areas, and since the travel time without stops is the value of interest in reliability
calculations, processing and filtering is required to ensure that reliability computations are based
on representative travel times for a given route. Techniques to perform this data processing are
described in Chapter 3, and an example is illustrated in the Sacramento/Lake Tahoe case study in
Chapter 4. Inaccuracies can also be reduced by deploying sensor readers at frequent intervals, to
reduce the likelihood that a vehicle took a different route than the one assumed in the
computation. Supplement B provides guidance about sensor spacing and sampling. The
following technologies are sources for AVI trip time data.
33
BluetoothTM
34
35
36
37
38
39
40
BluetoothTM receiver technology has only recently been applied to traffic data collection,
but appears promising for measuring trip times, especially on arterials. In the context of traffic,
Bluetooth detectors record the public Media Access Control (MAC) address of a driver’s mobile
phone or other consumer electronic device as the vehicle passes a point. This recorded ID
number (or a curtailed version of it, to reduce privacy concerns) can then be matched as the
vehicle passes subsequent detectors, allowing travel times between points to be calculated. Some
vendors are also developing Wi-Fi detectors based on similar principles.
2-5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
This technology is advantageous in that it is accurate, low-cost, and portable. A
drawback, however, is that currently only a small percentage of drivers have Bluetooth-enabled
devices in their vehicles; estimates range from 5 percent in the Washington D.C. metropolitan
area to 1 percent outside of Indianapolis (3,4). Still, it can be assumed that these percentages will
grow as commercial Bluetooth applications, particularly smart phones, become more prevalent,
making Bluetooth an important data collection alternative for future projects.
There are a few issues with Bluetooth measurements that need to be accounted for in the
data filtration process. Firstly, Bluetooth readers frequently record the same wireless network ID
more than once as a vehicle passes, especially when vehicles are traveling slowly. These
duplicate addresses have to be removed to avoid counting a vehicle’s travel time more than once.
Secondly, Bluetooth readers have a wide detection range that could collect travel times that do
not reflect actual conditions. For example, a Bluetooth sensor station on a freeway might detect a
vehicle that is in a queue on an entrance ramp, thus generating a travel time that is not
representative of mainline freeway traffic flow. These unrepresentative trip times have to be
filtered out during data processing. Additionally, on arterial streets, Bluetooth readers report trip
times from non-vehicular modes like walking or cycling; these times have to be removed in the
data cleaning process.
18
License Plate Readers
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
License plate readers (LPR) employ cameras that capture a digital image of a vehicle’s
license plate and use Optical Character Recognition (OCR) to read the plate number. While
primarily used for toll enforcement, LPR can also be used to calculate trip times for vehicles that
pass by two or more cameras.
The advantage of LPR is that it can collect trip time samples from vehicles without
requiring the presence of any specific device within the vehicle. This method, however, is not
well-suited for data collection on high-speed freeways. Additionally, plate matching is not
always accurate, especially during adverse weather conditions (5). The equipment needed is also
costly, and there are privacy concerns that come with tracking a vehicle by its license plate
number. In a travel time context, LPR has been used by the Florida Department of
Transportation to distribute travel times to Changeable Message Signs in various Florida cities,
as well as on the Arizona State Route 68 Work Zone Project near the Hoover Dam. (6)
The FHWA Travel Time Data Collection Handbook includes the following guidelines for
using LPR (5):
Place cameras at mid-route checkpoints such as major interchanges, intersections,
jurisdictional boundaries, and transition points between different roadway crosssections or land uses with the following spacing (which can vary depending on the
roadway network and the desired detail of study):
o Freeways/Expressways, high access frequency: 1 to 3 miles
o Freeways/Expressways, low access frequency: 3 to 5 miles
o Arterial Streets, high cross-street frequency: 0.5 to 1 miles
o Arterial Streets, low cross-street frequency: 1 to 2 miles
50 license plate matches is the target sample size for a given roadway segment and
time period to accurately represent travel time variability. It should be noted that the
percentage of successful matches is about 5 percent to 20 percent in a given period.
2-6
1
2
3
Due to LPR’s accuracy issues and high cost, it is recommended that only those locations
that have already installed LPR infrastructure use it as a primary method of data collection for
reliability monitoring.
4
Radio-Frequency Identification
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Radio-Frequency Identification (RFID) technology is employed in electronic toll
collection (ETC) and can be used to re-identify vehicles for travel time purposes. RFID is
embedded in toll-tags such as E-ZPass on the East Coast and FasTrak in the San Francisco Bay
Area. More than 20 states currently have locations that use RFID toll tags, including some that
have collected travel time data from their systems, including the San Francisco Bay Area, New
York/New Jersey, Houston, and Orlando (7,8). The iFlorida toll tag travel time project found that
toll tag penetration is high in urban areas with toll roads, but much lower in other areas. This
means that this data collection option is best suited for urban areas with a high toll tag saturation
rate. The study found comparable rates of saturation between urban freeways and urban arterials;
however, the percentage of vehicles that could be re-identified at a second sensor was lower for
arterials due to the fact that more vehicles enter and exit the facility between sensor stations. As a
result, in Orlando, toll tag readers usually only generated between 10 and 20 travel time
estimates per hour (8). Agencies should thoroughly evaluate their regional saturation rate of
RFID toll tags to determine whether this technology can supply the number of travel time
samples needed to robustly estimate reliability measures over time.
RFID is also an alternative for areas that do not have ETC if agencies distribute RFID
tags to volunteer drivers. This was done in San Antonio as part of the TransGuide traffic
monitoring program, where 30,000 tags were distributed in the first 10 months of operation (9).
The FHWA’s Travel Time Data Collection Handbook offers the following suggestions to
increase the number of RFID tags in the regional vehicle fleet:
Priority passage through toll facilities;
Priority passage through weigh/inspection stations (for trucks);
Ability to receive in-vehicle travel information;
Free electronic tag;
Voucher for merchandise or fuel;
Registration for roadside maintenance service; or
Some type of user incentive payment.
Aside from sample size concerns, privacy issues are raised because RFID transmits data
that is identifiable to an individual vehicle. Therefore, if RFID is used to collect travel times, the
system will need to encrypt data to remove personal information. The iFlorida deployment does
this by sending the DOT database an encrypted hash that represents the toll tag number, rather
than the actual toll tag number itself.
37
Vehicle Signature Matching
38
39
40
41
42
43
Vehicle signature matching refers to methods that match the unique magnetic signature of
a vehicle as it passes over a detector to the same signature from an upstream detector. Single
loop, double loop, and wireless magnetometer detectors all have this capability. While loops are
not capable of matching every vehicle, research and testing of this method has shown that it can
match enough vehicles to provide accurate travel time distributions for both freeways and
arterials. (10)
2-7
1
2
3
4
5
6
7
One advantage of this method is that it can use existing detectors in new ways that
improve travel time data accuracy. For arterials, it is advantageous over traditional detector data
since it estimates travel times without the need for signal phase information. It also offers an
additional benefit over other AVI technologies: it avoids potential privacy concerns through
anonymity. This technology has only seen limited use in practice thus far, with projects in a few
locations in California, but appears promising for measuring travel times on both freeways and
arterials. (10)
8
AUTOMATED VEHICLE LOCATION SOURCES
9
10
11
12
13
14
15
Automated Vehicle Location (AVL) refers to technologies that track a vehicle along its
entire path of travel. These methods provide the most accurate and direct measurements of travel
times, but have not yet seen deployment sufficient to provide reliable data on a regional scale.
This will change as more vehicles become equipped with AVL technologies and agencies
become more accustomed to using them for real-time data collection. An example of
incorporating AVL equipment to analyze transit performance is illustrated in the San Diego case
study in Chapter 4.
16
Global Positioning Satellite
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Any vehicle equipped with a Global Positioning Satellite (GPS) receiver can be tracked
along its path of travel to calculate route-based travel times and other traffic data. GPS
technology is well-suited for accurate travel time calculations because it can pinpoint a car’s
location within a few meters and its speed within 3 miles per hour. (13)
GPS has traditionally been used to calculate travel times through test probe vehicles
equipped with GPS receivers. The value of this data is limited because of the small number of
test probe vehicles typically deployed, and it does not provide real-time data on a permanent
basis. However, even in a more advanced system that monitors all GPS-equipped vehicles in-real
time, the low market penetration rate of GPS technology will be a constraint on the ability to
accurately represent travel time variations. However, it can be reasonably assumed that more
vehicles and devices will have GPS capabilities in the future.
