ATMS TESTBED FINAL REPORT: Volume 1 - Executive Summary October 2006
California ATMS Testbed
PHASE III: Operational Research Implementation
Final Report
Volume 1
Executive Summary
submitted to:
California Department of Transportation
Office of Technology Applications
Division of Research & Innovation
c/ o Transportation Management Center
6681 Marine Way
Irvine, Ca 92618
submitted by:
Institute of Transportation Studies
University of California, Irvine
October 2006
ATMS TESTBED FINAL REPORT: Volume 1 - Executive Summary October 2006
i
TABLE OF CONTENTS
ORGANIZATION OF THE EXECUTIVE SUMMARY ..................................................................................... III
PART 1: TESTBED RESOURCES/ MANAGEMENT/ TECHNICAL ASSISTANCE........................................ 1
TESTBED INFRASTRUCTURE AND DATA STREAMS..................................................................................................... 2
ATMS TESTBED LABORATORY DEVELOPMENT AND MAINTENANCE........................................................................ 4
TESTBED MANAGEMENT ............................................................................................................................... ........... 5
PART 2: TESTBED DEPLOYMENTS.................................................................................................................... 6
TRAFFIC DETECTOR AND SURVEILLANCE SUB- TESTBED ( TDS2) .............................................................................. 7
IRVINE ARTERIAL SUB- TESTBED........................................................................................................................ ...... 9
V2SAT DETECTOR VERIFICATION AND DATABASE SYSTEM................................................................................... 11
TESTBED ATMS TRAINING AND DEVELOPMENT SYSTEM....................................................................................... 13
CALTRANS DISTRICT 12 REAL- TIME DATA INTERTIE .............................................................................................. 15
MOBILE TMC/ MOBILE SURVEILLANCE UNITS ........................................................................................................ 17
ANAHEIM NETWORK TESTBED ............................................................................................................................... 18
ORANGE COUNTY MICRO- SIM DATA LIBRARY........................................................................................................ 20
TRACER ............................................................................................................................... ................................... 21
ATMS TESTBED RESOURCE WEBSITE..................................................................................................................... 23
PART 3: TESTBED RESEARCH AND DEVELOPMENT.................................................................................. 25
DEVELOPMENT AND DEPLOYMENT OF CORRIDOR MANAGEMENT PROTOTYPE ....................................................... 26
DISTRIBUTED PARAMICS DEVELOPMENT................................................................................................................. 28
DEVELOPMENT OF THE CAPABILITY- ENHANCED PARAMICS SIMULATION ENVIRONMENT................................... 30
REAL- TIME ADAPTIVE SYSTEMWIDE RAMP METERING AND SIGNAL CONTROL...................................................... 31
PERFORMANCE EVALUATION OF ADAPTIVE RAMP METERING ALGORITHMS.......................................................... 33
PEER- TO- PEER ( P2P) TRAFFIC INFORMATION PROPAGATION.................................................................................. 35
DYNAMIC VEHICLE NAVIGATION IN A P2P SYSTEM................................................................................................ 37
ANALYTICAL STUDIES OF P2P INFORMATION PROPAGATION.................................................................................. 39
MOBILE THROUGHPUT OF 802.11B........................................................................................................................ . 41
IMPLEMENTATION TESTS OF A P2P NETWORK FOR INCIDENT EXCHANGE............................................................... 43
IMPLEMENTATION OF A TOOL FOR MEASURING ITS IMPACTS ON FREEWAY SAFETY PERFORMANCE ..................... 45
DEVELOPMENT OF A TOOL FOR MEASURING NON- RECURRENT CONGESTION IMPACTS OF ACCIDENTS .................. 47
ATMS TESTBED FINAL REPORT: Volume 1 - Executive Summary October 2006
ii
ORGANIZATION OF THE EXECUTIVE SUMMARY
This report summarizes research and development that has been conducted to position the Testbed to
support prototype deployment and evaluation of Advanced Transportation Management Systems ( ATMS)
products and services. The various elements contained in the report generally involve both research and
development; funding for the research aspects reported herein was provided by the Caltrans Division of
Research and Innovation ( DRI) under the Partners for Advanced Transit & Highways ( PATH) research
program with Testbed support funding for those aspects of the research that involve development and
management of the ATMS Testbed infrastructure in support of the Department’s research agenda. This
joint effort targeted research activities that, with Testbed support, might lead to candidate products for
deployment. In that sense, the research funded under this agreement has been in support of the Advanced
Transportation Management and Information System ( ATMIS) research program of DRI, PATH,
California Center for Innovative Transportation ( CCIT), and other such organizations.
The work accomplished under this agreement is divided among three basic categories:
1. Testbed Resources/ Technical Assistance/ Management
2. Testbed Deployment
3. Testbed Research and Development
They are summarized in the following sections; the Testbed Research and Development Activities are
discussed in detail in the Final Technical Report Volume II, which includes a set of accompanying
Testbed Technical Reports.
ATMS Testbed Report: Executive Summary October 2006
1
PART 1
TESTBED RESOURCES/ MANAGEMENT/ TECHNICAL ASSISTANCE
ATMS Testbed Report: Executive Summary October 2006
2
Testbed Infrastructure and Data Streams
The ATMS Testbed Program was initiated in early 1991 to provide an instrumented, multi- jurisdictional,
multi- agency transportation operations environment linked to university laboratories for real- world
development, testing and evaluation of near- term technologies and applications, and to serve as an
ongoing testing ground for California and national Intelligent Transportation Systems ( ITS) efforts.
Located in Orange County, California, and under the direction of the University of California, Irvine
( UCI) Institute of Transportation Studies, the Testbed is intended to:
• accelerate deployment through advanced technology research;
• demonstrate the readiness of advanced systems;
• implement and evaluate operations of an integrated multi- jurisdictional, multi- agency
transportation operations system.
The Testbed is based on real- time, computer- assisted traffic management and communication. The
transportation operations system that forms the backbone of the Testbed is structured to provide
intelligent computer- assisted decision support to traffic management personnel by integrating network-wide
traffic information ( both surface street and freeway) in a real- time environment. The Testbed
currently either has, or is developing, direct links to three traffic operations centers ( TMC), Caltrans
District 12 TMC, City of Anaheim TMC, and City of Irvine Transportation Research and Analysis Center
that provide real- time data links from area freeways and major arterials directly to dedicated Testbed
research laboratories located at UCI.
The broad mission of the Testbed Program is to work toward overcoming institutional, technical and
philosophical barriers to introducing innovative technologies into the management of complex
transportation systems. Working together with CCIT, California PATH and the Testbed Partners, the
Testbed Research Implementation and Prototype Development Program is designed to establish an
intermediary link between basic research in ATMS/ ATIS technologies and their full deployment.
The Testbed covers the entire freeway system in Orange County and two contiguous sub- areas
comprising an arterial system that includes most of the major decision points for freeway travelers in the
region. The City of Anaheim sub- area encompasses the City’s major special event traffic generators and
is centered about two of its designated " smart streets,” Harbor Boulevard and Katella Avenue. This sub-area
is ideal for network- wide applications of advanced technologies in traffic management. The City of
Irvine sub- area provides freeway access to many business and office complexes on both sides of the 1- 5
freeway and is ideal for corridor- level integration of real- time communication and control in traffic
management.
A comprehensive testing and evaluation system has been established to support activities in the Testbed.
The system has been developed to interface with existing traffic surveillance and control components and
provide a common integrated real- time traffic database for ATMS research conducted within the Testbed.
The system design is built upon a wide- area communications network backbone linking the Cities of
Anaheim and Irvine Transportation Management Centers ( TMCs) to the California Department of
Transportation’s District 12 TMC and with the ATMS Research Laboratories at the UCI Institute of
Transportation Studies. The communications network is configured to permit easy future expansion to
accommodate appropriate private/ public sector research implementation projects that may be conducted
within the Testbed.
ATMS Testbed Report: Executive Summary October 2006
3
The Testbed has the following main components:
1. Testbed Labs
2. Data Streams
a. Historical
b. Real- time
3. Surveillance
a. Freeway
b. Arterial
4. Wireless Data Capabilities and Real- time Vehicle Tracking
5. Travel Behavior Monitoring
6. Microscopic Simulation
7. Mobile Surveillance
8. Mobile Transportation Management Capability
9. Testbed Resource Web Access
The Testbed laboratories form a computerized research environment connected to the real world
transportation system. The laboratories are a testing ground for the development of particular ATMIS
modules and of integrated ATMIS applications. The goal is for the Testbed laboratories to have a
complete simulation of the transportation systems that are part of the Testbed. The Paramics ( parallel
micorsimulation) traffic model is the core simulation for the Testbed laboratories. It can simulate all of
the existing and currently envisioned traffic measurement and control devices associated with ATMS.
The Testbed Real- time Integrated Control and Evaluation Prototype System ( TRICEPS) was developed as
an implementation platform to provide plug and play capabilities for the testing and evaluation of a range
of ATMIS modules. TRICEPS permits modules to be configured for various ATMIS applications and run
in simulated, real- time, and integrated modes. Any particular ATMIS application thus can be
implemented by being connected to both simulated and real- world transportation systems so that overall
effectiveness can be first assessed in the laboratory and then evaluated in the field.
Access to Testbed research, testing and evaluation resources is provided via an integrated website. The
web page features a real- time graphical user interface whereby researchers and Caltrans staff can
download information ( both historical and real- time) from Testbed data collection services ( e. g., real- time
freeway and arterial loop data and signal status, probe vehicle information, video surveillance, detector
testbed data, etc.), as well as coded networks, various API’s and other software tools developed under the
Testbed using a browser for query. The web page is available on the Testbed network as well as via the
Caltrans Wide Area Network.
Technical Assistance to Caltrans, CCIT, and Partners
The purpose of this element is twofold: 1) to provide a continuing capability for Caltrans to analyze and
measure the expected net benefits of ATMS projects proposed by Caltrans, typically by applying a
microscopic simulation approach to evaluation of traffic improvement projects, and 2) to provide
technical assistance to Caltrans, CCIT, and Testbed Partners in their use of Testbed resources.
Work conducted under this element focused on providing an “ on demand” environment for Caltrans’
strategic planning exercises involving both ATMIS as well as capital improvement projects, using the
analysis capabilities of Paramics, the physical testbed and the rest of the Testbed Workbench applications.
ATMS Testbed Report: Executive Summary October 2006
4
In addition, human resource capability was maintained to service “ on demand” requests for such analyses
and evaluations.
The UCI Testbed Team provided assistance in four broad categories:
1. On- Call Direct Support of Paramics
2. Technical Guidance on Micorsimulation
3. Micorsimulation Research- related Support
4. Technical Support of Use of Testbed Resources
The UCI Testbed team provided technical support for Application Program Interface ( API) developed by
Testbed personnel on an on- call basis. Due to PARAMICS new builds and updated versions, the released
APIs needed occasional maintenance to ensure full PARAMICS compatibility.
As different scenarios that could not be handled by current API versions were identified by Caltrans, an
assessment was made by the project team to define necessary API modifications and these enhancements
then programmed. For any modified or new APIs developed as part of this task, both source code and
technical documentation were provided to Caltrans.
Microscopic simulation models have the capability to evaluate transportation facility plans and design
alternatives as well as traffic operation strategies. However, lack of appropriate guidance and/ or direction
may lead to inappropriate use of and inaccurate results from these models. It is anticipated that, as
Caltrans' districts use micorsimulation models with increasing frequency, a standardized guidance and
evaluation process would facilitate model application and analysis, as well as the comparative
interpretation of model results. The UCI Testbed team provided Caltrans with guidance and evaluation
for simulation studies relative to traffic modeling and micorsimulation.
Upon request, the UCI Testbed team evaluated Caltrans on- going simulation studies, and provided input
to guidelines for application of micorsimulation models. It is anticipated that experience gained with the
current micorsimulation applications in California will contribute significantly to the development of
guidelines.
As expected, there is increasing demand to utilize Testbed resources in support of deployments within the
Testbed. The Testbed team maintained a capability to technically support such deployments.
ATMS Testbed Laboratory Development and Maintenance
The UCI Institute of Transportation Studies is responsible for the overall development and management of
the ATMS Testbed Laboratories. Acting on behalf of the Caltrans Division of Research and Innovation, it
continued to:
• develop necessary software enhancements to bring prototype Testbed model simulation systems to
full operational capability.
