U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

 
SUMMARY REPORT
This summary report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-15-085    Date:  December 2015
Publication Number: FHWA-HRT-15-085
Date: December 2015

 

The Exploratory Advanced Research Program

Next Generation Traffic Control Systems Workshop Summary Report - February 3–4, 2015

Introduction

On February 3–4, 2015, at the Turner– Fairbank Highway Research Center in McLean, VA, the Federal Highway Administration's (FHWA's) Office of Operations Research and Development (R&D), with support from the Exploratory Advanced Research (EAR) Program, convened a workshop on "Next Generation Traffic Control Systems." The purpose of this 2-day workshop was to bring together researchers and technology developers from industry, academia, and public agencies to discuss the direction of technological advances in traffic control systems and sensors. Workshop participants discussed a different set of issues on each day of the workshop. On the first day, the participants primarily focused on the promise of advances in infrastructure-based and mobile-sensor technology to offer new and substantial capabilities in measuring traffic speeds, volumes, origin–destination pairs, and other data that enable improved traffic control. The participants identified research gaps, barriers, and needs that could be addressed to improve the utility of sensors for traffic management, particularly to enable the next generation of adaptive signal control. On the second day, workshop participants focused on the importance of researching and developing new traffic-signal control algorithms and the complex simulation infrastructure that it requires. The participants discussed how to accelerate the development of tools for future signal control research, the need for these tools, the potential benefits, and future considerations for their development and dissemination.

Day 1: Next-Generation Sensors

Opening Remarks

Advances in infrastructure-based and mobile- sensor technology promise to offer new and substantial capabilities in measuring traffic speeds, volumes, origin–destination pairs, and other data that enable improved traffic control. This first day of the workshop involved identifying research gaps, barriers, and needs that could be addressed to improve the utility of sensors for traffic management, particularly to enable the next generation of adaptive signal control.

Dr. Joseph Peters and David Gibson of FHWA provided an overview of traffic control system and sensor technology and how they relate to their work, after which they described the mission of the Office of Operations R&D and its goal to advance state-of-the-art transportation operations. They described how the specific objectives of the office are to (1) drive technology work and evaluation of concepts, (2) engage state-of-the-art professionals from industry and Ph.D. students from academia, (3) support professional development to provide the value necessary for investment, and (4) support the mission of FHWA.

David Kuehn, EAR Program Manager, then discussed the purpose of the EAR Program. Kuehn stated that the EAR Program works to understand the various roles of technology and connectivity for the future. Kuehn noted that a vision of a connected and automated future is a big part of this role, and commu nication between vehicles and infrastructure is required to take full advantage of the technology and reduce congestion. Kuehn highlighted speed harmonization as one example of a way that traffic can be controlled through connectivity to improve mobility.

Gibson explained that the motivation for this workshop is the need to move beyond contact-closure sensors (i.e., devices that detect the open or closed status of a circuit) toward those that can support vehicle identification, re-identification, and location. Traffic control systems with these types of sensors could transform how optimization algorithms work and provide new types of data and information. Research on how to develop these sensors must answer the following questions:

Next, expert speakers from industry and academia presented their research to the workshop participants. These presentations are summarized in the following pages.

How Can Advances in Science and Technology Enhance or Supplant Current Sensors or Control Systems?

Dr. Jakob Eriksson
Assistant Professor, University of Illinois, Chicago

Overview

Dr. Eriksson began his presentation with some background information on traffic sensors and introduced three categories of sensors: passive, semi-passive, and active. He then described the developments that need to occur within the sensor technology field to provide improved accuracy and more comprehensive data. Dr. Eriksson highlighted that his presentation was intended to spark a discussion about the future of traffic-sensing technologies.

Major Themes Discussed

During his presentation, Dr. Eriksson discussed a range of traffic sensors. These are summarized as follows:

Passive Sensors

Dr. Eriksson explained that passive sensors require no participation from the driver and may include inductive-loop detectors, radar, camera or computer vision, and license-plate readers. He noted that computer vision holds many promising uses for the future, including continuous turn counts and the ability to sense from arbitrary perspectives, that could overcome today's limitations caused by heavy occlusion.

Semi-Passive Sensors

Dr. Eriksson described how semi-passive sensors require participation from drivers devices without their knowledge and can include applications such as roadside Bluetooth®, Wi-Fi, tire-pressure sensor sniffers, radio-frequency identification, cellular hand-off signals, and applications on smart phones. Figure 1 illustrates how reidentifying wireless sniffers can be positioned to overhear and record unique addresses contained within transmissions from these devices. This could include the transmission of a unique address from Bluetooth or tire-pressure sensors. Dr. Eriksson stated that cellular hand-off signals are useful because they produce a lot of data; however, service providers might restrict access to the reidentification or tracking information and granularity is coarse (i.e., less detailed) and inconsistent across providers. He also noted that some of these uses have privacy issues, especially those associated with smartphone applications that report global positioning system (GPS) coordinates.