GPS is also used by many transit agencies to monitor bus locations and schedule
adherence in real-time. As such, another alternative for agencies looking to monitor reliability is
to use equipped buses as travel time probes. By identifying and separating out bus-specific
activities, such as dwell times and different acceleration rates, arterial travel times can be
estimated from bus AVL data (11).
33
Connected Vehicle
34
35
36
37
38
39
40
41
The Connected Vehicle (CV) research program, sponsored by the U.S. Department of
Transportation (USDOT), is focused on leveraging wireless technology to allow vehicles and
roadway facilities to communicate with one another, with the aim of improving safety,
monitoring conditions, and providing traveler information (12). The majority of Connected
Vehicle research will be completed by 2013; therefore, at this point it is impossible to know the
full scope of the contributions that the research will make to reliability monitoring efforts. At this
point, however, it seems that Connected Vehicle technologies could provide a rich source of
travel time information, since the Vehicle to Infrastructure (V2I) communication channels
2-8
1
2
implemented through the Connected Vehicle program could be used to send collected vehiclespecific location data to a central data server for travel time processing.
3
Cellular Networks
4
5
6
7
8
9
10
11
Cellular telephone networks track cell phones to hand them off to different base stations
as they travel, and travel times can be calculated through this information. The precision of
location data increases with the number of cellular towers for which a phone is in range. In urban
areas, location accuracy can be within a hundred feet (15), which in some cases is too large to
assign vehicles to a specific link, especially in dense urban networks. In more rural areas,
location accuracy can be off by more than a mile, which would negate the value of travel times
estimated in this manner. To obtain cellular travel times for reliability monitoring, agencies
would either have to partner with cell phone companies or buy data from a third-party provider.
12
PRIVATE SECTOR-BASED SOURCES
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Interview results from this project established that many public transportation agencies
are interested in obtaining data from private sources, in order to save time and money on data
collection and processing. While these private sources can provide data for facilities that are
otherwise unmonitored (such as arterials), the lack of transparency on their proprietary methods
of data collection could present challenges for agencies seeking to monitor reliability.
The data disseminated by these sources vary. In most cases, the commercial offerings
provide a speed range (i.e. 30 to 40 mph) for stretches of roadway defined by Traffic Message
Channel IDs. These data are, by their very nature, opaque to agencies. For example, it is not clear
from the data where on that stretch of roadway the speeds were observed, or when during the
time period they were observed. More importantly, little information is given on the methods
used to calculate the speeds. For example, the speeds may have been calculated from multiple
GPS probe readings on the roadway, and thus be highly accurate, or they may have been
interpolated entirely from historical data because no real-time samples were collected during the
time period.
27
Transportation Data
28
29
30
31
32
33
34
35
36
37
38
39
40
These private source firms collect data from a variety of sources, including GPS probes,
road sensors, toll tags, and cellphones. Many of these firms also collect incident and event data.
Typically, the business model of these firms revolves largely around the traveler information
market, and they sell this data to public agencies as a sideline to their core business. As such, the
data itself largely reflects the needs of traveler information systems and not performance
measurement.
The simplest data these firms collect are fixed roadway sensors. These are largely the
result of a series of public-private partnerships, stretching back to the mid-1990s, in which firms
were allowed to install and maintain fixed detectors on public roadways, usually in exchange for
an exclusive concession to sell the traffic data to another market, such as the local media market.
Typically, these data are available already to the public agency, as part of the concession. In
some cases, the agency might procure this data, or additional rights to data they already receive
(as part of a new travel time reliability system, for example).
2-9
1
2
3
4
5
6
7
8
9
10
Increasingly, these firms are collecting probe data. This probe data has historically been
the purview of freight companies, who were the only firms that had the necessary cost incentives
to equip their vehicles with GPS. For example, freight companies can rent or purchase tracking
devices to place on vehicles, and then pay a flat communication fee to receive web access and
real-time alerts on vehicle locations. Thus, the first data sources for private providers were
primarily freight carriers. However, in a world of cheaper GPS and ubiquitous smart phones, this
is rapidly changing. Currently, an estimated 51% of drivers have smart phones, many of whom
use the devices’ GPS capabilities in-vehicle for navigation assistance (14). Firms are
increasingly acquiring data directly from consumers as part of the growing personal navigation
market. As a consequence, the size and diversity of the probe data sets are exploding.
11
Data Transparency
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
While some providers may supply metadata on the data quality (for example, a ranking
scale), the methods for the quality assessment remain opaque. For the most part, these limitations
are inherent to the business model of the data provider. Generally speaking, these private source
data providers have built their competitive advantage on their network of data sources and data
fusion methodologies. Because of this, they are unlikely to reveal the underlying sources and
methodologies to the agencies that may wish to know.
This fact must be considered by agencies interested in using private source data to
produce or supplement reliability information. Generally speaking, an agency must ask itself two
questions:
1.
Can we perform the necessary (travel time reliability) calculations with this form
of data?
2.
Do we believe that the data are sufficiently accurate?
If the answer to both of these questions is yes, then this “black box” approach to data
acquisition may be entirely appropriate for a given agency. Both of these questions can be
answered, but they generally must be answered on an application-by-application basis. For
example, to answer the second question, an agency could purchase a data sample from a firm and
independently test its accuracy. To answer the first, an agency might attempt to build travel time
distributions out of data binned by speed to examine if these simplified distributions were
adequate to its needs.
31
DATA MANAGEMENT FOR RECURRING CONDITIONS
32
33
34
35
36
37
38
39
40
41
42
43
Travel time reliability is often monitored through a combination of real-time and off-line
systems. Off-line, regimes are developed that characterize the manner in which segments and
routes operate. Then, in real-time, tests are performed to determine what regime is operative for a
segment or route at any given point in time and what regime(s) are likely to be operative in the
short-term future. The regimes are described by probability density functions that portray the
distribution of travel times likely to be experienced by travelers.
The best database design for this type of situation is a data warehouse model, which
means that data are stored in the database in a non-normalized format, at multiple levels of
aggregation and imputation. This ensures that the system can support report generation at several
levels of temporal and geographic granularity.
A data warehouse is a type of relational database, which is composed of two-dimensional
storage elements (tables). Each table has fields (columns) and records (rows). Tables relate to
2-10
1
2
3
4
5
6
one another through common fields. An example relationship between tables in a travel time
database is shown in Table 2-1. The top table contains background information on routes in the
system, and the bottom table contains 5-minute travel times. The two tables relate to one another
through the Route ID field.
Table 2-1: Example Database Table Relationship
Route ID
Route Name
Length (miles)
009
Golden Gate Bridge to Bay Bridge
6
020
Walnut Creek to Oakland Coliseum
15
116
Downtown Berkeley to Rockridge
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Time ID
Route ID
Travel Time (mins)
Operative Regime
11/25/2009 17:00:00
009
25
1
11/25/2009 17:00:00
020
32
3
11/25/2009 17:00:00
116
15
2
Five general types of tables are commonly used in reliability monitoring systems:
1.
Configuration information. This is information that is not in real-time and
does not change in the field. Examples of configuration information
include the definitions of the routes for which reliability will be monitored
and the locations of detectors and sensors on freeways and arterials.
2.
Raw data. These data are directly received by devices in the field and not
processed in any significant way. The form of the raw data is dependent
on the sensor technology; for example, data from loop detectors contains
different fields than data from a GPS device. As a result, different raw
data tables are needed for each of the four data sources: (1) and (2)
infrastructure-based; (3) AVI-based; and (4) AVL-based. Examples of
these are included in Supplement A as discussed below.
3.
Travel time information by sensor type. Since the computational process
for deriving travel time information from the raw data is dependent on the
technology used in data collection, different database tables are needed to
store the travel time information derived from the different source types.