• develop protocols for “ plug and play” operation.
• develop API’s for research and prototype deployments conducted within the Testbed.
• develop and coordinate an equipment and software upgrade and replacement program to keep
Testbed capabilities within current industry standards.
ATMS Testbed Report: Executive Summary October 2006
5
• incorporate and integrate appropriate simulation/ modeling advancements emanating from associated
Caltrans research programs and other national/ international efforts within the Testbed modeling
environment.
• maintain and manage the Testbed Communications Network, including security provisions.
Testbed Management
The UCI Institute of Transportation Studies is responsible for the overall management of Testbed operations.
Acting on behalf of the Caltrans Division of Research and Innovation, it continued to:
• develop and maintain the Testbed facilities in a manner that facilitates the overall goals of Caltrans
DRI, CCIT, and the Testbed Program.
• establish and coordinate the activities of the various management committees charged with
overseeing Testbed operations.
• act as liaison between CCIT, PATH researchers, private and public sector users and the Testbed
operating agencies that comprise the Testbed partnership.
• identify and coordinate enhancements to the Testbed that serve its basic functions.
• coordinate and assist in the procurement of hardware/ software necessary for the conduct of
deployments within the Testbed.
• develop and execute a strategic plan for realizing the maximum potential from Testbed activities.
• represent the interests of the Testbed at national and international forums.
• manage the Testbed Communications network.
• supervise and coordinate the activities of Testbed researchers assigned to the Testbed Program.
• act as Caltrans agent in all such other Testbed activities not specifically delegated.
ATMS Testbed Report: Executive Summary October 2006
6
PART 2
TESTBED DEPLOYMENTS
ATMS Testbed Report: Executive Summary October 2006
7
Traffic Detector and Surveillance Sub- Testbed ( TDS2)
Summary
• What is it?
– A real- world laboratory for the development and evaluation of emerging traffic detection
technologies relative to: appropriateness for ITS operations and performance
measurement, data quality and consistency, ease of use, ease of installation, and overall
cost.
• What does it consist of?
– The latest- generation IST detector cards installed along both the freeway mainline as well
as on the associated on- and off- ramps; these installations provide data for signature
analysis associated with the Caltrans REID vehicle reidentification development.
• How has it been used?
– This facility has already proved extremely useful in the testing and evaluation of
overhead- mounted detectors ( laser), side- mounted detectors ( OMRON Vision Sensor
Detector, Wavetronics and RTMS radar sensor detector, Pan- Tilt- Zoom Webcam, Spread
Spectrum Radio), and in- pavement detectors ( enhanced double loop, High- resolution
Detector Blade).
• How can it be used?
– Test and evaluate new sensors for traffic management; obtain section- related speeds;
vehicle classification; estimate O- D patterns
Description
To fully exploit the benefits of the new generation of Intelligent Transportation Systems now under
development, including applications for homeland security, more accurate and appropriate real- time
traffic data need to be collected from the urban highway transportation network and communicated to
traffic management centers, traffic operations personnel, travelers, and other agencies.
To assist in identifying and evaluating detector
technologies for ITS deployment, the Testbed initiated
development and construction of a Traffic Detector and
Surveillance Sub- Testbed ( TDS2) within the Testbed.
The overall purpose of the Traffic Detector/
Surveillance Sub- Testbed is to provide a real- world
laboratory for the development and evaluation of
emerging traffic detection technologies relative to:
appropriateness for ITS operations and performance
measurement, data quality and consistency, ease of use,
ease of installation, and overall cost.
The first phase of the development of TDS2 focused on
the installation of “ ground truth” sensing equipment
and attendant communications infrastructure at two
contiguous sites along the I- 405 freeway within the
Testbed — at I- 405 and Laguna Canyon Road, and at I- 405 and Sand Canyon Avenue. At these sites,
Ground Truth Camera Installation
ATMS Testbed Report: Executive Summary October 2006
8
“ ground truth” video cameras were installed over each lane of traffic, connected to a bank of VCRs and to
an automated video image capture and re- identification system. Data from these systems are transferred
to the UCI Testbed Laboratories via a combination of wireless technology ( SSR) and fiber optic ( F/ O)
cable. This facility has already proved extremely useful in the testing and evaluation of overhead-mounted
detectors ( laser), side- mounted detectors ( OMRON Vision Sensor Detector, Wavetronics and
RTMS radar sensor detector, Pan- Tilt- Zoom Webcam, Spread Spectrum Radio), and in- pavement
detectors ( enhanced double loop, High- resolution Detector Blade).
Work under this component continued the development of a real- world corridor- level testbed facility that
would permit field testing and evaluation of advanced traffic sensor, detection and surveillance
technologies to enhance the ability of Caltrans to deploy such technologies as well as advanced systems
for improved traffic management and control of California's freeways and signalized arterials.
Specifically, 18 detector cards were purchased to populate the TDS2 cabinets. IST Inc., a Testbed private
sector partner, developed and implemented web- based software to collect, store and make available the
traffic data collected on NB Rte. 405 to researchers over the internet ( a mysql database server was created
to collect all incoming vehicle event records). A total of 120 additional signature cards were secured for
the 18 new sites along NB Rte. 405 from the El Toro Y to Jamboree. IST also provided enhancements to
the GUI web interface.
In its current configuration, TDS2
comprises a fully instrumented facility
for the testing and evaluation of
advanced detectors— one that has been
utilized extensively ( with considerable
success) by Caltrans staff to evaluate
several alternative detection
technologies. With the completion of the
current phase of construction, an 8 km
corridor comprising both a freeway ( I-
405 NB) and adjacent arterial ( Alton
Parkway, both directions, including 18
intersections) has been fully
instrumented to provide real- time data
streams detailing traffic movement
through the corridor; in addition to
providing raw loop data for all detectors
on all approaches, the arterial subsystem
is instrumented to furnish real- time status
information of the signal displays at each
of 18 intersections. The freeway
locations have the latest- generation IST detector cards installed along both the freeway mainline as well
as on the associated on- and off- ramps; these installations provide data for signature analysis associated
with the Caltrans REID vehicle reidentification development. For the arterial sites, 2070 controllers and
associated detection software ( ACTRA) have been deployed at the 18 intersections and are planned to be
outfitted with IST detector cards and blade sensors. Software elements of the project included
modification and enhancement of existing field to central server communications and real- time data base
and web- site access, deployment and evaluation of advanced vehicle reidentification methods for
anonymous real- time tracking of individual vehicles, real- time travel time and performance measurement,
real- time management and control, and potentially homeland security applications.
Freeway IST Detector Installation
ATMS Testbed Report: Executive Summary October 2006
9
Arterial 2070/ ACTRA Installation
Irvine Arterial Sub- Testbed
Summary
• What is it?
– An arterial detection and signal system built on 2070 technology designed to test and
evaluate combined freeway/ arterial traffic management strategies.
• What does it consist of?
– The latest- generation 2070 controllers installed along Alton Parkway; these installations
provide stopline and extension detection, signal phasing data.
• How has it been used?
– This facility is the centerpiece in the testing and evaluation of the CARTESIUS corridor
real- time management tool
• How can it be used?
– Test and evaluate adaptive signal control algorithms; develop and evaluate use of
signature analysis to estimate turn counts in real time; test and evaluate arterial “ point- of-inception”
ramp metering effectiveness; test and evaluate combined arterial/ freeway
incident management approaches
Description
A state- of- the- art Siemens ACTRA Central
Traffic Control System and eighteen 2070NL
intersection controllers with SE- PAC
firmware was installed along an 18-
intersection adaptive control testbed in the
City of Irvine ( COI) that supplies real- time
traffic data to the UCI ATMS Testbed
laboratories utilizing single mode fiber optic
communication operating over 4
communication channels. The Eagle ACTRA
system provides detection and adaptive control
capabilities utilizing existing Ethernet
communications.
The communications architecture for the
complete systems is designed to isolate the
Testbed functions from the COI’s traffic
system, thus allowing researchers to
interact with the traffic system without the
possibility of disrupting normal city
operations.
Input Acquisition Software ( IAS) was developed to enable transmission of all detector inputs and signal
displays to the UCI Testbed laboratories. The 2070 controller and its corresponding field I/ 0 module
function as two separate computers. They each have their own memory, CPU, and firmware. The job of
the field I/ 0 module is to interpret serial data coming from the controller via a synchronous serial link ten
ATMS Testbed Report: Executive Summary October 2006
10
times a second and set the state of the hardware outputs to whatever the traffic program in the 2070 wants
them to be. The field I/ 0 module reads the state of the hardware inputs ten times a second and reports
these back to the main 2070 CPU over a synchronous serial
link. None of the inputs or outputs has meaning to the 2070
hardware, only to a traffic program running in the 2070.
The IAS replaces the SE- PAC traffic software with a
program ( not a traffic control program) that simply looks at
the serial data stream coming back from the 2070- 8 NEMA
base every tenth of a second, picks out the input status bits
( either on or off) and reports these back to a host program
via a UDP packet. This approach is very elegant because the
host can see the state of all of the inputs into the 2070- 8
NEMA base for every 10th second of time. The host may
then archive or process the data as desired. The software
supports the following:
• Allow communications via Ethernet using the UDP protocol on port 20000.
• Allow the user to remotely query the date and time of the unit.
• Allow the user to remotely set the date and time of the unit
• Allow the user to specify a rate at which the state of the filtered input bits on the - 8 NEMA base
are transmitted to the user.
• Periodically send the user the state of the filtered input bits on the - 8 NEMA base at the specified
rate.
Communications Architecture
ATMS Testbed Report: Executive Summary October 2006
11
V2SAT Detector Verification and Database System
Summary
• What is it?
– A video image processing system that detects and classifies passing vehicles.
• What does it consist of?
– 14 video detection cameras above the same number of traffic lanes, attached to the bridge
railing with unique mounting brackets, together with specialized software employing
fuzzy inference methods.
• How has it been used?
– Provided ground truthing of vehicle reidentification research based on IST magnetic
induction signature analysis.
• How can it be used?
– Enhanced surveillance system that provides accurate speed estimation, vehicle
classification, and O- D information
Description
This project involved the design and deployment of specialized equipment and technical services to
support the testing of various types of traffic detectors in the Caltrans Irvine Detector Testbed. The
primary hardware/ software deliverable was the V2SAT ( Video
Vehicle Signature Analysis and Tracking) system, a distributed
video data acquisition system configured to serve as a ground-truth
system for the Testbed. The V2SAT system includes
vehicle re- identification capabilities for tracking of vehicles
among the three primary test sites located on the NB Route 405
freeway at Laguna Canyon overcrossing, Sand Canyon
overcrossing, and Sand Canyon off ramp. In addition, we were
responsible for the design and installation of all video cameras
and video- related components, the performance of most of the
field tests, and the design, operation and maintenance of the
central Testbed server which functions as the central repository
for all detection test data and verification images.
Hardware/ software deliverables included:
• 14 V2SAT site detection computers installed at three field locations
• One central server located at UCI/ ITS
• Software for video detector ground- truth verification, verification and archiving on central server, and
vehicle reidentification
• Deployment of 14 video detection cameras above the same number of traffic lanes, attached to the
bridge railing with unique mounting brackets
• Equipment for precise time- synchronization of video- tape records acquired at all sites
• 3kW UPS at all field sites
V2SAT Camera Installation
ATMS Testbed Report: Executive Summary October 2006
12
• System Operators Manual and Technical Supplement
The V2SAT computer vision systems acquire an image of every
vehicle passing below each down- looking video camera. These images
and the related detection data are maintained in a distributed database
accessible from the central server located at ITS. During daylight
hours, reliability of detection has been found to be over 99%, with
fewer than 0.2% false detections. Most errors are due to ambiguous
lane positions of vehicles during lane changes or merging.
Detector data verification and reduction is done on the server, which
receives data from all field units and can be remotely accessed by
authorized UCI and Caltrans personnel, as well as vendors participating
in studies.