A diagram illustrates the roadside layout of reidentifying wireless sniffers. A birds-eye view of a section of highway shows two black boxes next to the road, marked sniffer 1 and sniffer 2. A blue car is shown about to pass these sniffers.

Figure 1. Roadside layout of reidentifying wireless sniffers.

Active Sensors

Dr. Eriksson noted that active sensors require that a driver participate knowingly and have a personal interest in providing data. He provided some example applications for these sensors, including fleet-tracking networks and mobile applications that allow users to report scenarios such as accidents and construction. Dr. Eriksson mentioned that there may be some legal concerns associated with these applications; however, creating an incentive for the user to provide data could prove to be useful. He suggested engaging the driver directly to establish a customer relationship, which can then be used to incentivize changes in driving patterns. Dr. Eriksson noted that a "frequent-driver" program such as this would create a large amount of data that could prove useful for many purposes. He also suggested that willing participants often lead to more accurate and reliable data.

Next Generation Sensors

Dr. Eriksson suggested that any next generation sensing technology must be statistics-based and actionable in real time. He explained that statistics-based sensors are those that can estimate congestion state, hourly-turn probabilities, and current traffic volume, speed, and vehicle mix. Real-time actionable sensors are those that are useful and can detect the presence of vehicles and pedestrians. These sensors are able to count vehicles, predict vehicle-arrival times, and detect vehicles that are obstructing upstream or downstream intersections.

In summary, Dr. Eriksson highlighted the importance of using several different types of sensors together in the system, known as sensor fusion. He also suggested that combining passive, real-time, and active sensors with "frequent-driver" programs could provide more comprehensive and accurate data.

Advances in Infrastructure-Based Sensors

Dr. Lianyu Chu
President, CLR Analytics Inc.

Dr. Henry Liu
Professor, University of Michigan Transportation Research Institute

MODERATOR

Raj Ghaman
Texas Transportation Institute

Improving Loop-Detectors

Dr. Chu presented his research on loop detectors and made some recommendations for their improvement. Dr. Chu explained that the basic function of a conventional loop detector is to obtain volume, speed, and lane occupancy data. He suggested that advanced loop detectors could obtain this basic information in addition to a unique vehicle signature to classify each vehicle.

Vehicle Signatures

Dr. Chu described how a signature can be determined from a vehicle's size, number of axles, metal mass, and height of undercarriage. Dr. Chu noted that different types of vehicles have different vehicle signatures and that the signatures from vehicles of the same vehicle class show some similarities. He mentioned that researchers are currently developing and testing detectors that can read these vehicle signatures. Detectors can produce very similar vehicle signature readings for the same vehicle with different detector loop shapes at different locations (e.g., a circle loop on the upstream traffic and a square loop on the downstream traffic). Dr. Chu highlighted that the two core algorithms that can use the signature data are vehicle reidentification and vehicle classification. Based on a dataset collected in California, he noted that the rate of accuracy for vehicle matching is 66.8 percent, which can be used to derive more reliable and accurate travel times. He stated that the vehicle classification accuracy within the FHWA vehicle classification scheme is 92.4 percent.

Future Applications

Dr. Chu suggested that a new signature-capable detector card is needed to advance this sensing technology. Figure 2 shows a traffic controller cabinet used to field test a signature-capable detector card. Dr. Chu mentioned that development is underway and includes field tests in California and Minnesota. He suggested that these detector cards will be useful for both freeways and arterials and that they will work with existing traffic controller cabinets compatible with both 170/2070 and National Electrical Manufacturers Association standards. Dr. Chu suggested that some potential applications could include: high-definition traffic system performance monitoring, conversion of vehicle- detection stations to vehicle-classification stations, estimating emissions, collecting origin–destination data, counting intersection turns, tracking heavy vehicles, and detecting bicycles.

A photo of a traffic controller cabinet. The cabinet is open and shows the cards and cables inside.

Figure 2. A traffic controller cabinet used to field test a signature-capable detector card.

Improving Data Quality

Dr. Liu presented his research on infrastructure- based traffic control system data and next generation traffic control systems. He explained that the current generation of intelligent traffic-signal systems uses fixed-location sensors to produce performance measures. These performance measures are used to create signal-control algorithms, which feed traffic-signal devices. Dr. Liu noted that most systems are closed-loop systems, and bicycles and pedestrians are often passive parts of the systems. He highlighted that these types of systems do not produce good data, but vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) communication can improve the data collected from traffic signals. Dr. Liu explained that his research team developed a data-collection device to improve the quality of the data collected from traffic signals.