Where routes are being monitored by multiple technologies, data from the
different sources can be fused to improve the travel time information. In
these cases, a final database table is needed to store the fused travel time
data, which ultimately are used to update the travel time density functions
and regimes and facilitate the reliability computations.
4.
Travel time density functions. These density functions are derived off-line
from the travel times kept in the data tables described above. They are
commonly separated by segment and route regimes that have been
identified by the travel time reliability monitoring methodology described
in Chapter 3.
2-11
1
2
3
4
5
6
7
8
9
10
11
Database design usually takes the form of a schema, which formally describes the
database structure, including the tables, their relationships, and constraints on data value types
and lengths. A companion document to this guidebook, Supplement A: Monitoring System
Architecture, presents example tables that can store information generated during all steps of the
reliability monitoring computation process, from the raw data to the travel time density functions
and reliability metrics. The exact tables, fields, and relationships are flexible to the needs of the
agency, the data available, and the desired reporting capabilities.
12
DATA MANAGEMENT FOR NON-RECURRING EVENTS
13
14
15
16
17
18
Additional information beyond the already-discussed traffic data is required to
understand the impact of non-recurring events on travel time reliability. The primary nonrecurring events that affect reliability are incidents, weather, construction, and special events.
The ability of agencies to collect data on these events, and the types of data they can collect, will
vary between locations. As such, this section is meant to be flexible, so that agencies can tailor
recommendations to fit the data that they have at their disposal.
19
Data Characterization
20
5.
Reliability summaries. These tables contain information that portrays the
manner in which segments, routes, and special areas have been operating
in specific spatial and temporal timeframes.
This section presents four of the major sources for non-recurrent congestion.
21
Transportation Incidents
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
Traffic incidents can reduce available roadway capacity and include accidents, vehicle
disablements, debris on roadway, and vehicle abandonments. Higher frequency of incidents on
specific roadway segments is typically associated with greater travel time variability. There are
many viable sources for collecting incident data. Most state (and some local) emergency
response agencies use Computer Aided Dispatch (CAD) systems to respond to incidents; these
systems have feeds to connect with. The benefit of this data source is that it is in real-time, but
the drawback is that the data has not been cleaned (for example, locations may not be clearly
specified and durations may be inaccurate). Many State Departments of Transportation have
databases with cleaned up incident records for state highways (for example, the Caltrans
Accident Surveillance and Analysis System), for the purpose of performing detailed analyses.
These sources can also be leveraged for reliability monitoring. Another potential source for
incident data is the local Transportation Management Center (TMC), where operators usually
enter incident information into their management software. Finally, private sources such as
Traffic.com often collect incident data at a high level of specificity from various sources,
including video, mobile (patrol) units, and emergency communication frequencies. While many
potential sources for incident data exist, it should be noted that these data are often incomplete,
many times lacking severity indicators, clearance times, and exact incident locations.
The following variables can be used to relate traffic incidents with travel time variability:
Location
Date
2-12
1
2
3
4
5
6
7
8
Type
Starting time and duration
Full time to clearance
Severity
Lanes impacted
In addition, transit incidents, such as bus collisions or disablements can disrupt the
operations of a transit system and cause major delays. Such incidents are increasingly being
detected by the AVL systems used by transit agencies.
9
Weather
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Severe weather events include rain, snow, extreme temperatures, strong winds, blowing
dust, and wildfires. Adverse weather typically results in lower vehicular speeds and higher
incident frequencies, increasing travel time variability. One source for weather data is existing
weather stations operated by various governmental organizations or research bodies. For
example, the most accurate sources of weather information are the Automated Surface Observing
System (ASOS) and Automated Weather Observing System (AWOS) stations maintained and
used for real-time airport operations by the Federal Aviation Administration (FAA). Another
good source is an online interface from the National Climatic Data Center (NCDC) of the
National Oceanic and Atmospheric Administration (NOAA), which provides hourly, daily, and
monthly weather summaries for 1,600 U.S. locations. For mountainous rural areas, the major
sources of weather-related delay are closures and chain control stations. These data are
frequently available from rural traffic management centers, although collecting feeds of such
data is rare and problematic. One of the richer sources of this data may be Highway Advisory
Radio (HAR) networks, which broadcast closure and chain control locations and are frequently
available via statewide feed. Any weather data obtained from sources not directly on a monitored
route will have to be associated with nearby routes in the system.
Another option for collecting weather data is to install Environmental Sensor Stations
(ESS) at key roadway locations. Many states have used these to build Road Weather Information
Systems, which archive weather data and use it in roadway-related decision making.
The following variables can be used to relate weather with travel time variability:
Air temperature
Type of precipitation
Amount of precipitation
Visibility
Wind speed
Pavement temperature
Surface condition
Transit agencies can use similar methods to monitor weather conditions and develop
operational plans that can help them deal with potential disruptions in service and variability in
travel times during a variety of adverse weather events.
As of this writing, the process for connecting weather data to travel time monitoring
systems is largely manual. Examples from the San Diego and Atlanta case studies are presented
in Chapter 4.
2-13
1
Work Zones
2
3
4
5
6
7
8
9
10
11
12
13
14
Work zones typically reduce the capacity of roadways, increase driver distractions,
increase incident frequency, and cause major delays in the transportation network. They are
typically associated with higher travel time variability. There are a few different sources for
construction-related lane closures. Many states have lane closure systems that serve as a
communication interface between the contractors and state agencies to facilitate lane closure
management; this data source can be obtained in real-time. Private sources are another option;
for example, Traffic.com reports both scheduled and unscheduled construction events. Another
option is to manually obtain construction-related information from Changeable Message Sign
logs or feeds.
The following variables can be used to relate work zones with travel time variability:
Start time and duration
Start and end locations
Lanes impacted
15
Special Events
16
17
18
19
20
21
22
23
24
25
26
Frequently, special event data must be collected by individuals rather than systems, since
there are no standardized, centralized sources for this information. One option is to manually
review calendars for major event venues near a route. Another option is to obtain event data from
TMCs, many of which collect event logs to know when and where to activate event-based signal
timing plans.
The following variables can be used to relate special events with travel time variability:
Location
Routes affected
Duration
Type of event
Attendance
27
Data Storage
28
29
30
31
32
33
34
35
36
37
38
39
40
41
The data storage regime for the non-recurring events is dependent on exactly which
variables are collected, and at what granularity. The spatial and temporal resolution of nonrecurring events data is an important consideration that impacts the strength of the relationships
developed with travel time variability. Data on non-recurring events, to some degree, needs to be
aggregated to the same temporal and spatial resolution, in that it all needs to be spatially
collected by route and temporally collected for each day in the analysis period. Collecting data
on some of the sources at higher spatial and temporal resolutions would lead to more accurate
analysis.
The case studies and use case analyses, described in Chapter 4, suggest that a granularity
of five-minutes is sufficient to link the effects of non-recurring events like weather to their travel
time impacts. Granularities greater than 15 minutes are too coarse. Finer granularities, such as
one minute or 30 seconds, are helpful but not necessary. The data on non-recurrent events does
not need to be stored in the same tables as the route travel times, since the analysis to link travel
time variability with its causes is typically a manual exercise. As such, the database for non-
2-14
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
recurring events can be uniquely designed to store the data that each agency is able to collect.
Examples of typical data tables are provided in Supplement A: Monitoring System Architecture.
The most useful way to store the event data is in separate files tailored to the event type
where beginning and ending times can be recorded along with descriptive information. This is
because these events have no specific synchronization with the traffic sampling intervals and
should not be forced to be synchronous. Moreover, the events can have temporal and spatial
impacts outside the bounds of their duration and physical location. For example, in the case of
the Lake Tahoe analyses, the heavy snowstorms had impacts extending well beyond the time
when the snow stopped falling. Also, in Sacramento, accidents on intersecting freeways, near the
interchanges with the subject facility, had impacts on the travel times observed.
An important corollary to this is that the queries seeking explanatory information for high
travel rates or times at a specific location and time must look outside the bounds of that location
and time. Although the weather data may indicate that the weather event is over, the query must
seek causal explanations from earlier weather events. Similarly, although there may be no event
at the location being examined, the query must look for events on intersecting facilities. (It must
also look downstream of the location being examined.)