In addition, the vehicle re-identification
feature of the
system matches vehicles
detected at the Laguna Canyon Site with vehicles detected at
the Sand Canyon Site. The computer vision systems analyze
acquired video images and generate relatively small numeric
vectors, which characterize each vehicle. Vectors acquired at
each site are transferred over the Testbed wireless network to
the V2SAT server located at UCI/ ITS. This function does not
work at night since it requires ambient illumination.
Vectors are compared using fuzzy inference methods, to match
vehicles, which were detected at both sites, and reject vehicles
detected at only one site. Computer verification tools such as
the one shown at the right are provided to confirm the accuracy of the reidentification, while also
providing a means for monitoring and cataloging all traffic during test periods at all Testbed locations.
Vehicle images shown in the user interface screenshot can be set to update in near real- time, or may be
selected from archives of previously acquired data.
V2SAT Images
V2SAT Server
ATMS Testbed Report: Executive Summary October 2006
13
Testbed ATMS Training and Development System
Summary
• What is it?
– A high- end simulator for use in Caltrans operator training efforts as well as in the
development of a true testing environment for ATMS upgrades and enhancements within
an environment with the full functionality of a Caltrans TMC.
• What does it consist of?
– An ITS Test Center and TMC Simulator that provides connectivity to the Testbed’s real-time
data streams ( field and simulated). Provides both “ live” and “ virtual” connections to
Caltrans field traffic control systems including vehicle detector stations, dynamic
message signs, and camera stations. An adjoining “ classroom” has data and video
input/ output capability to ATMS Simulator software ( ATMSSIM). The classroom has
capacity for fifteen Caltrans trainees.
• How has it been used?
– Provides Caltrans TMC operator training.
• How can it be used?
– Test and evaluate upgrades/ enhancements to TMC operations in a secure environment
that exactly duplicates actual TMC software systems and data feeds.
Description
The Testbed ATMS Training and Development System brings operational capability to the TMC
Simulation Laboratory, providing networking that links the simulator to the Testbed Laboratories. This
provides a foundation for the evolution of an ATMS development environment that can be used to
support isolated, independent testing and evaluating of ATMS products prior to their deployment in a
district operating environment, as well as to provide operator training using the latest version of the
ATMS software in a simulated environment.
The system features a stable network in support of the ITS Test Center and TMC Simulator that provides
connectivity to the Testbed’s real- time data streams ( field and simulated). The network has been
designed to isolate network traffic depending on the nature of services needed or open communication
among different subnets. A simulation environment is in place that features:
1. a “ live” connection to Caltrans field traffic control systems including vehicle detector stations,
dynamic message signs, and camera stations;
2. a “ virtual” ( through simulation) connection to Caltrans field traffic control systems including
vehicle detector stations, dynamic message signs, and camera stations;
3. an ATMSV2 Graphical User Interface ( GUI) with geographic map depicting freeways and field
element icons for displaying traffic status by levels of speeds, volumes, or occupancies at vehicle
detector stations, and simulated status for dynamic message signs, and camera stations;
4. Connection of ATMSV2 software to Paramics for simulation of real- time operation.
In a joint effort among the Testbed, Cal Poly, and Caltrans Division of Traffic Operations, a realistic
training environment has been provided for Caltrans TMC operators based on a simulation version of
ATMSV2 that has been developed by NET specifically for use by the Testbed in support of Caltrans’
ATMS Testbed Report: Executive Summary October 2006
14
TMC operator training program. The ATMSV2 Simulation ( ATMSSIM) is based on the port of District
7’ s ATMSV2 with 7.2 functionality to District 11 using upgraded COTS software.
With this system, the Caltrans TMC Operator Training is being conducted in the Testbed facility at UCI.
The Testbed ATMS Training and Development System both supports Caltrans’ operator training efforts
as well as establishes a true testing environment for ATMS upgrades and enhancements.
Training operations are supported by a freestanding,
custom designed, Display Wall System. The Display
Wall System includes ( 5) Christie Digital XGA DLP
projectors. The system features a Christie Digital FRC
5100 Display Wall Processor, designed for 24/ 7
operation as an integral part of the video wall, capable
of driving ultra high resolution tiled wall displays of
virtually unlimited local or network- based software
applications, real- time video inputs and multiple RGB
Inputs. The processor can be controlled through a
touch panel, RS- 232 and IP with software on local
workstations.
The processor hardware is configured to accept ( 20)
video feeds and ( 4) RGB sources. A VGA and a
Composite video feed are routed to adjacent rooms for
viewing. A Crestron TPS- 5000 connected via
Ethernet to the Caltrans Pro2, switches the camera inputs at the matrix switcher. A Crestron Pro2 CPU
provides control of local TMC equipment.
Operator training itself is carried out in an adjoining
“ classroom,” that has data and video input/ output
capability to ATMSSIM. The classroom has capacity
for fifteen Caltrans trainees. Video and audio
surveillance of the trainees is provided to the
instructors in the control room. Using this facility,
TMC operator personnel will be able to interact both
with actual, real- time or “ playback” field data as well
as with specific operational scenarios developed from
loop- simulated data generated by the Paramics real-time
microscopic traffic model.
TMC Operator Training Classroom
TMC Display Wall
ATMS Testbed Report: Executive Summary October 2006
15
Caltrans District 12 Real- time Data Intertie
Summary
• What is it?
– A real- time data retrieval subsystem that interfaces directly with the Caltrans District 12
Front End Processor ( FEP), providing loop detector data from all loop stations and real-time
video from any two CCTV stations, in a dedicated, independent environment that is
functionally isolated from the Caltrans District 12 operating system.
• What does it consist of?
– A real- time data feed of 30- second “ poll- data” from all of Caltrans District 12’ s SATMS
170 controllers via a secure network connection between D12 FEP and the UCI Testbed
private network, restricted by the D12 firewall to permit transmission only from the D12
FEP to the UCI Testbed Labs; transmissions to the D12 FEP from the UCI Testbed Labs
are strictly prohibited. The real- time data is received by a dedicated server at the UCI
Testbed Labs; the data is then stored in a database and made available for distribution to
various real- time research and development activities.
• How has it been used?
– Has provided data for numerous PATH and other Caltrans research projects, as well as
been the foundation for TMC Operator Training.
• How can it be used?
– Test and evaluate upgrades/ enhancements to TMC operations, including adaptive control,
in a secure environment that exactly duplicates actual field conditions and data feeds.
Description
The data intertie with Caltrans District 12 has been configured based on a system architecture that, to the
extent possible, renders the UCI ATMS Testbed Intertie to function in a dedicated, independent
environment that is functionally isolated from the Caltrans District 12 operating system. Specific care has
been given to adopting a design that will ensure that the Testbed Intertie remains operational, with
minimal revision, as the Caltrans District 12 ATMS is upgraded or is otherwise changed.
The real- time data retrieval subsystem has been designed to interface directly with the Caltrans District 12
FEP system that includes a dedicated computer that collects data from traffic controllers via modems and
interfaces with the outside world via TCP/ IP network connection. It acts as a small central repository for
the data collected from all of the traffic controllers that are connected to it. As it has limited memory it
cannot hold large quantities of data, but does have the ability to respond to requests from a remote
computer that wants to receive a copy of the traffic controller data that it collects. During normal
operation the FEP continuously polls the traffic controllers and stores the returned data internally,
together with diagnostic information regarding the communication link, timestamps, etc. This
information is held in a small internal buffer and the oldest data is overwritten by the newest as the data
arrives.
In the Testbed laboratories, a RECEIVER program that runs on a separate workstation under the Solaris 7
operating system ( a version of UNIX) retrieves this internally stored data over a standard TCP/ IP
network. This program instructs the FEP to send whatever traffic controller data that it has stored in
RAM, together with diagnostic information, back to the RECEIVER program. The RECEIVER program
then formats this information in a human readable way and outputs it to a stream ( e. g., console or disk
file).
ATMS Testbed Report: Executive Summary October 2006
16
Under the current Testbed integration with D12’ s ATMS and SATMS- 170 controllers, the Testbed Labs
at UCI receives the same type of data feed from the D12 FEP that the D12 ATMS receives. This data
stream is a real- time data feed of 30- second “ poll- data” from all of D12’ s SATMS 170 controllers. The
network connection between D12 FEP and the Testbed private network is restricted by the D12 firewall to
permit transmission only from the D12 FEP to the Testbed Labs; transmissions to the D12 FEP from the
Testbed Labs are strictly prohibited.
The real- time data feed from the D12 FEP is received by machine Nemesis at the Testbed Labs; the data
is then stored in a database and made available for distribution to various real- time research and
development activities. The diagram below outlines the current configuration.
ITS Priv ate Netw ork
Control Algorithm
D12- FEP
...
Nemesis @ ITS
( Sun/ SPARC/ Solaris 2.7)
DB- receiver
Recorder
repeater
Oracle
DB
Other
Development
Machine
( typical)
UNICON
D12 Firewall
D12
ATMS
Schematic of Caltrans District 12 Real- time Data Intertie
High- speed switching and firewall equipment provides a secure gigabit Ethernet data link between the
two sites. Data from the D12 FEP is sent from the TMC to the Testbed Labs over a fiber link. Two
simultaneous video feeds from the Caltrans D12 TMC cameras across the fiber link to the Testbed Labs
are available to visually verify traffic conditions. A touch screen panel is used as the physical interface
for the Testbed Labs to select the traffic feeds from the Caltrans D12 TMC.
ATMS Testbed Report: Executive Summary October 2006
17
Mobile TMC/ Mobile Surveillance Units
Summary
• What is it?
– A mobile laboratory equipped with full functionality of a Caltrans TMC.
• What does it consist of?
– The Mobile TMC and the Mobile Ramp Metering and Surveillance units are capable of
operating in conjunction with the Caltrans District 12 TMC or independent of it. These
two systems provide on- site traffic data collection, video surveillance, and management
functions to support Caltrans TMC operations.
• How has it been used?
– Provides on- site traffic data collection, video surveillance, and management functions to
support Caltrans TMC operations
• How can it be used?
– Provide surveillance in construction zones; provide surveillance and ramp metering
functions for event and major incident management.
Description
The Mobile TMC and its integration with mobile ramp- metering and surveillance trailers that were
developed under previous Testbed contracts are designed to improve incident/ disaster response and
managing special events traffic situations through the use of mobile transportation management systems
and field elements. The Mobile TMC supports TMC operations by performing the following functions:
1. Assist in incident management and highway system recovery
2. Act as a relay station for mobile surveillance trailers
3. Assist in interagency field coordination on the scene of major events or incidents
4. Provide a temporary facility in the event a TMC becomes disabled due to a catastrophic event.
The Mobile TMC was developed and has been maintained for Caltrans Division of Research and
Innovation ( DRI) between 1999 and 2004 as part of the ATMS Testbed. The Mobile Ramp Metering
( MRM) system, developed previously, has been modified and integrated to operate in the field with the
Mobile TMC. The Mobile TMC and the MRM are capable of
operating in conjunction with the Caltrans District 12 TMC or
independent of it. These two systems provide on- site traffic
data collection, video surveillance, and management functions
to support Caltrans TMC operations. Performance
specifications for both the Mobile TMC and the Mobile Ramp
Meter have been developed, allowing procurement of
additional units to support Caltrans traffic operations ( see
Volume II, TTR3- 23).
Baxall codecs allow relay of video from the Mobile TMC to
the District 12 TMC over the Mobile TMC’s satellite
communications link. A 5.5kw Onan generator powers all AC
electrical equipment on board the Mobile TMC, including on
board computers and satellite communications system.
Mobile TMC
ATMS Testbed Report: Executive Summary October 2006
18
Anaheim Network Testbed
Summary
• What is it?
– A system of 2070 controllers running the SCOOT adaptive control system in the
convention/ Disneyland area of the City of Anaheim ( COA).
• What does it consist of?
– Implementation of an interface for remote control access of the COA’s SCOOT system
field elements, providing the capability for closed- loop control, based on industry-standard
communication technologies.
• How has it been used?
– Test and evaluation of implementation of SCOOT adaptive control system using
California- standard loop detector placement
• How can it be used?