Another significant issue is that event data seem to lack quality information about
duration. Moreover, the end of the interval during which the incident has an impact on travel
times tends to be larger than the duration of the incident event itself, e.g., the increased travel
times persist for some time after the incident has been recorded as “over”.
A final note is that AVI and AVL data show travel time (and rate) transients with a
granularity dictated by the frequency with which such observations are obtained. The case
studies and use cases tend to suggest that with today’s penetration rate for such devices, keeping
all the raw data makes sense (sanitized to protect privacy). Summarizing it based on some
granularity (e.g., 5 minutes) has no value and can obscure very useful information about travel
time/rate transients.
27
SUMMARY
28
29
30
31
32
33
34
35
36
37
38
39
40
The final database design for a reliability monitoring system must reflect the technologies
used to collect data and the processes used to derive travel time estimates from the raw data. In
general, four types of tables are needed to fully describe the monitoring system and its outputs.
Configuration tables are needed to define the routes and route segments for which travel times
are to be computed, including their starting and ending point, where their detectors are located,
and what type of detection they have. Raw data tables are needed to store the unaltered inputs
from the various detection types. Travel time tables are needed to store the representative travel
times calculated for each route segment, time period, and detection type, as well as the
intermediary information generated during the computation. Finally, reliability tables can be
used to store high-level monthly, quarterly, and yearly summaries of reliability statistics for
individual routes or higher spatial aggregations. One-way (irreversible) encryption can be
applied to protect the privacy of personally-identifiable information such as license plate
numbers.
41
REFERENCES
42
43
1. California Center for Innovative Transportation. Evaluation of Wireless Traffic Sensors
by Sensys Networks, Inc. University of California, Berkeley, October 2006.
2-15
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
2. Klein, L., M. Mills, D. Gibson. Traffic Detector Handbook: Third Edition—Volume 1.
Report No. FHWA-HRT-06-108. Turner-Fairbank Highway Research Center, Federal
Highway Adminstration, October 2006.
3. Venere, Emil. Method uses Bluetooth to track travel time for vehicles, pedestrians.
Purdue University, May 27, 2008.
4. Traffax, Inc. http://www.traffaxinc.com. Accessed October 16, 2009.
5. Texas Transportation Institute. Travel Time Data Collection Handbook. Report FHWAPL-98-035. Federal Highway Administration, 1998.
6. PBS&J. License Plate Reader Operations. Florida Department of Transportation, 2006.
7. Dahlgren, Joy and John Wright. Using Vehicles Equipped with Toll Tags as Probes for
Providing Travel Times. California Partners for Advanced Transit and Highways, 2001.
8. Haas, R., M. Carter, E. Perry, J. Trombly, E. Bedsole, and R. Margiotta. iFlorida Model
Deployment Final Evaluation Report. FHWA Report No FHWA-HOP-08-050. Federal
Highway Administration, January 2009.
9. Riley, J. D. Evaluation of Travel Time Estimates Derived from Automatic Vehicle
Identification Tags in San Antonio, TX. MS thesis. Virginia Polytechnic Institute and
State University, Blacksburg, 1999.
10. Kavaler, R., K. Kwong, R. Rajagopal, and P. Varaiya. “Arterial Travel Time Estimation
Based on Vehicle Re-Identification Using Wireless Magnetic Sensors.” Transportation
Research Part C: Emerging Technologies. Volume 17, Issue 6, December 2009, Pages
586–606.
11. Berkow, M., J. Chee, R. Bertini, and C. Monsere. “Transit Performance Measurement
and Arterial Travel Time Estimation Using Archived AVL Data.” Presented at ITE
District 6 Annual Meeting, Portland, Oregon, July 2007.
12. Research and Innovative Technology Administration. “Achieving the Vision: From VII
to Intellidrive.” http://www.its.dot.gov/press/2010/vii2intellidrive.htm. Accessed
September 2, 2012.
13. University of California at Berkeley. “New research project captures traffic data using
GPS-enabled cell phones.” February 2008.
http://www.physorg.com/news121845452.html. ccessed October 19, 2009.
14. J.D. Power and Associates. Smartphone Ownership Drives Increased Vehicle Owner
Interest in Communication- and Connectivity-Related Features. Press release.
http://businesscenter.jdpower.com/news/pressrelease.aspx?ID=2010095. Accessed
September 2, 2012.
15. Wunnava, S., K. Yen, T. Babij, R. Zavaleta, R. Romero, and C. Archilla. Travel Time
Estimation Using Cell Phones (TTECP) for Highways and Roadways. Florida
Department of Transportation, 2007.
2-16
1
CHAPTER 3
2
COMPUTATIONAL METHODS
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
The computational methods used by travel time reliability monitoring systems (TTRMS)
need to be designed to generate information about the ways in which travel times and rates vary
during system operation and the reasons why those variations occur. Ultimately, system
managers want to use the TTRMS to improve the reliability of the system and to convey its
current and projected performance to the people who use the system.
This guidebook presents ways to create probability density functions for the travel times
and travel rates (TT-PDFs and TR-PDFs) experienced by network segments and routes and tools
that show how these PDFs are affected by influencing factors, both internal and external to the
system.
The TTRMS can be thought of as comprising three parts. The first is a travel time
processing unit: it takes the raw data outlined in Chapter 2, cleans it, fills in gaps, estimates
errors, and produces the travel times upon which the PDFs are based. The second is a data
processing unit: it takes these travel times and creates the PDFs that characterize the performance
of the system under various operating conditions. The third is an analysis tool: it sees how the
various influencing factors, both internal and external, affect the reliability performance of the
system.
In this chapter, seven subsections are used to describe the TTRMS:
Network Concepts: how the TTRMS represents travel times.
Trip-Making Concepts: how the TTRMS represents trip travel times.
Operating Conditions and Regimes: how the impacts of influencing factors are
studied.
Imputation: how the TTRMS should impute estimates for missing or invalid data.
Segment Travel Time Calculations: the steps and computations that transform raw
sensor data into observations of segment travel times.
Route Travel Time Computations: how travel times are assembled into probability
density functions for segments and routes.
Causal Factor Analysis: how the TTRMS can be used to examine the influence
on reliability of various causal factors, both internal and external.
As discussed in Chapter 1, both operators (the supply side) and travelers (on the demand
side) need to be kept in mind in describing the TTRMS. Operators and managers want to
minimize variability in the travel times so that network performance is consistent. They want to
find ways to reduce this variability when it is unreasonably high. Travelers want to make trips
that avoid highly variable travel times so they can get to destinations on time. They want to make
astute route and departure time decisions.
In all instances, the fundamental unit of observation is travel times and rates for
individual vehicles (or people or packages) making trips across the network. These are the
entities that are moving. This is the raw data upon which the TTRMS is based. Either the
TTRMS receives this raw data directly from AVI and AVL-equipped vehicles, or it receives it
implicitly through average speeds and travel times provided by single and double loop detectors
or third-party data sources (e.g., TMC segment travel times).
3-1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Some of this raw data is eventually archived for future use in developing the PDFs for
analyses of the past. In the case studies presented in Chapter 4, all the data have been kept, even
the individual AVI and AVL vehicle records, to support methodological development across the
project. But for operating agencies, this is unlikely to be a suitable practice, especially for the
AVI and AVL data because of the large amounts of storage required (and privacy issues). It will
be sufficient to keep PDFs that describe the system performance under various conditions along
with samplings of the AVI and AVL data (scrubbed so identities are removed) and the system
detector data (as is the current practice). The sufficiency test is this: are the data that are being
kept sufficient to develop any and all PDFs that are likely to be of interest in comparing past
performance with the current status quo.
For both the present and the past, the raw data are summarized into PDFs that look at
trends across time for the same time period (e.g., the same 5-minute periods on weekdays across
the year) or across vehicles at the same point in time (using cross-sectional data) to look at
variations in the trip times or rates for travelers who are all making the same trip at the same
time, or traversing the same parts of the network at the same point in time.
The longitudinal PDFs are often used by system operators to understand how travel time
variability is affected by internal and external influencing factors such as congestion (internal)
and weather, incidents, and special events (external). The cross-sectional PDFs are used by
travelers to understand how long it will take them to make a trip at a particular point in time and
the probability that they will arrive on time.