– Test and evaluate adaptive signal control algorithms; test and evaluate event management
approaches
Description
A key operational element of the Testbed is the TMC located in the City of Anaheim ( COA) and used to
manage the flow of traffic in the Anaheim portion of the Testbed. The City SCOOT system is an
enhanced deployment that enables ATC Model 2070 controllers to be integrated into SCOOT with a more
robust protocol to accommodate real- time traffic information exchange to the TMC. A real- time data
intertie between the Anaheim TMC and the Testbed Laboratories allows exchange of data from this
system with other Testbed partners, enabling research into network- level control and traveler information
systems, utilizing real- time traffic information for input into the various models and algorithms, and also
providing closed- loop control capability of the field devices for validation of the technologies under
development.
With this system, the City’s traffic control system disseminates real- time traffic signal controller
information to the Testbed labs. This provides a framework for establishing SCOOT real- time control
capabilities from the Testbed Laboratories, and can be used in the development and implementation of an
interface for remote control access of the COAs SCOOT system field elements, providing the capability
for closed- loop control, based on industry- standard communication technologies.
The diagram outlines the City of Anaheim TMC system configuration. Arrows in the diagram reflect data
flows related to Testbed data acquisition. The SCS ( SCOOT COM. Server) is currently capable of
delivering data to two network nodes. Node- 1 is the SCOOT system and node- 2 is the COA Data Views
Application ( DVA) system. The approach taken was to redirect SCS node- 2 to the Testbed SAF ( Store &
Forward) system. SAF system software is then be responsible for storing the data locally and forwarding
the SCS data back to the COA DVA system ( via XMIT- 1) and also forwarding the SCS data to machine
‘ Nemesis’ at UCI ( via XMIT- 2). The diagram also shows the various components and the normal
operational data flow. The Testbed hardware is installed at the COA TMC and consists of the following
components:
• The Store- and- Forward ( SAF) system is a rack- mounted mini- server that is configured with two
network adapters. This system runs the Store- and- Forward ( SAF) application components.
ATMS Testbed Report: Executive Summary October 2006
19
• The Firewall- VPN box provides network security for the SAF system and indirectly for the COA
TMC.
• DSL supports the Testbed access to the SAF system within the COA TMC.
As the COA TMC real- time traffic data is acquired by the SAF system, it is stored locally in a circular
buffer located in a shared memory data structure ( SMDS). Once the data are stored in the SMDS they are
then retransmitted to the COA DVA machine as well as to a machine at UCI. The local buffering on the
SAF system allows the data acquisition process to proceed without concern for the speed of outgoing
transmission processes. It also allows multiple transmitting agents to operate completely independently
and without side effects. The
design allows for the continuous
recording of data locally and the
reliable forwarding of data to
the COA DVA system even if
the network link to UCI should
go offline. The storage
requirement on the SAF system
is minimally only a few
seconds. This minimal interval
provides sufficient storage to
prevent a loss of data during
typical momentary networking
delays; however, this buffer
area has been sized to
accommodate several days of
COA TMC data. In this way an
extended network downtime can
be tolerated without an actual
loss of data— data would just
continue to be stored locally
until the network link is restored
( up to the limit of the buffer used).
This allows Testbed management
a reasonable amount of time to
notice a networking problem and arrange for link repair prior to a loss of data. Once the link is restored,
the SAF system software would simply forward the data stored locally over to system Nemesis at UCI as
fast as the network link will allow.
On the UCI side of the connection, a COA TMC data reception application RCVR- 2 has been developed
to accept the COA TMC real- time data stream as it arrives from the SAF system. This application stores
the data in the existing Testbed Oracle database for distribution on the Testbed website.
COA TMC
UCI Testbed Lab
COA TMC Equipment Lay out ( High- Lev el View)
SCOOT Com
Server ( SCS)
Included in this proposal DataViews
App ( DVA)
SCOOT
COA TMC Network # 1
ACTRA
ACTRA
Workstation
COA TMC Network # 2
Store & Forward System
COA Data
Acquisition
COA Data
XMIT- 2
NIC
NIC
COA Data
XMIT- 1
Firewall - VPN
UCI machine
Example COA
Data RCVR- 2
Firewall - VPN
ACTRA
Workstation
UCI/ Testbed IP- Network
Future
NIC
Replaced by ( A) & ( B)
Oracle
DB
( A)
( B)
COA Testbed Network Configuration
ATMS Testbed Report: Executive Summary October 2006
20
Orange County Micro- sim Data Library
Summary
• What is it?
– A series of coded Paramics networks, and associated data.
• What does it consist of?
– Coded freeway and arterial networks within Caltrans District 12, including: API plugins
developed by the Testbed researchers; network geometry data; OCTAM planning model;
Caltrans Tachrun data; traffic counts from cities of Irvine and Anaheim.
• How has it been used?
– Used in a variety of micorsimulation studies
• How can it be used?
– Test and evaluate traffic management strategies
Description
Microscopic simulation studies need various data as inputs. Data used for network coding includes
background images, network data, traffic control data, traffic detector data, traffic zones and demand
tables, and vehicle data. Data used for model calibration include traffic counts and travel time along
freeways and arterials. Due to the requirement of large amount of data for microscopic simulation studies,
researchers may take a long time to collect data. The Orange County Micro- sim Data Library establishes
a secure web site that allows traffic agencies, such as Caltrans and OCTA, and universities to share all
available data securely.
The secure website was implemented using Apache,
PHP, and MySQL. Briefly, the web service is designed
to run on a Windows machine, using the Apache web
server, together with PHP ( a scripting language for
writing web- based applications) and MySQL ( the
database that the PHP application uses). The application
itself is a PHP application built from the ground up,
making use of standard HTML Forms as the basis for the
interface. Necessary security technologies ( SSL) were
setup and integrated into the PHP application.
The secure website provides functions for system
administrators to manage the website and its data. The
registered users can upload or download data. Currently,
the site has about 1GB data from Caltrans District 12,
City of Irvine, Orange County Transportation Authority
( OCTA), including ( 1) API plugins developed by the
Testbed researchers; ( 2) Network geometry data; ( 3) Previously coded Paramics networks; ( 4) OCTAM
planning model; ( 5) Caltrans Tachrun data; ( 6) Traffic counts from cities of Irvine and Anaheim.
The secure website provides an efficient mechanism to share data between different traffic agencies and
universities; its continued usefulness will depend on the involved traffic agencies making a practice of
uploading any new data to the secure website.
Sample Paramics Network
ATMS Testbed Report: Executive Summary October 2006
21
Tracer
Summary
• What is it?
– A GPS- based travel data collection and analysis system.
• What does it consist of?
– Software and hardware designed to collect and store GPS data for later retrieval and
analysis. These devices consist of a Garmin GPS antenna ( either a Garmin 16 or Garmin
35), and a small embedded computer running Linux, which enables data collection,
communication, and so on. The data are stored locally on a CompactFlash card, and can
also be transmitted wirelessly via CDPD modem, via WiFi wireless network connections,
or via a traditional wired ethernet connection.
• How has it been used?
– Used to generate travel diary information, travel speeds and route choice.
• How can it be used?
– Generate vehicle trajectory information. Can use this geographic data fusion to enable
Caltrans staff to identify loop detectors that need recalibration in order to accurately
report traffic characteristics.
Description
At its core, the fundamental task of the Tracer system is to collect and store GPS data for later retrieval
and analysis. From this simple core, however, has grown a large, complex system. The Tracer database,
as produced by this project, has been designed to manage this complex, evolving system.
The device that performs this data collection is the Extensible Data Collection Unit, or EDCU. Owing to
its flexibility and extensibility, there are multiple configurations of the EDCU that can be used at any
given time. All of the device attachments and optional configurations are included in the database
schema.
Any qualified Testbed researcher can request one or more EDCUs. The EDCUs might be used by the
researcher directly in a mapping study or a traffic speed survey, and so on. Alternately, the researcher
might be conducting a travel behavior study, and therefore will need to subsequently distribute those
EDCUs to any number of survey participants, if necessary. The Tracer database is able to track who is
using the EDCU, and who has rights to view which GPS records.
A GPS point is based on well- defined latitude and longitude coordinates. Specifically, the GPS antennas
used by the EDCUs record their data in what is known as the WGS- 84 datum. Therefore, in order to be
compared with data from another datum ( such as TIGER/ Line files) or using another projection, one must
be able to convert the GPS data correctly. The Tracer database handles this by leveraging the open source
Post- GIS database library, built on top of the PostgreSQL database. In addition, TIGER/ Line files from
the US Census have been downloaded for Los Angeles and Orange Counties, and included in the database
as well. Thus using only the data available in the Tracer database, one can generate a map of the local
area and plot recorded GPS points on it. The Tracer website uses a large Java library to access and
manipulate the data.
The primary elements of the Tracer database and website has been up and running for approximately one
year. Additional data visualization interfaces are being developed as needed and as time permits. Taking a
broad view of the system, there are two aspects to the Tracer system: inventory control, and GPS data
ATMS Testbed Report: Executive Summary October 2006
22
management. Both of these are transferable to other projects. First, the inventory control aspect of Tracer
could be used to manage any circulating Testbed resource. The database schema, described in detail in the
project final report ( Report TTR3- 08), links Testbed researchers with Tracer devices through intermediate
join tables. This approach can be used with any other resource, while preserving the core database and
Java code that manages user log in, device use, and so on.
While the details of managing the highly configurable EDCUs
cannot be directly transferred to another device, the approach
can serve as a model, which may simplify implementation.
Second, the GPS data management aspect of Tracer is directly
transferable to any other geocoded data resource. The Tracer
system has to cope with moving devices, and track data across
time, space, and across different devices for different users.
Other geocoded resources, such as the data feeds from
embedded loop detectors, can be managed using a subset of the
EDCU Java code and database tables since there would be no
need to track the device itself beyond its initial placement.
As was stated above, data visualization enhancements to the
website are being added as time permits and research needs
dictate. One recent enhancement has been to leverage the
public API of Google Maps to display the collected GPS data. The figure shows a recent screen- shot,
with the GPS data being plotted on top of USGS aerial orthoimagery. In addition, the core Java code and
geographic tables are being used to track data from inductance loop detectors throughout Orange County.
Since the locations of loop detectors and the positions of vehicles are being tracked in the same database,
when this work is complete it will be easy to compare the two sources of data. One goal is to use this
geographic data fusion to enable Caltrans staff to identify loop detectors that need recalibration in order to
accurately report traffic characteristics.
Tracer “ Google Map” Display
ATMS Testbed Report: Executive Summary October 2006
23
ATMS Testbed Resource Website
( www. atmstestbed. net)
Summary
• What is it?
– An interactive real- time graphical user interface for obtaining traffic information from the
Detector Testbed, area freeways and arterial, as well as access to historical databases
using a browser for query.
• What does it consist of?
– A “ virtual gateway” both to the traffic data streams ( real- time and historical) generated in
the Testbed as well as the analysis tools ( e. g., Paramics APIs) developed by Testbed
researchers.
• How has it been used?
– Has provided Caltrans and university researchers with virtual access to the full
complement of services available through the Testbed.
• How can it be used?
– Run experiments, retrieve data, test and evaluate ATMS components within the Testbed
from virtually any location without requiring physical proximity to the Testbed.
Description
Access to the resources of the Testbed are provided through the ATMS Testbed website. The website has
a real- time graphical user interface for obtaining traffic information from the Detector Testbed, area
freeways and arterial, as well as access to historical databases using a browser for query. The website
was developed and is maintained by the Testbed
staff.
The ATMS Testbed website can provide
university researchers, Caltrans staff, and public
and private sector partners with a “ virtual
gateway” both to the traffic data streams ( real-time
and historical) generated in the Testbed as
well as the analysis tools ( e. g., Paramics APIs)
developed by Testbed researchers. Examples of
some of the data currently available on the
Testbed website are given below.
The ATMS Testbed website maintains a rich
complement of both real- time and historical data
generated within the Testbed that are being
made available to researchers. ATMS Testbed Website Homepage
ATMS Testbed Report: Executive Summary October 2006
24
Through the website, members ( researchers,
Caltrans, public and private sector partners) can
obtain access to the various resources
maintained by the Testbed, under varying levels
of access privilege. For example, information
pertaining to the Traffic
Detector and Surveillance Sub- Testbed ( TDS2)
can be accessed through a “ pull- down” menu
from which either historical data, either raw or
statistical can be retrieved.
Utilizing vehicle re- identification systems
developed jointly by PATH and the Testbed, real- time section data can be obtained both for arterial and
freeway networks within the Testbed.
The full complement of resources ( e. g., data
streams, streaming video, software packages)
available at the Testbed are incorporated in the
ATMS Testbed web page in a manner that
provides “ virtual” access to these resources.
The web page has a real- time graphical user
interface for Orange County freeways and traffic
information from the cities of Irvine and
Anaheim, access to historical database using a
browser for query, and IP streaming video. The
web page is available on the Testbed network as
well as via the Caltrans Wide Area Network.
Traffic Detector Sub- Testbed
Historical data
Advanced Loop Detector Locations
ATMS Testbed Report: Executive Summary October 2006
25
PART 3
TESTBED RESEARCH AND DEVELOPMENT
ATMS Testbed Report: Executive Summary October 2006
26
Development and Deployment of Corridor Management Prototype
Why was this research undertaken?
A central Advanced Transportation Management Information System ( ATMIS) capability is a timely and
efficient response to non- recurring congestion. The complexity of traffic in urban corridors requires
substantial interaction between the various agencies that share responsibilities for corridor management.
Coordinated response to congestion phenomena between these agencies avoids the implementation of
responses that may be conflicting and therefore counter- productive.
The problems of transportation management in urban areas are complicated by jurisdictional as well as
operational problems. The spatial and administrative organization of transportation management agencies
in metropolitan networks requires a coordinated solution effort that preserves the different levels of
authority, guarantees privileged data control, and in general reflects the inherent distribution of the
decision- making power. A coordinated response to congestion avoids the implementation of operational
solutions that may otherwise conflict, and therefore be counterproductive.
What was done?
The Testbed functionality in this broadly defined ATMIS application is centered on the real- time multi-agent
incident management system CARTESIUS ( Coordinated Adaptive Real- Time Expert System for
Incident management in Urban Systems), deployed within the Testbed Real- time Integrated Control and
Evaluation Prototype System ( TRICEPS). CARTESIUS approaches this problem by employing advanced
cooperation and conflict resolution methodologies for coordinated traffic management operations among
multiple agents. This system is at the cutting edge
of the application of agent technology to traffic
management and has been tested extensively using
laboratory simulation with positive outcomes
reported across the scenarios evaluated. To date,
however, it has not been deployed in the field.
In an ongoing joint Testbed/ PATH project,
TRICEPS/ CARTESIUS is in the process of being
field tested in two evaluation modes. In the first
mode, the system processes real- time data coming
from sensors in the field and provides advisory
management strategies and control actions for the
consideration of Caltrans District 12 ( Caltrans
D12) and City of Irvine Traffic Management
Center ( ITRAC) personnel. The second evaluation mode involves developing CARTESIUS as a client
under CTNET, and deploying the system in its real- time, interactive, mode in the TDS2 corridor. The
deployment will utilize dynamic O- D patterns and travel times derived from the vehicle reidentification
research ( REID), which is a separate ongoing PATH research effort, to assess alternative routing of
vehicles through the corridor. Signal timing recommendations selected with the aid of
TRICEPS/ CARTESIUS will be implemented via the Cartesius/ CTNET system that will be deployed by the
Testbed along Alton Parkway; ramp meter settings will also be implemented based on enhancements to
the Caltrans D12 Testbed intertie as part of the follow- on Testbed contract.
CARTESIUS
Distributed
Paramics
Simulation
Optimizing
Control
Path- based
Demand
Estimator
Advanced
Detector
Technology
Delay estimator
Demand estimator Problem detection
Control module
FA/ C search
CARTESIUS Components
ATMS Testbed Report: Executive Summary October 2006
27
What can be concluded?
The product of this work will be deployment of a multi- jurisdictional, multi- agent traffic management
decision support system using an extensible implementation architecture. The deployment will utilize the
products of current and prior Testbed research and evaluate their use in practical settings. The field
deployment of TRICEPS/ CARTESIUS will produce a functional multi- agent, multi- jurisdictional traffic
management system for the real- world transportation corridor that is part of the Testbed. This system will
coordinate management strategies between Caltrans D12 and the City of Irvine. The TRICEPS/ CARTESIUS
system is a mature research project
utilizing agent technology to implement
strategies that “ coordinate freeway and
arterial operations among multiple
agencies.” The work conducted here
will produce a pilot deployment of this
system in a specific real- world setting.
This will permit further evaluation of
both the performance of the architecture
as well as provide feedback on the
practical usability of the system from
the operators' perspectives.
What do the Researchers recommend?
Because CARTESIUS will be deployed in the Irvine I- 405 transportation corridor, it offers the potential to
conduct study in the following important areas:
1. Assessments of impacts of application of technologies: The impact of TRICEPS and CARTESIUS on
various non- recurrent congestion scenarios in Irvine corridor will be assessed.
2. Studies of institutional coordination issues: The efficacy of CARTESIUS mediated inter-jurisdictional
coordination of traffic management strategies between the City of Irvine and
Caltrans D12 will be evaluated.
3. Post- deployment evaluations of effectiveness: While full evaluation of the deployment will likely
require longer- term calibration of the system and subsequent monitoring, preliminary evaluations
of its performance will be produced and used to feed further fine- tuning of the deployment.
Contact Information
Dr. Michael G. McNally ( e- mail: mmcnally@ uci. edu)
Dr. Craig Rindt ( e- mail: crindt@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
Incident: blocked lanes
Caltrans Freeway Agent
Divert flow around incident
Adjust metering rates upstream
Monitor problem until resolved
Municipal Arterial Agent
Divert freeway traffic back to freeway
Adjust signals to handle diversion
Monitor problem until resolved
CARTESIUS Operational Scenario
ATMS Testbed Report: Executive Summary October 2006
28
Distributed Paramics Development
Why was this research undertaken?
Micro- simulation modeling is an increasingly popular and effective tool for analyzing a wide variety of
dynamical problems, which are not amendable to study by other means. As a suite of ITS- capable, user-programmable,
high- performance microscopic traffic simulation package, PARAMICS offers very
plausible detailed modeling for many components of an ‘ ideal’ simulator. It has become a widely used
microscopic simulation model in the US ( especially in California). However, for it to be incorporated
effectively for real- time applications involving California’s urban freeway networks, it needs to have
scalability, so that these large networks can be simulated in something close to ( or, even exceeding) real
time.
What was done?
Research is being conducted under the auspices of both PATH and the Testbed to develop a distributed
version of Paramics that will be truly scalable. In this project, we developed a distributed modeling
framework for large- scale microscopic traffic simulation and implemented a scheme with global control
and independent subnets using the PARAMICS model. Unlike the previous studies using the dedicated
high performance machines, our efforts are to utilize the low- cost networked PCs that are commonly
available. By using the Application Programming Interface ( API) functions supported by off- the- shelf
PARAMICS, we are able to simulate the traffic in a large region simultaneously with independent subnets
running on separate desktop PCs and vehicle transferring from one subnet to another.
In the distributed simulation environment being developed, the targeted large network is divided into sub-networks,
and each sub- network is simulated on a separate desktop PC. The general distributed
architecture includes: 1) a “ controller” simulator running the " master network," and 2) several sub-network
simulators. Although the controller may
have various tasks related to coordinating the
traffic simulation itself, the essential task from a
computational architecture standpoint is the
synchronization of the time in each sub- network,
either at every simulation time- step or at specified
time intervals. To synchronize the simulation
time, the controller has the ability to start and stop
the sub- network simulation at any time. In
addition, such information as boundary zones and
their corresponding ownerships are established in
the controller computer.
During a simulation run, the controller and
simulators communicate over the distributed
platform. The sub- network simulators act as
slaves to the controller. During a time step of
simulation or certain time interval, a simulator
executes a non- blocking loop ( asynchronous
communication) while waiting for a new request
from the controller. A request is simply a message
associated with a specific task. When the request arrives into a sub- network simulator, it starts with an
execution of the corresponding sequential code. When the request task is completed, a notification is sent
back to the controller. When all simulators are “ checked in”, the simulation master clock advances by
Distributed Paramics Schematic
ATMS Testbed Report: Executive Summary October 2006
29
one step and broadcasts the new times to every simulator in the system. Each simulator then proceeds
until it reaches the master clock time. A pictorial description of the scheme used for distributed
processing is shown in the figure. The communication between the simulators and controller is through
the CORBA distributed platform.
What can be concluded?
Performance testing and analysis of the implemented prototype demonstrate that the proposed framework
is very promising. Since synchronous communication is used among simulators and controller, each
simulator can only run as fast as the slowest one; proper and balanced decomposition of the network is
critical to the overall performance. Because the total computational requirement for a microscopic traffic
simulation is dominated by the number of vehicles in the network at any time, the ideal division of
network is to create N regions that each has exactly V/ N vehicles, where V is total number of vehicles in
the simulation and N is the target number of processors. The speed- up performance of the simulation in
distributed processing is also dependent on the communication to computation overhead: if there are a
large number of communication operations for each computational operation, the overall process will
reduce in speed. In order to minimize the communication to computation overhead, distributed
simulations require methodological decomposition of the large network to find a subdivision where there
are as few boundaries as possible and the computational load is spread evenly across the processors.
In the design, not only does the controller synchronize the time clock of simulators but it also manages
the global abstract network, global O- D matrix and global routing table. The benefit from this design is
that vehicle’s origin- destination and its path are all controlled at the global level, as opposed to the local
level. In this aspect, the simulator’s design is similar to the simulation over single processors in terms of
routing, with the distinction of updating vehicle’s location over distributed processors. But the
communication load between the controller and simulators is also significantly higher than that of the
light controller- heavy simulator, which may slow down the simulation.
The controller has the global abstract network, global O- D matrix and global routing table. Each
simulator also has its own local routing table. When a vehicle is generated in the sub- network, if its
origin and destination belong to different sub- network, its temporary destination in the originating sub-network
will be determined from the global routing table, but its path in the originating sub- network will
be determined locally from the local routing table. Instead of routing every individual vehicle at the
global level, the design allows a vehicle’s route calculated at the local level, and significant
communication overhead is reduced.
What do the Researchers recommend?
The platform should be tested and on several large networks. For early testing, it is recommended that the
large Orange County Testbed network recently fine- tuned at UCI be used. This network is one of the
largest networks coded in Paramics, with as many as 100,000 vehicles being present at any given time.
Contact Information
Dr. R. Jayakrishnan ( e- mail: rjayakri@ uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
ATMS Testbed Report: Executive Summary October 2006
30
Development of the Capability- enhanced PARAMICS Simulation Environment
Why was this research undertaken?
Micro- simulation modeling is an increasingly popular and effective tool for analyzing a wide variety of
dynamical problems, which are not amendable to study by other means. As a suite of ITS- capable, user-programmable,
high- performance microscopic traffic simulation package, PARAMICS offers very plausible
detailed modeling for many components of an ‘ ideal’ simulator. It has become a widely used microscopic
simulation model in the US ( especially in California). Its typical applications are the evaluation of
different traffic control and management strategies. To be qualified as an appropriate simulator for these
studies, Paramics should have the capabilities to model the real- world traffic condition and various
operational strategies. However, PARAMICS does have some functional deficiencies. For example,
Paramics can basically model the fixed- time signal control but not the commonly used actuated signal
control. Using the powerful API programming ability of Paramics, this project aimed to complement its
functionality in signal control and enhance its capabilities in the modeling of ITS strategies.
What was done?
This project developed a capability- enhanced PARAMICS simulation environment through integrating
ten plug- in modules implemented in Paramics API. These ten plug- ins included actuated signal, multiple
actuated signal timing plan, actuated signal coordination, detector data aggregator, ramp metering control,
on- ramp queue override control, ALINEA ramp metering control, BOTTLENECK ramp metering
control, SWARM Ramp metering control, and Freeway MOE ( Measure of Effectiveness). They
complemented the current Paramics simulation model and enhanced its functionalities.
What can be concluded?
Our tests using the capability- enhanced PARAMICS simulation environment show that the commercial
PARAMICS model functionalities can be effectively complemented and enhanced through API
programming. The enhanced PARAMICS simulation can better model and evaluate ITS.