21
NETWORK CONCEPTS
22
23
24
The first task in establishing a TTRMS is to define a network topology that can be used
to study the system’s travel time variability. Notions of monuments, segments and routes need to
be created.
25
Monuments
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
The first idea is a monument. Similar in concept to monuments used by land surveyors, it
is a reference point (node) to and from which the travel times are measured. As illustrated in
Figure 3-1, while monuments could be placed at the ends of the physical links (i.e., in the middle
of intersections or interchanges), the findings from this study suggest this is not wise; the travel
times become ambiguous because turning movements confound the observations. Placing the
monuments at the midpoints of the physical links removes this confusion. It ensures that the
correct turning movement delay is included in each monument-to-monument travel time. This is
clearly important for arterials but it is also important for freeways. Ramp movements can have
different travel times (e.g., direct ramp or loop ramp, as well as and any traffic control on the
ramp—such as a signal—as is sometimes the case in Los Angeles).
The monuments should be—and need to be—linked to locations that the traffic
management center uses to monitor the system, as in the location of system detectors on both the
freeway and arterial networks. This minimizes the database management tasks involved in
keeping track of where the monuments are located. They can also be the location of toll tag
readers and AVI sensors. They should not be placed at locations where standing queues occur.
3-2
1
2
3
4
5
Figure 3-1: Illustration of Monuments
Passage Times
6
7
8
9
10
11
12
13
14
15
16
The second idea is a passage time. It indicates when a vehicle has passed a given
monument. When the monument is at the location of a detector (e.g., a loop in the pavement or
an AVI reader), the passage time is the instant at which the vehicle crosses the detector. In the
case of an AVI reader used for collecting tolls, it is the time when the toll is collected. In the case
of other AVI sensors, where the tag may be identified more than one time when the vehicle is
passing the sensor, it should be the time when the vehicle is closest to the sensor. For Bluetooth
sensors, the response with the greatest signal strength should be used, or an interpolation of the
response pulses to identify this location, or the mean of the strongest responses. For AVL
vehicles, as is the case for some AVL applications, it should be the time when the vehicle passes
the monument location based on comparing the latitude and longitude of the vehicle with the
latitude and longitude of the monument.
17
Segments
18
19
20
21
22
23
24
25
26
27
28
A segment is a path between two monuments. In some cases there will be only one
segment between the monuments. In other cases there may be two or more. It is these segments
for which the travel times ought to be monitored and estimated. The segments can be, and should
be, related to the physical network. A good option is to map them to demarcations of the network
used for other purposes (e.g., TMC segments, which are explained later in this chapter). It is the
segments for which PDFs are estimated.
Figure 3-2 shows examples of segments that tie the midpoint of a north-south link south
of an intersection to the midpoint of an east-west link to the east of the intersection. The
northbound-to-eastbound segment connecting the monuments on these two links represents the
northbound right turn at the intersection. The segment in the other direction represents the
westbound left.
3-3
1
2
3
4
Figure 3-2: Illustration of Segments
5
Routes
6
7
8
9
Routes are sequences of segments. Routes are defined by the sequence of segments, and
thus monuments, that they pass through; routes with the same origin and destination are unique
unless they go through the same segments in the same order. Figure 3-3 shows an example of
two routes.
10
TRIP-MAKING CONCEPTS
11
12
13
14
15
16
For system operators, monitoring the performance of segments and routes is likely to be
sufficient unless reliability goals and objectives are established for specific sets of users (i.e.,
travelers or shippers). For travelers and shippers, however, additional trip-making concepts are
needed to appropriately analyze the raw data. This includes: the travelers that are of interest, the
geographic definitions of the origins and destinations, the routes and segments involved, and the
conditions under which the trips were made.
3-4
1
Users
2
3
4
5
6
7
8
9
In this context, users are people (or packages) making trips across the network. (They
may also be users of the TTRMS, but that is a different thought.) Users can be organized into
groups with the same or similar trip-making behavior. An example would be commuters making
weekday trips from zone A to zone B. Another would be vehicles traveling from a residential
zone the airport. The analyst needs to determine how “similar” these users need to be for them to
be classified as a single group, or how heterogeneous they can be without undermining the
usefulness of the travel time reliability information.
10
11
12
Figure 3-3: Illustration of Routes
13
Route Bundles
14
15
16
17
18
19
20
21
Route bundles are sets of two or more routes. The routes in a bundle can have the same
monument pair as the start and end points (a set of different paths for the same point-to-point
pair) or they can be for slightly different monument pairs, as might be the case for all the routes
leading from the monuments in one zone (e.g., a ZIP code or traffic analysis zone) to all the
monuments in another zone. The majority of the intervening segment sequences might be largely
the same (one general route) or varied (several general routes exist). Hence, the bundles arise
because a) there are many routes or b) there are many originating or terminating monuments, or
c) both. These are illustrated in Figure 3-4.
3-5
1
2
3
Figure 3-4: Illustration of Route Bundles
4
Markets
5
6
7
A market is formed by a set of users in combination with a route bundle. An example
would be commuters who have toll tag equipped vehicles (user group) that travel in the morning
(time period) from a suburban area to downtown destination (using a specific route bundle).
8
OPERATING CONDITIONS AND REGIMES
9
10
11
12
13
14
Operating condition concepts form the basis for describing the conditions under which
the segment, route, or network is operating at a given point in time. They also serve as the basis
by which the operation at a given point in time can be classified. Ultimately, it is for these
conditions that the PDFs are developed and via these conditions that the operation is understood
(i.e.., its variability) and for which actions are taken to improve reliability.
15
Congestion Conditions
16
17
18
19
20
21
22
23
24
25
It is useful to categorize 5-minute time periods on the basis of the level of congestion
(demand-to-capacity ratio) that would normally be occurring at that point in time. For example,
7:55–8:00 on weekdays that are workdays (not holidays) might be classified as “high
congestion” if it were a 5-minute time period that would normally be experiencing high
congestion during the a.m. peak. It is well recognized that the level of congestion affects the
variability of the travel times and a categorization scheme that codifies that affect is helpful.
It is also valuable to use four categories for these congestion conditions: uncongested,
low, moderate, and high. Other analysts might find it helpful to use additional categories. Also, it
is not necessary to apply all of these categories to every segment or route. Uncongested is
defined as meaning that the vehicle interactions were not having a perceptible impact on the
3-6
1
2
3
4
5
6
7
8
distribution of the vehicle travel times. Travelers were able to use their desired speeds. Low
means there was some evidence of increased variability in the travel times (more than would
arise under uncongested conditions) so the vehicle interactions had to be interfering with the
ability of travelers to move at their desired speeds, but not so much as to create queues.
Moderate means that the demand-to-capacity ratio was sufficiently high to create short-term
queuing, as often occurs in queuing systems where the queues form and dissipate quickly. High
meant that there were sustained queues caused by demand in excess of capacity that were only
cleared when the demand-to-capacity ratio again diminished and the system was able to catch up.
9
Non-Recurring Events
10
11
12
13
14
15
16
17
18
19
The second dimension for categorizing system behavior is based on the non-recurring
events that occurred, including “none” or “normal operations”. These designations were intended
to be consistent with the FHWA ideas about sources of congestion. The authors were able to
identify the impacts of incidents, weather, special events, traffic control devices, and unusually
high demand. The impacts of “insufficient base capacity” were captured by the nominal
congestion condition categories described above (i.e., situations where the demand-to-capacity
ratio is high enough that sustained queuing occurs). The high demand category is equivalent to
“fluctuations in demand”. Work zones were not present in any of the networks for which
analyses were conducted, but it is clear from the congestion-related analyses that they would
have an impact on reliability.
20
Regimes
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Regimes are the resulting categories of conditions under which the segment, route, or
network is operating at a given point in time (or from one time to another). It is effectively the
“loading condition” on the system. It is helpful to define these regimes based on combinations of
the congestion condition that would normally be occurring, reflecting the demand-to-capacity
condition that would nominally arise at that point in time , as well as the non-recurring event (if
any) that was occurring at that point in time.