Our experiences also show that API can be used to access the core models of a micro- simulator and
potentially, researchers can use commercial micro- simulators as a shell for testing their own models and
algorithms. Since other commercial micro- simulators, such as VISSIM and AIMSUN 2, also provide
users with their own API functions, users can replicate our methods to enhance their capabilities.
What do the Researchers recommend?
These developed plugins need to be continuously maintained because the upgrade to a new version of
Paramics may cause them to work abnormally. This is because these plugins were developed based on the
core models of Paramics, which may be changed or modified in a new version.
Our current capability enhancements of Paramics only cover some aspects of the microscopic simulator.
More efforts are expected in order to make the enhanced PARAMICS better fit to more ITS- related
studies.
The ten developed plugins have been released to Caltrans for use. Since the current plugins are developed
for the research purposes, extra development for user- friendly graphical interfaces may be needed.
Contact Information
Dr. Lianyu Chu ( e- mail: lchu@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
ATMS Testbed Report: Executive Summary October 2006
31
Real- time Adaptive Systemwide Ramp Metering and Signal Control
Why was this research undertaken?
Ramp metering has been recognized as an effective freeway management strategy to avoid or ameliorate
freeway traffic congestion by controlling access to the freeway. However, studies show that significant
benefits can be obtained from ramp metering only when implemented correctly and operated effectively.
In order to ensure the success of the implementation, it is important to first investigate during a pre-implementation
phase questions related to whether or not ramp metering is warranted, which kind of ramp
metering algorithm is suitable to the peculiarities of the target network, how to implement a specific
metering algorithm, and how to setup the parameters of the algorithm. With the advancement of
technology, more complicated metering algorithms, such as System Wide Adaptive Ramp Metering
( SWARM), have been ( and will be widely) implemented in the real world. These algorithms have more
parameters and depend on accurate detector data. Their correct implementation and performance
optimization depend on greater expertise from operators and more research on the part of researchers.
What was done?
In support of Caltrans’ ramp metering deployment efforts, an evaluation platform based on PARAMICS
simulation is being developed jointly by the Testbed and PATH. The ramp metering design tool handles
how to implement a metering algorithm within a target freeway corridor. The ramp metering evaluation
tool employs the testing and evaluation of a metering
algorithm and its design based on performance
measures obtained from simulation. A library of
metering algorithms either currently or potentially
applied in California is included in this platform.
These algorithms include District 3, 6, 8, and 11’ s
SDRMS, and District 7 and 12’ s SATMS, as well as
some adaptive metering strategies ( e. g., ALINEA, and
SWARM) that potentially can be applied. The
platform has intuitive graphical interfaces in order to
facilitate Caltrans practitioners.
The overall framework of the platform is shown in the
figure. The input data of PARAMICS simulation are
network geometries, traffic control data, and traffic
demand data estimated from real- world loop data. The
platform has four main modules, and each module is
made up of several components. The components
within the red dotted box represent the newly-developed
modules. The light green module is the core
module, including two functional tools, design and
evaluation. The bright green module is the graphical
interface module through which users can access these
tools. The blue module is metering algorithm module
that consists of a library of Caltrans’ metering
algorithms ( implemented as PARAMICS plug- ins).
The pink module is the supporting module, which
includes several PARAMICS API plug- ins used to support various metering algorithms and measurement
data collected from the simulation world.
Evaluation Design
SJRMS ALINEA
SDRMS
SATMS SWARM
Override
Strategies
MOE Loop Ramp
PARAMICS Simulation
Network Traffic control Loop data
Graphical Interfaces ( GUI)
URMS
Evaluation Framework
ATMS Testbed Report: Executive Summary October 2006
32
The latest web- based programming technologies have been used to implement the platform; the web
pages containing the GUI have been developed using HTML and JavaScript; the core module of the
platform is implemented in Java language, which has strong capabilities of the development of graphical
interfaces; XML ( eXtensible Markup Language) is used for data exchange between the core module and
two PARAMICS plug- ins modules, including metering algorithm module and supporting modules.
A series of performance measures for evaluating various aspects of ramp metering can be extracted,
including the following time- dependent measures:
1. Average mainline travel speed / travel time and its standard deviation
2. Total delay at an on- ramp
3. On- ramp queue length
4. Time percentage of queue spillback to the local streets at an on- ramp
5. Travel time from an on- ramp to downstream end of freeway and its standard deviation
The evaluation tool of the ramp- metering platform employs the evaluation of the metering design of a
metering algorithm, and various traffic operational aspects of the ramp metering control. It links a specific
ramp metering study, such as a study of freeway mainline efficiency, with required performance
measures. Examples of these ramp- metering studies include freeway mainline efficiency, on- ramp delay,
and equity analysis of a metered freeway corridor. As a result, the applications of the evaluation tool are:
1. Investigate whether metering has been operated correctly and efficiently
2. Analyze, evaluate and improve the current metering operations
3. Test new algorithm and fine- tune parameters
These studies can be done based on trial- and- error method and selected performance measures. Through
the GUI, users can dynamically observe the performances of various aspects of metering control.
What can be concluded?
Compared to field tests, the platform should provide a quick and cost- effective way to conduct ramp-metering
studies in the microscopic simulation environment. The platform can be used to analyze and
improve current metering operations, to test a new algorithm, to fine- tune parameters of an algorithm, to
evaluate various aspects of performance of a metering strategy, and to conduct a series of deeper level
analysis of the performance of the algorithm.
What do the researchers recommend?
The platform can serve as a training environment for Caltrans personnel to gain experience with various
metering algorithms, especially those coordinated metering algorithms that are characterized by many
parameters and complicated control logic. This platform should be used to guide Caltrans personnel on
how to successfully manipulate the various aspects of the ramp- metering systems, including initializing
parameters, fine tuning of parameters, performance analyses, and hypothetical " what if" simulated testing.
Contact Information
Dr. Lianyu Chu ( e- mail: lchu@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
ATMS Testbed Report: Executive Summary October 2006
33
Performance Evaluation of Adaptive Ramp Metering Algorithms
Why was this research undertaken?
Adaptive ramp metering has undergone significant theoretical developments in recent years. However, the
applicability and potential effectiveness of such algorithms depend on a number of complex factors that
are best investigated during a planning phase prior to any decision on its implementation. A field
operational test is one way to evaluate metering operation. The method can provide the ultimate fair
evaluation of the actual performance of a ramp metering system as applied in the real world, but they are
usually very expensive and time- consuming to conduct. Moreover, without prior testing, such tests have
the potential to adversely impact traffic conditions, and such uncontrollable factors as incidents or
variations of demand patterns often make it difficult to fairly compare algorithms or even before- after
results. As an alternative, this project studied a microscopic simulation based evaluation method.
What was done?
Three well- known adaptive ramp- metering algorithms, ALINEA, BOTTLENECK and ZONE were
selected for evaluation. ALINEA is a local feedback control algorithm and the other two are area- wide
coordinated algorithms. Paramics was adopted as the simulation platform for further evaluation of these
metering algorithms. Several Paramics plugin modules, including loop detector data aggregation ( used for
on- line data collection), ramp metering control ( used to mimic ramp signal operations), and ramp
metering algorithms ( metering logic implementations of ALINEA, BOTTLENECK and ZONE), were
developed to build a simulation based ramp metering evaluation framework. The evaluation was then
conducted over a stretch of the I- 405 freeway in California. The performances of these metering
algorithms were compared with respect to different demand patterns under both recurrent congestion and
incident scenarios.
As an alternative to the difficult task of fine- tuning them in real- world testing, a hybrid Genetic
Algorithms ( GA) simulation method was developed to optimize four operational parameters of a
particularly effective algorithm, ALINEA, including the update cycle of the metering rate, a constant
regulator, the location and the desired occupancy of the downstream detector station. In the hybrid
method, GA- simulation was used for parameter optimization, and micro- simulation ( i. e. Paramics) was
used for performance evaluation. The objective of the study was to achieve system optimum for all
metering freeways and ramps. Simulation results showed that the genetic algorithm was able to find a set
of parameter values that can optimize the performance of the ALINEA algorithm for the whole controlled
system.
What can be concluded?
Simulation results showed that adaptive ramp- metering algorithms can reduce freeway congestion
effectively compared to the fixed- time control. ALINEA shows good performance under both recurrent
and non- recurrent congestion scenarios. BOTTLENECK and ZONE can be improved by replacing their
native local occupancy control algorithms with ALINEA. Compared to ALINEA, the revised
BOTTLENECK and ZONE algorithms using ALINEA as the local control algorithm are found to be
more efficient in reducing traffic congestion than ALINEA alone. The revised BOTTLENECK algorithm
performs robustly under all scenarios. The results also indicated that ramp metering becomes less
effective when traffic experiences severe congestion under incident scenarios.
Since our simulation network does not contain arterial routes, traffic diversion to alternative routes is not
considered and thus the performance improvement through ramp metering control is not fully revealed.
Ideally, one should consider a corridor network and integrate a variety of control measures, including
ramp metering, traffic diversion, and signal timing, to combat traffic congestion. We should also note that
ATMS Testbed Report: Executive Summary October 2006
34
all of the algorithms evaluated in this study are reactive, rather than proactive, control strategies.
Algorithms with state estimation and/ or OD prediction capabilities are desirable. The development and
evaluation of these integrated control strategies will be left to future studies.
Because of the good performance and easy implementation of ALINEA metering algorithm, ALINEA
was identified as a good algorithm that can be potentially implemented in the field. The ALINEA
algorithm was implemented together with a typical queue override strategy in Paramics as a plugin. The
queue override strategy was used to avoid vehicles to spillback to arterials during ALINEA operation
with low metering rate.
It was found that the system performance is not sensitive to the variation of the constant regulator under
ALINEA control. When the updated cycle ranges between 30 to 60 seconds, mainline detector is placed
between 120~ 140 meters downstream of the on- ramp nose, and the desired occupancy is set to 19% to
21%, the ALINEA control produces the best system performance with most stability in our testing
network.
The desired occupancy is the most sensitive parameter among all four selected operational parameters in
our study. Choosing a suitable value for the desired occupancy was essential to optimize the performance
of ALINEA control. Simulation results show that the occupancy at capacity at the mainline downstream
detector station is the best value to be selected. If less on- ramp delay is expected from metering control, a
higher desired occupancy ( around 30%) could be used.
What do the researchers recommend?
This study showed that micro- simulation with genetic algorithm can be used to calibrate and optimize the
operational parameters of ramp metering control algorithms, as a necessary part of successful
implementation. Potentially, micro- simulation and genetic algorithm may also be used to fine- tune
various other ITS strategies.
The study provides a parameter setting guideline for the ALINEA ramp metering control. Practitioners
can use the recommended parameter values as a basic operational reference when they setup ALINEA
control in the field.
Contact Information
Dr. Lianyu Chu ( e- mail: lchu@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
ATMS Testbed Report: Executive Summary October 2006
35
Peer- to- Peer ( P2P) Traffic Information Propagation
Why was this research undertaken?
Since the introduction of intelligent transportation systems ( ITS) in the early 1990s, there has been
growing interest in potential applications of the use of wireless communication between vehicles, usually
referred to as inter- vehicle communication ( IVC). In IVC- based system, vehicles are envisioned to
exchange precise position information from satellite navigation data ( GPS) via IVC at low cost to
optimize traffic flows and provide valuable, real- time traffic information to the drivers. Our focus is on
the potential for the concept of IVC- based systems to serve aspects of transportation systems
management. We attempt to give answers to two questions: ( 1) What IVC equipment penetration rate is
needed for information propagation to extend to a substantial part of the whole network under various
traffic conditions?, and ( 2) What IVC system requirement is needed to disseminate incident information
faster than the attendant traffic/ vehicle wave propagates through the network?
What was done?
This project investigated the feasibility of a distributed traffic
information system based on IVC technology, focusing on
such systems arising principally from the transportation
application. Specifically, via simulation modeling techniques,
we determined the thresholds for some of the parameters
necessary to support the systematic collection and provision of
useful and in time ( real- time or close to real- time) traffic
information in a self- organized, distributed traffic information
system, dubbed Autonet, that is based upon the peer- to- peer
information exchange among vehicles.
Potential applications of this Autonet concept for traffic
management and traveler information are intrinsically based on
achieving information propagation throughout the traffic
network; however, because penetration of the necessary
technology to the fleet of vehicles can be expected to be
gradual, a " mixed" network of IVC- capable vehicles and non-
IVC capable vehicles will exist for some period of time. In our
research, we attempt to give answers to two questions: ( 1)
What IVC equipment penetration rate is needed for
information propagation to extend to a substantial part of the whole network under various traffic
conditions, and ( 2) What IVC system requirement is needed to disseminate incident information faster
than the attendant traffic/ vehicle wave propagates through the network?
We analyzed the two issues mentioned above for various possible IVC technologies and for different
roadway network formats and different road traffic conditions. In addressing these open questions that
may be barriers for implementation of self- organizing, IVC- based traffic information systems, our focus
is on the potential for the concept of IVC- based systems to serve aspects of transportation systems
management; the traffic- oriented abstraction evaluation framework is developed without detailed
electronic engineering and computer science modeling. Nonetheless, the results serve to identify, at least
roughly, the system implementation requirements for both software and hardware sides as well.
We used PARAMICS ( PARAllel MICroscopic Simulator) to build our simulation modeling framework.
Three IVC scenarios were investigated in the research: uni- directional and bi- directional information
Peer- to- Peer Schematic
ATMS Testbed Report: Executive Summary October 2006
36
propagation in one- dimensional highway network, and information propagation in two- dimensional
arterial streets networks.
There are three major outputs from the simulation that were used in our analysis: 1) IVC success
probability, representing the average chance that an individual IVC- equipped vehicle can find other IVC
vehicles in the communication radius range and communicate successfully in the traffic network at any
particular time; 2) communication bandwidth, indicating the average maximum amount of data that needs
to be transmitted by each IVC vehicle in the traffic network, defining the basic requirement for the
software and hardware implementation in the proposed system; 3) maximum information propagation
distance, at any time an indicator of how fast the information flow is traveling in the traffic network, a
key factor in determining whether or not this information flow may potentially benefit the traffic system.
What can be concluded?
Our results indicate that it may be extremely difficult to evolve the proposed self- organizing vehicle- to-vehicle
based system to support information propagation for location sensitive, real- time traffic
information in freeway networks in which communication is only among vehicles moving in the same
direction, especially if the IVC equipment market penetration rate is low and communication radius range
is short – two conditions that are likely to characterize the proposed system in its start- up period. Under
incident conditions, for such market conditions and available IVC technologies, the incident information
wave generally travels slower than does the traffic shock wave due to the incident.
Compared to information propagation employing information exchange only among vehicles in the same
stream of traffic, efficiencies and effectiveness of bi- directional propagation systems built upon inter-vehicle
communication technology are significantly easier to achieve because mechanisms for
information propagation in this case not only include “ hopping” along vehicles moving in the same
direction of flow as well as “ cross transference” of information to vehicles moving in the opposite
direction. For market conditions and available IVC technologies that are likely to prevail during the
system’s start- up period, the incident information wave generally travels faster than the traffic shock
wave due to the incident freeway networks in which vehicles that are moving in opposite directions in
close proximity to each other can exchange information.
Traffic information dissemination in two- dimensional urban arterial networks via information exchange
among IVC- equipped vehicles is also easier to achieve than in one- direction freeway network cases;
however, propagation speed is generally slower than in two- direction freeway network cases.
Bandwidth/ data rate requirements for IVC in urban arterial streets are relatively high because of the
representation of the complex network configurations and high density of vehicles in the traffic network
due to the distribution of vehicles in two- dimensional space.
What do the researchers recommend?
This study showed that a traffic information/ management system based on P2P communication is
feasible. Significant additional study should be undertaken to identify traffic management efficacy,
limitations and hardware requirements.
Contact Information
Dr. Will Recker ( e- mail: wwrecker@ uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
ATMS Testbed Report: Executive Summary October 2006
37
Dynamic Vehicle Navigation in a Peer- to- Peer ( P2P) System
Why was this research undertaken?
Although inter- vehicle communication and information dissemination in traffic networks have been
modeled both from various academic perspectives ( transportation engineering, electronic engineering and
computer science) and at different levels ( highly abstract, software protocols and hardware products), no
efforts have been found that systematically model and test such a distributed traffic information system
based upon peer- to- peer information exchange relative to drivers’ dynamic on- line routing behaviors
within it. This work focuses on detailed modeling of a self- organizing, distributed traffic information
system built upon vehicle- to- vehicle information exchange, with case testing of the pre- trip route- choice
and in- trip re- route behaviors of drivers with access to traffic information from the proposed information
system.
What was done?
This project focused on detailed modeling of a self- organizing, distributed traffic information system built
upon vehicle- to- vehicle information exchange, with case testing of the pre- trip route- choice and in- trip re-route
behaviors of drivers with access to traffic information from the proposed information system. Two
different large- scale traffic networks, one comprising grid arterial streets and the other a freeway corridor,
are tested with respect to different assumptions regarding drivers’ route choice behavior including both
pre- trip route choice and in- trip re- route behaviors, and different levels of knowledge of daily recurrent
traffic patterns. Potential benefits arising from this proposed information system both for travelers with
IVC and for the whole traffic systems, including all travelers with and without equipment, are
demonstrated based on simulation study results.
We used the PARAMICS simulation software to build our simulation modeling framework. Within the
micro- simulation modeling framework, some vehicles in the traffic network, equipped with IVC systems,
geographic information systems ( GIS), global positioning systems ( GPS), on- board navigation systems,
and in- vehicle computing processors, are assumed to generate floating car data information based on their
own experiences, exchange traffic information through peer- to- peer communications, and process
incoming traffic information in real- time using their on- board processors. In addition, each vehicle within
this distributed traffic information system is assumed to optimize its personal route based on its
estimation of current traffic conditions obtained from real- time traffic information propagated in the
information network and its understanding of recurrent traffic pattern from its historical traffic
information database; based on the assumption that each driver is a rational entity, re- routing decisions
are examined. Path- based vehicle navigation is implemented
for each driver, in which IVC vehicles follow changeable
paths; no- IVC vehicles follow no- changeable paths.
The first test case concerned non- recurrent congestion
scenarios in a grid network. The 5,000m x 5,000m network
evaluated consists of equally spaced two- lane local street
roadways with speed limit of 45 mph; the distance between
any two neighboring signalized intersections is 1 km. An
incident is assumed to have occurred on a link close to the
center of the study grid network. This incident is assumed to
cause passing vehicles to reduce speed to 5 mph in the
direction of the roadway in which the incident occurs and to 10
mph in the opposite direction of the roadway ( due to
speculator slowing). Three levels of O/ D demand were used in
5,000 m
5,000 m incident
Network Schematic
ATMS Testbed Report: Executive Summary October 2006
38
our simulation studies to generate light, moderate and heavy traffic flow conditions in the network.
The second test case considered non- recurrent
congestion scenarios along a freeway corridor with a
neighboring alternative arterial street. The freeway
and parallel arterial street are connected by other
arterial streets running perpendicular to the freeway
and freeway on/ off ramps spaced at 2000- meter
intervals. In the simulation, an incident 4000 meters
from the far end of the network occurs 30 minutes
into the simulation and lasts for 30 minutes; owing to
the blockage caused by the incident, vehicle speed in
that direction is reduced to 5 mph near incident
location. Three levels of O/ D demand are used to
generate light, moderate and heavy traffic flow conditions, both for the freeway and the arterials.
What can be concluded?
Several interesting results are obtained from our simulation studies under non- recurrent congestion
scenarios in the grid arterial streets network. First, only when IVC market penetration rate is higher than
some threshold values, can IVC- capable vehicles make sufficiently accurate estimations of real- time
traffic conditions to take re- routing actions. This threshold value is relatively high compared to freeway
networks due to the dimensional characteristics of grid arterial networks compared to freeways. Although
the relative travel time saving benefits for IVC- capable vehicles from re- routing decrease after reaching
maximum values, IVC- capable vehicles always maintain an advantage compared to their counterparts and
total system performance improves by their re- routing.
For most cases in the freeway corridor network, as increasing numbers of drivers have accessibility to
real- time traffic information from inter- vehicle information exchange, the redistribution of IVC- capable
vehicles enables the system to move toward user equilibrium. The characteristics of this freeway corridor
network are such that a traveler has significantly fewer choices than in the grid arterial streets. Under
normal traffic conditions the freeway system typically has much more capacity and a better level of
service than its surface street alternative, leading to many more vehicles choosing to use the freeway
system under these conditions; thus, even a relatively small amount of IVC- capable vehicles’ re- routing
from the freeway to arterial streets may dramatically worsen the traffic conditions on the arterial streets,
resulting in a general worsening of system performance under heavy demand.
What do the researchers recommend?
This study showed that significant travel time savings may be possible under a traffic
information/ management system based on P2P communication. Significant additional study should be
undertaken to identify traffic management efficacy, limitations and hardware requirements.
Contact Information
Dr. Will Recker ( e- mail: wwrecker@ uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
16,000 m 4,000 m
freeway freeway
urban arterial streets
Incident
ramp
Freeway Schematic
ATMS Testbed Report: Executive Summary October 2006
39
Analytical Studies of Peer- to- Peer ( P2P) Information Propagation
Why was this research undertaken?
The rapid advance in available information technology, especially, the development of wireless
communication technologies, now makes feasible the exploration of traffic information systems that
decentralize the tasks of collecting and disseminating traffic information. Compared to centralized ATIS
systems, a decentralized system based on Inter- Vehicle Communication ( IVC) offers some significant
advantages: ( 1) the IVC components of the system require no capital investment on the part of the
transportation agencies, and can evolve to a full system gradually, once a threshold market penetration is
achieved; ( 2) both the monetary and labor costs to build and operate the system are directly distributed to
the users of the system; ( 3) the system is much more resilient to disruption, particularly in the event of
disasters, when communications, management, and control are most important; ( 4) the system can be
anchored to the Internet as a platform for additional applications.
What was done?
In this project, we developed a novel analysis framework for the performance of inter- vehicle
communication in a traffic stream. With the assumption that information propagation is instantaneous
compared to vehicle movements, we consider the probability of success for information to propagate
beyond a location. Most- forward- within- range
communication chains and transmission cells were
defined, measurement probabilities were modeled
regressively, and examples were studied for different
traffic scenarios, transmission ranges, and market
penetration rates. This mathematical model was found to
be consistent with simulation- based studies.
In an effort to understand the structure of the
complicated P2P system, we first make an assumption
that information propagation through IVC is
instantaneous. This simplification is based on the
observation that vehicles’ movements are
inconsequential during the short transmission time of a
message ( in the order of one- tenth second). Therefore,
the effect of traffic dynamics is omitted, reasonably, but
the distribution pattern of vehicles caused by road
geometry and car- following rules are included.
We start by defining “ most forwarded within range” communication chains and splitting a traffic stream
into a number of cells based on the transmission range of wireless units. Then by considering the
relationship between communication chains and transmission cells, we obtain two basic components for
constructing an arbitrary communication chain. With the understanding of the structure of information
propagation, we are able to derive a regressive model for computing the success rate for information to
travel beyond a certain point.
What can be concluded?
With this model, the user can evaluate the performance of certain wireless communication units for
different market penetration rates and traffic patterns.
Figure 1: Success rates v. s. transmission ranges: penetration
rate = 10% and four- lane road with density of 56 veh/ km
ATMS Testbed Report: Executive Summary October 2006
40
For example, we evaluated the influence of different communication ranges on the information
propagation as shown in Figure 1. As expected, longer transmission ranges yield better propagation
results. The results in this figure are consistent in magnitude with those in the literature by simulation
studies. This further argues for the assumption of instantaneous information propagation.
We also applied this model to study the impact of a gap
in a traffic stream on information propagation. As shown
in Figure 2, a gap at [ 0, 0.4 km] can seriously deteriorate
the success rate ( by more than 10 percentiles) for
different locations of information sources.
In Figure 3, we show the influence of a shock wave on
information propagation. Here, the information source
vehicle runs into a congestion shock wave and slows
down ( from 104 km/ h to 97 km/ h). For a number of time
instants, we consider the propagation distances of 95%
success rate in both forward and backward directions.
We can clearly see the bending effect of traffic density
change on information propagation. This example shows
that a small change in traffic density ( 60 veh/ km vs 80
veh/ km) can have substantial consequence on
information propagation distances ( 4 km vs 27.1 km).