Table 3-1 illustrates how these regimes were used. It shows the amount of variability
experienced during a year can be categorized based on these regimes for three routes in San
Diego that were examined in one of the use cases. The particulars are not important here, only
the illustration of the regimes. That having been said, notice that for the I-5 route, that 50% of
the extra time attributable to variable travel times is attributable to periods of high congestion
when no non-recurring event has arisen. High congestion under periods of higher-than-normal
demand accounts for another 13% and non-recurring weather and incident events account for
another 13% and 16% respectively. On the basis of these regimes, analysts can identify times
and conditions for which mitigating actions would be appropriate (including adding more
capacity or altering the use of traffic control devices) to reduce the amount of extra travel time
created by the variability in the travel times.
3-7
1
2
Table 3-1: Illustration of Regimes
Route
Condition
I-5
Uncongested
High
Uncongested
Low
Moderate
High
Uncongested
Moderate
High
CA-15
CA-163
Normal
(%)
8
50
4
4
7
35
4
12
35
Demand
(%)
1
13
0
0
0
17
0
1
19
Weather
(%)
1
8
0
0
0
6
0
1
5
Special Events
(%)
0
3
0
0
0
6
0
2
4
Incidents
(%)
1
16
0
0
0
19
0
3
14
Total
(%)
11
89
4
5
8
83
4
19
77
Facility Total
(%)
100
100
100
3
4
5
6
7
8
9
10
11
12
13
14
Readers should recognize that the right set of regimes for a given segment, route, or
system will be site-specific. Care should be taken in identifying these regimes so that they
represent the smallest set of categories that clearly describe how the system operates and
performs. The incentive for using fewer (rather than more) is so that the data are categorized into
groups of sufficient size that meaningful PDFs can be developed that describe the system’s
operation under that condition (regime). It is also important that these regimes be the system
states that are most critical to study, manage, and improve from the operating agency’s
perspective. The ones that are important are likely to evolve over time as the agencies fix
problems and develop more insight into travel time reliability. Akin to the development of signal
timing plans, the system manager will determine over time what conditions are the most
important ones to monitor.
15
Sample Space
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
The sample space (observation set, observation timeframe, sample frame) is the set of
raw data that pertain to each context for which a PDF is being developed (a regime or some other
basis). A simple example would be the observations that pertain to the high congestion / normal
regime. Another, which is not a regime, is observations of travel times (or rates) between two
points in time for some non-recurring condition (e.g., between 7:00 a.m. and 9:00 a.m. on
weekdays that are workdays when it was snowing). This latter sample frame is a very logical
one; it is just not necessarily a regime. The timeframe of 7:00 a.m. to 9:00 a.m. is likely to be one
across which the nominal level of congestion is varying considerably, and the fact that the nonrecurring condition is “snowing” is a specific instance of a weather condition. It is certainly
possible to develop PDFs for this condition, to study them and see how they indicate system
performance (variability in the travel times) is different for this situation than it is for others
(e.g., high congestion during “no” non-recurring event), and the main point is that the sample
space would be observations of travel times (or rates) that occurred during this time frame when
it was snowing. The main point is that it is always important to understand what data were used
to conduct the analysis so that the findings are correctly understood and interpreted in a manner
that is consistent with the data used in the analysis.
33
IMPUTATION
34
35
For a TTRMS to function effectively, it needs complete information about the travel
times and rates for all segments in the network. Missing data makes it impossible to both analyze
3-8
1
2
current (and past performance) and to develop projections of future conditions. This section
provides guidance about how best to deal with situations where field data are missing.
3
Imputation for Infrastructure-Based Sensors
4
5
6
7
8
9
10
11
12
13
The most prevalent type of roadway data currently comes from point sensors, such as
loop and radar detectors. While these data sources provide valuable information about roadway
performance, they have two main drawbacks: (1) detectors frequently malfunction; and (2) they
only provide data in the temporal dimension (speeds at a point, or their inverse, spot rates),
whereas travel times require data over spatial segments. The first issue is best dealt with through
an imputation process, which fills in data holes left by bad detectors. This is the issue addressed
in this section.
Rigorous imputation is a two-step process. First, the system needs to systematically
determine which detectors are malfunctioning. Then, the data from the broken detectors must be
discarded and the remaining holes filled in with imputed values.
14
The Imputation Concept
15
16
17
18
19
20
21
22
23
24
25
Once the system has classified infrastructure-based detectors as good or bad, it must
remove the data transmitted by the bad detectors so that invalid travel times are not incorporated
into the reliability computations. However, removing bad data points will leave routes and
facilities with incomplete data sets for time periods where one or more detectors were
malfunctioning. Since travel time computations for infrastructure-based sensors rely on the
aggregation and extrapolation of values from every detector on the route, travel time
computations for a route would be compromised each time a detector failed (which, given the
nature of detectors, will occur frequently). For this reason, imputation is required to fill in the
data holes and thus enable the calculation of travel times and variability. Figure 3-5 shows the
high-level concept of imputation.
26
27
28
29
30
31
Figure 3-5: Imputation of Traffic Data
From a data management perspective, agencies are advised to store and archive both the
raw data and the imputed data in parallel, rather than discarding the raw data sent by detectors
diagnosed as bad. This ensures that, when imputation methods change, the data can be re-
3-9
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
imputed where necessary. It also gives system users the ability to skeptically examine the results
of imputation where desired.
When imputation is needed for infrastructure-based data, it must begin at the detector
level. The strength of an imputation regime is determined by the quality and quantity of data
available for imputation use. The best imputations are derived from local data, such as from
neighboring sensors or historical patterns observed when the detector was working. Modest
imputations can also be performed in areas that have good global information, by forming
general regional relationships between detectors in different lanes and configurations. It is
important to have an imputation regime that makes it possible to impute data for any broken
detector, regardless of its characteristics. As such, the best practice is to implement a series of
imputation methods, employed in the order of the accuracy of their estimates. In other words,
when the best imputation method is not possible for a certain detector and time period, the nextbest approach can be attempted. That way, data that can be imputed by the most accurate method
will be, but imputation will also be possible for detectors that cannot support the best or most
rigorous methods.
The following imputation methods are recommended to be employed where possible, in
this order, to fill in data holes for bad detectors (1). The first two methods make use of real-time
data, while the third and fourth methods rely on historically archived values.
19
Linear Regression from Neighbors Based on Local Coefficients
20
21
22
23
24
25
26
27
28
29
30
31
This imputation method is the most accurate due to the fact that there are high
correlations between occupancies and volumes of detectors in adjacent locations. Infrastructurebased detectors can be considered neighbors if they are in the same location in different lanes or
if they are in adjacent locations upstream or downstream to the bad detector. In this approach, an
offline regression analysis is used to continuously determine the relationship between each pair
of neighbors in the system. The dependent variable is the flow or occupancy at a detector (when
the detector was good) and the independent variables are the flow or occupancy at adjacent
detectors. Then, when a detector is broken, its flow and occupancy values can be determined by
using the estimated regression parameters. The regression equations can take the form given in
Equations 3-1 and 3-2 as follows:
32
33
34
35
36
37
38
39
40
41
42
q i (t ) 0 (i , j ) 1 (i , j ) q j (t )
Equation 3-1
ki (t ) 0 (i, j ) 1 (i, j ) k j (t )
Equation 3-2
where:
(i,j) is a pair of detectors,
q is flow,
k is occupancy,
t is a specified time period (for example, 5 minutes), and
0 ,1 , 0 , 1 are parameters estimated between each pair of loops using five days of
historical data. These parameters can be determined for any pair of loops that report data
to a historical database.
3-10
1
Linear Regression from Neighbors Based on Global Coefficients
2
3
4
5
6
7
8
9
This imputation method is similar to the first, but is needed because there are some loops
that never report data because they are always broken. This makes it impossible to compute local
regression coefficients for them. For these loops, it is possible to generate a set of global
parameters that expresses the relationship between pairs of loops in different configurations. The
configurations refer to location on the freeway and lane number. The linear model is developed
for each possible combination of configurations and takes the following form:
10
11
12
13
14
15
16
17
18
19
20
21
22
23
q i (t ) 0* ( , l i , l j ) 1* ( , l i , l j ) q j (t )
Equation 3-3
k i (t ) 0* ( , li , l j ) 1* ( , li , l j ) k j (t )
Equation 3-4
where:
(i,j) is a pair of detectors;
δ = 0 if i and j are in the same location, 1 otherwise;
li = lane number of detector i; and
lj = lane number of detector j.