What do the researchers recommend?
This project demonstrated the effectiveness of the
mathematical model for evaluating the performance of
IVC in a traffic stream. Use of this model can
significantly reduce the time required for evaluating the
influence of communication ranges, market penetration
rates, and traffic pattern on the success rate of
information propagation. The model can be used for
estimation of the performance of Autonet system.
Continued efforts are needed to extend this analysis
framework for studying other important issues related to
IVC systems, such as communication capacity, and the
impact of road network topology.
Contact Information
Dr. Will Recker ( e- mail: wwrecker@ uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
Figure 3: Information progress with 95% success rate in a
shock wave
Figure 2: Success rates of information propagation for
different locations of information so Figure 1: Success rates
v. s. transmission ranges: penetration rate = 10% and four- lane
ATMS Testbed Report: Executive Summary October 2006
41
Mobile Throughput of 802.11b
Why was this research undertaken?
Simulation research has shown that a traffic information/ management system based on P2P
communication is feasible. However, significant additional study is needed to identify traffic
management efficacy, limitations and hardware requirements. In particular, the ability of current wireless
communication technology, e. g., 802.11b, to successfully transmit traffic information between vehicles
moving at high relative speeds to each other is questionable.
What was done?
The project investigated the capabilities of wireless local area networking ( WLAN) technology for
vehicle- to- vehicle and vehicle- to- roadside communication. An 802.11b ( WiFi) card is inexpensive, and
provides plenty of bandwidth. WiFi technology is being incorporated into PDAs, cellphones, and in
Intel’s Centrino laptop chipset. It requires very little imagination to suppose that drivers will soon have
easy access to a computing device of some sort, which supports WiFi or a related technology. This project
focused exclusively on vehicle- to- roadside communications. The work is part of a larger effort to
benchmark the practical throughput of off- the- shelf wireless technologies. The work is on- going, with
further vehicle- to- roadside and vehicle- to- vehicle tests planned.
The first step in testing wireless throughput was to install hardware both on the roadside and in a vehicle.
Two Cisco AP 350 series wireless access points were mounted on buildings on the UC Irvine campus
fronting Peltason Road. The access points were connected to building- mounted 13.9dBd antennas. The
antenna specifications are as follows: 3dB Beamwidth, Degrees E- Plane 30, 3dB Beamwidth, Degrees H-Plane
34, Front to Back Ratio, dB 18, Gain dBd 13.9, and Impedance ( ohms) 50.
Inside of the vehicle, wireless connectivity was achieved using a Lucent Orinoco Silver PCMCIA card
plugged into a laptop computer. A 5db gain omni directional external antenna was mounted on the vehicle
roof using a magnetic base. The laptop was also connected to the Testbed’s extensible data collection unit
( EDCU), in order to obtain exact positioning for the wireless signal readings.
The primary study area is shown in the
accompanying figure. One antenna was mounted
on the Multipurpose Science and Technology
( MST) building, shown in the center of the figure,
and the other was mounted on the Engineering
Gateway ( EG) building, just to the right of the
MST building. Both of the directional antennas are
pointed in a westerly direction along Peltason
Road, more or less towards the crossing roadways
on the left side of the figure. In order to test the
wireless link capacity, the utility program Netperf was used. Netperf is designed to test various aspects of
a network connection. In this case, it was configured to repeatedly conduct throughput tests with one-second
duration. This resulted in one geocoded throughput measurement every two seconds.
Data were collected over several days. The maximum throughput observed was 4.62 Mbps, which
compares favorably with WiFi used in an indoor environment. This maximum throughput value was
measured while the vehicle was traveling at 30.7 mph, approximately 159 meters from the MST building
access point. As can be seen from the figure, the connectivity is highly directional. This is for two
reasons. First, the antennas are directional, and are pointing in a westerly direction along Peltason.
Second, 802.11b requires a clear line of sight. Peltason road is quite hilly, and the roadside antennas are
Study Area
ATMS Testbed Report: Executive Summary October 2006
42
completely obscured at the intersection of Peltason and Bison. However, at the intersection of California
and Bison ( near the crossing roads), there is a clear line of sight to the MST building once again.
Some simple regression tests were done to
examine the relationship between line of
sight, speed, and distance with the
observed throughput. The regression
results indicate that distance plays the
most significant role in determining the
throughput value. If all non- zero
observations are included rather than just
those clustered in the direction that the
antenna is pointing ( 192 degrees), then the
most important factor is the “ degrees away
from 192.” Interestingly, speed is less
important than we expected. This is perhaps due to the fact that our tests were performed on streets with a
speed limit of between 35 and 45 mph.
What can be concluded?
This project demonstrated that it is possible to establish a useful, high- bandwidth link between a moving
vehicle and a roadside, directional antenna. The connection strengths were surprisingly high, with a signal
of about 4 Mbps established at an intersection approximately 1 kilometer from the nearest antenna. While
this connection was not maintained for very long due to the road geometry, at this communication rate it
is possible to download a PDF file containing the UC Irvine campus map within 1 second. The biggest
factor limiting throughput was maintaining a clear line- of- sight to the roadside antenna. As the road
curved away or dipped down behind a hill, the throughput rapidly went to zero.
The implications of these tests for possible applications are clear. If we continue to use the same
directional antennas, it is possible to conduct short transactions with moving vehicle that are in range.
These transactions might include downloading a map, or a short session requesting the nearest available
parking to an intended destination. However, transactions that require long sessions, such as attending a
lecture wirelessly or checking email, are not possible with this kind of connection.
What do the researchers recommend?
There are several aspects of the vehicle to roadside connection, which still need testing. For example, the
test protocol did not repeatedly attempt to establish a connection with the base station. That is, a single
period of testing would include a single connection event. After that, the wireless antenna would be
assigned an IP address and could concentrate on testing throughput. It would also be interesting to test the
time required to negotiate the initial connection throughout the test area. The tests were also performed
with very little ambient interference. While there are 50 or more open WiFi access points in the study
area, they are all use low power omni- directional antennas, and did not offer any significant interference.
Contact Information
Dr. James Marca ( e- mail: jmarca@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
Connectivity
ATMS Testbed Report: Executive Summary October 2006
43
Implementation Tests of a Peer- to- Peer ( P2P) Network for Incident Exchange
Why was this research undertaken?
The existing paradigms for vehicular traffic monitoring and control have a strong infrastructure bias—
data are collected centrally, processed, and then redistributed to travelers and other clients in response to
requests. There are a number of research efforts to decentralize traffic state monitoring by leveraging
advanced local area wireless technology. One version of such a traveler- centric system is called Autonet.
Work to date consists of a preliminary implementation of some of the key Autonet concepts as well as
some field measurements.
One of the key gaps limiting the broad adoption of Intelligent Transportation Systems ( ITS) technologies
by drivers is the lack of usable real- time data. According to a 2002 national survey of ITS technology
adoption [ ITS Joint Program Office, 2004], only about 30% of all signalized intersections on arterial
streets had any form of electronic surveillance. Even if every highway were fully and accurately
monitored, drivers attempting to plot an alternate route around an incident would be unable to evaluate
conditions on the arterial street network.
Automobile manufacturers have been installing more and more computer electronics and control systems
into their cars, with high end models commonly sporting in- dash, GPS- based mapping systems. The
contrast between the high cost and low penetration rate of infrastructure based traffic monitoring, and the
declining cost and increasing capabilities of in- vehicle electronics devices implies that we should explore
the abilities of the latter to augment or even replace the former. We suspect that a decentralized, vehicle-based
traffic probe system may be cheaper, faster, and easier to implement than the centralized
monitoring and traffic probe systems defined in the National ITS Architecture.
What was done?
An in- vehicle client with an informative GUI has been developed. This client continuously listens for
other clients, and exchanges knowledge about network incidents once contact is made. We demonstrated
that knowledge about traffic conditions can be propagated successfully using this system. The client
programs were also used to test the actual throughput possible for messages sent from one vehicle to
another using 802.11b wireless hardware. These measurements establish the maximum throughput at
about 4,000 incidents for two vehicles moving in opposite directions at highway speeds.
The main result of this project was to design the
basic features of an Autonet system, and to then
implement an Autonet peer that could perform the
most important of those features.
The guiding scenario is shown in the figure. This
picture shows a vehicle passing a traffic jam,
collecting information about that jam from the
affected cars, and then “ carrying” that information
upstream. At some point, the information would be
used by travelers upstream of the incident to decide
their optimal course of action, in this case, to exit the
highway. This scenario led to a specification of the
system requirements, which in turn led to a software
architecture, shown in the accompanying figure, and
an implementation.
AUTONET Scenario
ATMS Testbed Report: Executive Summary October 2006
44
The implementation was installed on a number of laptops. Each laptop was equipped with a GPS antenna
to provide location and speed information, and a
WiFi card to allow the peers to communicate with
each other. Each peer was equipped with a common
network model, based on the freely available
TIGER/ Line file data for Orange County. This
provided a common framework for exchanging
incident information. In repeated trials, we
demonstrated the ability to transmit incident
information between vehicles, and to have incident
information be carried from one vehicle to another
just as depicted in the basic scenario discussed
above.
What can be concluded?
In addition to providing a proof of concept test, the initial Autonet peer implementation served a second
purpose: the ability to benchmark the capabilities of off the shelf communications hardware and protocol
stacks. In order to explore the feasibility of the Autonet idea, the proper approach is to implement a
large- scale traffic simulation, and to slowly ratchet up the adoption rates. However, such simulations
would be difficult to program without solid
evidence about the capabilities of each wireless
client. Using the initial Autonet peers, we were
able to demonstrate the ability to transmit data in a
highway setting between two oncoming cars, each
traveling at 70 mi/ h. The results are summarized
in the accompanying table.
This project demonstrated that a decentralized,
vehicle based real- time traffic information
system is technically feasible. We designed the necessary system architecture, implemented that
architecture, and then tested our implementation on the road. We were able to transmit incidents in real
time between moving vehicles using our software and off the shelf hardware.
What do the researchers recommend?
A large question is how useful such a system will be if it is adopted by the general public. To answer this
question, the next step is to expand simulation studies using the empirical capabilities of the prototype
device as a starting point. Through simulation, it will be able to determine the relationships between
adoption rates, information flow, and system awareness, which in turn will lead to traveler- centric
applications such as route planning and incident avoidance.
Contact Information
Dr. James Marca ( e- mail: jmarca@ translab. its. uci. edu)
Institute of Transportation Studies
University of California, Irvine
Irvine, CA 92697- 3600
40 mi/ h 55 mi/ h 70 mi/ h
# of incidents exchanged 5,996 4,295 3,556
Time within range 23 s 18 s 10 s
Round trip time 71 ms 80 ms 81 ms
Implementation Schematic
Test Results
ATMS Testbed Report: Executive Summary October 2006
45
Implementation of a Tool for Measuring ITS Impacts on Freeway Safety Performance
Why was this research undertaken?
Reduced congestion and smoothed traffic flow are likely to improve safety, as well as reduce
psychological stress on drivers. Using data from the Testbed, we have begun to document the
relationship between safety and improved traffic flow. Recent developments indicate that the time is right
to refine and implement analytical tools that can be used in real- time monitoring of the safety level of the
traffic flow on any instrumented segment of freeway. As opposed to tools that measure freeway
performance in terms of throughput or travel time, the data indicate that the key elements of traffic flow
affecting safety are not only mean volume and speed, but also variations in volume and speed. We further
determined that it is important to capture variations in speed and flows separately across freeway lanes,
and that such information is useful in differentiating types of crashes.
What was done?
The objective of this joint effort between the Testbed and PATH is to implement a real- time tool for
safety analysis. The overall project goal is to calibrate and verify a tool that translates traffic flow, as
measured by ubiquitous single loop detectors, into safety performance in terms of expected numbers of
crashes by type of crash per exposed vehicle mile of travel. This tool can be used in monitoring the safety
performance of freeway operations and to evaluate and document improvements to safety arising from
such ITS deployment as system- wide ramp metering ( SWARM), freeway service patrol ( FSP) and other
incident response measures, and driver information.
In work conducted thus far, we lay the groundwork for the development of a performance tool that gauges
the level of safety of any type o