The terms are derived using locally available data for each of the different
configurations. Figure 3-6 shows an example of this type of imputation routine.
Figure 3-6: Imputation Routine Example
24
Imputing Based on Temporal Medians
25
26
The above two linear regression methods cannot be applied during time periods when
adjacent detectors are not reporting data themselves (for example, when there is a widespread
3-11
1
2
3
4
5
communication outage). In this case, temporal medians can be used as imputed values. Temporal
medians can be determined by examining the data samples transmitted by that detector for the
same day of week and time of day over the previous ten weeks. The imputed value is then taken
as the median of those historical values. Note that if a large number of samples have to be
imputed using temporal medians, it may be difficult to assess travel time variability.
6
Imputing Based on Cluster Medians
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
This is the imputation method of last resort, for those detectors and time periods where
none of the above three methods are can be applied. A cluster is defined as a group of detectors
that have the same macroscopic flow and occupancy patterns over the course of a typical week.
An example cluster is all of the detectors that are monitoring a freeway into the downtown area
of a city. These detectors would all have comparable macroscopic patterns in that the a.m. and
p.m. peak data would be similar across detectors due to commuting traffic. Clusters can be
defined within each region. For each detector in the cluster, the joint vector of aggregate hourly
flow and occupancy values can be assembled and, from this, the median time-of-day and day-ofweek profile for each cluster can be calculated.
Once values have been imputed for all broken detectors, the data set within the system
will be complete, albeit composed of both raw data and imputed values. It is important that the
quantity and quality of imputed data be tracked. That is, the database should store statistics on
what percentage of the data behind a travel time estimate (and thus a travel time reliability
measure) is imputed and what methods were used to impute it. This allows analysts to make
subjective judgments on what percentage of good raw data must underlie a measure in order for
it to be usable in analysis.
23
Imputation for Vehicle-Based Sensors
24
25
26
27
28
29
30
31
32
33
34
35
Fundamentally, vehicle-based sensors are different from point sensors. They provide
observations of travel times across distances—not spot rates—so they are true indications of the
time required to travel from one monument to another and they are observations of individual
vehicle travel times—not averages or aggregates—so they provide a picture of the distribution of
the travel times as well.
The critical quality control issue with vehicle-based sensor data is filtering out
unrepresentative travel times collected from travelers who make a stop or take a different route.
This section describes methods for identifying malfunctioning AVI sensors, and imputing data
where sensors are malfunctioning or an insufficient number of vehicles samples were obtained in
a given time period. The emphasis is principally on generating estimates of the average or
median travel times, as opposed to imputing the entire distribution, but the latter is briefly
addressed as well.
36
Identifying Malfunctioning AVI Sensors
37
38
39
40
41
There are two scenarios for which an AVI sensor might transmit no data: (1) the sensor is
broken; or (2) no vehicle equipped for sampling passed by during the time period. Therefore, it is
important to implement a method for distinguishing between these two cases, so that
malfunctioning sensors can be detected and fixed. A simple test for flagging a sensor as broken
is if it did not report any samples throughout a single day.
3-12
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
It is important to recognize that travel times from AVI sensors are based on differences in
time stamps for vehicles passing two AVI detectors. The system computes travel times for
segments defined by two sensors positioned at the segment's beginning and ending points. For
example, sensors A and B define segment A–B. The number of vehicles measured at each sensor
can be written as nA and nB respectively, and the number of matched vehicles over segment A–B
as nA–B. As a rule:
36
Imputing Travel Times for Missing or Bad Vehicle-Based Sensor Data
37
38
39
40
41
42
43
The fact that vehicles can be traced from one AVI sensor to another provides additional
ways to impute travel times not available for infrastructure-based sensors. If vehicle-based
sensor B is broken, it may no longer be possible to measure travel times from A to B or B to C,
but it is still possible to measure travel times from A to C. Thus, with vehicle-based sensors,
malfunctions impact the spatial granularity of reported travel times, but not their accuracy.
The idea of “super segments” is the best way to impute travel times (and travel time
distributions) for segments whose endpoint detector is malfunctioning. Figure 3-7 illustrates this
nAB min(nA , nB )
Equation 3-5
If either sensor is broken or there is little to no traffic equipped for sampling, then nA 0 or
nB 0 , and therefore nAB 0 .
There are three scenarios for which a vehicle-based sensor pair would report no travel
time data:
Either or both sensor A and B are malfunctioning;
No equipped vehicles passed by A or B during the time period; or
No vehicle was matched between A and B during the time period.
For the calculation of travel times, the consequence of a malfunctioning sensor and few
to no equipped vehicles at the sensor are identical: there are not enough data points to calculate
the travel time summary statistics for the segment during the time period.
Similar to the individual sensor malfunction test, a segment can be considered to be
broken on day d if its sensors do not report any samples throughout the day. This "missing" data
case is straightforward to detect. It is more difficult to handle the case when there are too few
data points to construct meaningful travel time trends. For example, if only one data point per
hour is observed over a 24-hour period, detailed 5-minute or hourly travel time trends cannot be
recovered. Such data sparseness is expected during night when traffic volumes are actually low,
and does not cause problems during these free-flow periods. But such data sparseness is not
acceptable during peak hours in urban areas. As such, the system should employ a test for data
sparseness. A segment A–B can be considered to have sparse data on day d if the total number of
samples throughout the day is less than a specified threshold (for example, 200 samples per day).
This approach looks at an entire day of data to determine where sensors are broken and
data is sparse. More granular approaches are also possible, such as running the above test for
individual hourly periods. However, more granular tests would have to use varying thresholds
based on the time of day and day of the week to fit varying demand patterns.
3-13
1
2
3
idea. If AVI detector B is broken, the super segment A–C provides a way to impute vehicle
travel times for both segments A–B and B–C.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
When all three detectors are working properly, regression equations can be developed
that predict the travel times for A–B and B–C based on the travel time for A–C. Then, when
detector B is malfunctioning, these equations can be used to impute individual vehicle travel
times (or the average or some other percentile value such as the median) based on the travel time
observations between A and C. The same idea applies to the segments B–C and C–D if detector
C is malfunctioning, only there are two super-segments that could be used to impute the missing
values (i.e., super segments A–D and B–D). The super segment that is the best predictor of the
travel times on the subject segment (i.e., which might be either B–C or C–D) should then be used
to impute the missing travel times.
Where infrastructure-based detection is present, one can use the point speeds (spot rates)
from those sensors to adjust and/or cross-check the imputed distribution of travel times. Also,
equations (e.g., regression) can also be developed to use the system detector data directly in
doing this. (Of course, the infrastructure-based point speeds can be used directly to estimate
average travel times for the subject segments.)
A temporal median approach, equivalent to the one described for option 3 for
infrastructure-based imputation, can also be used. A temporal median is the median of the
historical, non-imputed route travel time values reported for that segment for the same day of
week and time of day over the most recent ten weeks. As a general rule of thumb, if more than
50 percent of the data are imputed, the results can have questionable validity.
26
SEGMENT TRAVEL TIME CALCULATIONS
27
28
29
30
31
Ultimately, the TTRMS needs to have segment travel times; and from the segment travel
times, route travel times. Moreover, these travel times are needed both on a longitudinal basis
(for the same points in time each day across months, seasons and years) and on a cross-sectional
basis (for multiple vehicles, or perhaps even multiple segments and routes at the same point in
time). This section describes how to develop the segment travel times.
32
Infrastructure-Based Sensors
33
34
35
Infrastructure-based sensors typically collect observations at a single location in the
network. Examples are single- and double-loop detectors, radar detectors, and video detectors.
The sensors all see individual vehicle inputs, but they report out only sums (volumes) and
Figure 3-7: Super Segment Examples
3-14
1
2
3
averages (e.g., average occupancies and speeds). Perhaps at some time in the future, these
devices will report other statistics for the observed inputs (perhaps even a PDF), but that
capability does not exist today.
4
Estimating Spot (Time-Mean) Speeds (for Single Loop Detectors)
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
In the case of many point sensors, the speeds are directly observed and averages of those
speeds are periodically reported out by the device. But in some cases, the speeds are not directly
observed and must be estimated. This is typically the case for single-loop detectors.
Estimation of the average speed is then done by using a "g-factor", which uses an
assumed average length of vehicle to estimate the average speed value from the flow and
occupancy data using the below equation:
flow
1
speed
Equation 3-6
occupancy g
Here, (1/g) is the average effective vehicle length, which is equal to the sum of the actual
vehicle length and the width of the loop detector. At a given loop, the average vehicle length will
vary by time of day and day of the week due to truck traffic patterns. For example, trucks make
up a larger percentage of traffic in the early morning or midday hours in urban regions, while
automobile traffic dominates during commute hours. Vehicle length is also dependent on the lane
of the loop; loops in the right-most lanes will typically have longer lengths than those in the lefthand lanes since trucks usually travel in the right lanes. Despite this variability, most agencies
use a constant g-factor that gets applied across all loops and all time periods in a region. Analysis
of this approach has shown that it can introduce errors of up to 50% into speed estimates (2).
Since average travel time computations require the aggregation and extrapolation of multiple
loop detector speed estimates, this has ramifications for the reliability of calculated average
travel times. As such, agencies are advised to implement methods that account for the time- and
location-dependency of the g-factor.
The g-factors can be estimated for each time of day and individual detector using a
weeks-worth of detector data. By manipulating the above equation to make speed the known
variable and the g-factor the unknown variable, g-factors can be estimated for all free-flow time
periods. Free-flow conditions can be identified in the data as time periods where the occupancy
falls below 15%. For these conditions, free-flow speeds can be approximated based ideally on
speed data obtained from the field for different lanes and facility types. An example of speeds by
facility type, lane, and number of lanes derived from double-loop detectors in the San Francisco
Bay Area is shown in Table 3-2.
Using the assumed free-flow speeds, g-factors can be calculated for all free-flow time
periods according to the below equation:
k (t ) *T
g (t )
V free
Equation 3-7
q(t )
where:
g(t) is the g-factor at time of day t (a 5-minute period),
k(t) is the observed occupancy at t,
q(t) is the observed flow at t,
T is the duration of the time-period (for example, 5 minutes), and
Vfree is the assumed average free-flow velocity.
3-15
1
2
3
4
5
Table 3-2: Example of Measured Speeds by Lane, San Francisco Bay Area
# of
Lanes
Lane 1
HOV
1
65
HOV
2
65
ML
1
65
ML
2
71.2
65.1
ML
3
71.9
69.7
62.7
ML
4
74.8
70.9
67.4
62.8
ML
5
76.5
74.0
72.0
69.2
64.5
ML
6
76.5
74.0
72.0
69.2
64.5
64.5
ML
7
76.5
74.0
72.0
69.2
64.5
64.5
Type
6
7
8
9
10
11
12
13
14
15
Lane 2
Lane 3
Lane 4
Lane 5
Lane 6
Lane 7
65
64.5
Applying this equation to the data transmitted by a loop detector over a typical week
yields a number of g-factor estimates at various hours of the day for each day of the week, which
can be formed into a scatterplot of g-factor versus time. To obtain a complete set of g-factors for
each 5-minute period of each day of the week, the scatterplot data points can be fitted with a
regression curve, resulting in a function that gives a complete set of g-factors for each loop
detector. An example of this analysis is shown in Figure 3-8.
Figure 3-8: Example of Five-Minute g-Factor Estimates at a Detector
3-16
1
2
3
4
5
6
7
8
9
10
11
In the figure, the dots represent g-factors directly estimated during free-flow conditions,
and the line shows the smooth regression curve (3). The 5-minute g-factors from the regression
function can then be stored in a database for each loop detector, and used in real-time to estimate
speeds from flow and occupancy values. An example of the g-factor functions for the detector in
the left-most lane (Lane 1) and the right-most lane (Lane 5) of a freeway in Oakland, California,
is shown in Figure 3-9. As is evident from the plot, the g-factors in the left-lane are relatively
stable over each time of day and day of the week, but the g-factors in the right-lane vary widely
as a function of time.
These g-factors should be updated periodically to account for seasonal variations in
vehicle lengths, or changes to truck traffic patterns. It is recommended that new g-factors be
generated at least every three months.
12
13
14
Figure 3-9: Example Estimates of g-Factors at a Detector (I-880 North, Oakland, CA)
3-17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Estimating Segment Travel Times from the Average Speeds
Traffic monitoring systems typically assign short segments of the highway network to
each point sensor. (These segments can be the segments by which travel time reliability is
monitored, or aggregations of these short segments may form the TTRMS segments.) The
sensor-related segments often extend from midway to the upstream detector to midway to the
downstream detector. This practice is illustrated in Figure 3-10, where the segments assigned to
10 point sensors are shown. As can be seen, this midpoint-to-midpoint designation is common,
but there are exceptions, as in the case of the detector highlighted in the middle by the red
triangles, where the entire six-lane section is assigned to that detector and in the case of the first
detector to the right of milepost 7.12, which is at the eastern end of the six-lane section, but the
entire six-lane section is assigned to it. Hence, this discussion focuses on how to develop average
travel times for these point-sensor-linked segments.
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
For these infrastructure-based detectors, the process of calculating segment average travel
times from raw data requires (1) estimating spot speeds from the raw data; and (2) transforming
these speed estimates into segment travel times. Almost all estimation methodologies today
compute mean speeds by dividing the sum of the speed observations by the number of data
points observed. This value is then inverted to obtain the average travel rate (spot rate) at that
location of the sensor; and that rate is multiplied by the segment length to obtain an estimate of
the travel time for the segment. A shortcoming of this technique is that it misses the effects of
bottlenecks or other influencing factors, such as on-ramps, that produce nuances in the travel
speed across the segment.
An important aside is that the median travel rate is actually a more useful and stable
statistic, although it is not often employed. It is less sensitive to the impacts of outliers. One can
diminish the impact of outliers by removing them, but this can make it more difficult to fully
capture the impacts of incidents and other disruptive events.
33
Estimating Vehicle Travel Times based on Average Travel Times
34
35
36
37
38
39
As indicated earlier, even though the infrastructure sensors observe individual vehicle
speeds (or occupancies) they only report averages (speeds and/or occupancies) along with the
number of vehicles observed. Consequently, the infrastructure sensors cannot be used directly to
generate observations (or estimates) of the individual vehicle travel times across a segment. To
do this, separate field observations of the individual vehicle speeds have to be obtained or an
assumption must be made about the distribution of those speeds based on similar facilities. There
Source: PeMS
Figure 3-10: Assignment of Segments to Point Detectors
3-18
1
2
3
4
is a use case that illustrates how to do this. For situations where only the average (or median)
travel times are of interest, estimation of these vehicle travel times (and their distribution) can be
skipped, but when distributions of segment or route travel times for individual vehicles is the
focus of the analysis, then the estimates must be developed.
5
Vehicle-Based Sensors
6
7
8
9
10
11
Vehicle-based sensors directly report individual vehicle travel times, but only for some of
the roadway travelers. Hence, in this situation, the main computational steps are to (1) compute
segment travel times for individual vehicles by differencing the timestamps reported by the
readers at either end of the segment or route; (2) filter out non-representative travel times, such
as those involving a stop; and (3) determine if enough valid travel time observations remain to
compute an accurate representation of the travel times for the time period of interest.
12
Generating Passage Times for Bluetooth Readers (and other AVI Sensors)
13
14
15
16
17
18
19
20
21
22
23
24
25
It is common for Bluetooth Readers (and other AVI sensors) to observe the same vehicle
multiple times while the sensed device is in the vicinity of the reader. However, most of these
devices report out only one passage time. Figure 3-11 shows an extreme case for the sequence of
readings for a single vehicle (device) obtained by a Bluetooth reader that was located on US-50
between Sacramento and South L