Data Collection

Dr. Liu noted that field tests of TS-1 signal- control cabinets, which exist in many locations but are no longer the current standard, show that they can collect and archive every vehicle actuation and signal change before sending this information to a traffic control center.(1) Dr. Liu highlighted that each vehicle produces two data points: actuation and de-actuation. These data can help with the measurement of performance at the intersection level. For example, he stated that they could help with the estimation of queue length and travel time, as well as identify oversaturated conditions. Dr. Liu mentioned that with queue-length data it is possible to derive other measures, including delay, level of service, and number of stops. Moreover, he noted that looking at several intersections in a row will allow for the analysis of corridor-level travel time, trip delay, number of stops, and emissions. Dr. Liu suggested that once the performance measures are derived, it is possible to identify if there are any problems with a traffic signal.

Future Testing

Dr. Liu mentioned that next generation intelligent traffic-signal systems will enable controllers to advise individual vehicles on speed and route options. He noted that in these systems, vehicles, pedestrians, and bicycles are all actively sending and requesting data and actions. Dr. Liu suggested that V2I communication could offer even greater options for these systems. Dr. Liu mentioned that his team participated in The University of Michigan Transportation Research Institute Safety Pilot, which incorporated V2I technology. For this study, he noted that the research team instrumented 19 intersections and plans to instrument up to 60 intersections and 9,000 vehicles. Their goal is to expand the testbed to southeast Michigan and, in time, have the infrastructure to communicate speed advisories and rerouting information.

New Applications for Infrastructure-Based Sensor Data

Walton Fehr
Program Manager, U.S. Department of Transportation's Intelligent Transportation Systems Joint Program Office

Overview

During his presentation, Walton Fehr discussed the cooperative, communica- tion-based, intelligent transportation system (ITS) technology pilot, conducted by the University of Michigan's Transportation Research Institute in 2013 (also known as the Safety Pilot Model Deployment). He stated that the pilot study investigated the benefit of communication-based ITS on crash avoidance. Fehr noted that the results of this pilot led the National Highway Traffic Safety Administration to investigate rulemaking that would require communication technology in all new vehicles. He mentioned that there will be a large round of deployment trials and that more information can be found at http://www.its.dot.gov/pilots/.

Major Themes Discussed

Fehr discussed several themes during his presentation. He highlighted several requirements that a system such as the ITS technology pilot would require to have in place before moving forward. These requirements include establishing a common process for all information flows that preserves privacy and security; ensuring availability of data for all users of the transportation system ubiquitously and in a standard form that makes use of many different types of media, including dedicated short-range communications; and safeguarding conservation of privacy for consumer acceptance.

Next, Fehr mentioned that the U.S. Department of Transportation's (USDOT's) ITS Joint Program Office (JPO) developed a reference implementation architecture to support all of the mentioned requirements. Fehr noted that outreach is a major part of this effort and includes the following:

Fehr highlighted that USDOT entered into more than 62 memoranda of agreement with public, private, and academic organizations as affiliated testbed collaborators. This status allows outside parties, with no previous connection to FHWA, to participate in and follow activities surrounding the reference implementation architecture as they occur. Fehr noted that the ITS JPO developed a data-flow visualizer online that all of these members may access. He explained that there are three different types of fundamental data flows that the ITS JPO is trying to understand to help them build the reference architecture, as follows:

  1. Traveler Situation. This involves actionable information that travelers should have at their disposal, flowing from the transportation system managers to the vehicle.
  2. Field Situation. This involves control devices that exist at the boundary between mobile and fixed elements, such as traffic-signal controllers, which direct data from the field equipment to both the transportation system managers and the vehicles.
  3. Vehicle Situation. This data originates in the vehicle and flows to field devices or the back offices.

Data Management

Fehr then discussed what is required to measure data quality. He suggested that a basic safety message, containing vehicle safety-related information that is periodically broadcast to surrounding vehicles, is required to meet performance requirements. He also suggested that there needs to be a common interpretation of standards, for example, among geometric intersection data, mobility application programs, and signal phase and timing. Fehr mentioned that every data unit must meet the fundamental performance requirement, in that it must be operable in every situation. He noted that traveler situation data should also be delivered to USDOT's data distribution warehouse using an agreed format. This would enable USDOT to make the data available in different ways to data users.

Fehr noted that USDOT is keeping track of all of the data contributions and users so that, sometime in the future, there can be a market built around it. Fehr also noted that the ITS JPO created a traveler situation data tool that helps practitioners create data units and deliver them to the distribution warehouse. The tool automatically encodes the data according to USDOT standards. In summary, Fehr highlighted that V2I and V2V communication is extremely important for traffic control and will continue to be important as automation technology in vehicles becomes available.

Data Standards and Sensors

PANELISTS

Dr. Christos Cassandras
Professor of Electrical and Computer Engineering, Boston University

Richard Denney
Operations Specialist,  FHWA Resource Center

Dr. Stan Young
President, Traffax, Inc.

MODERATOR
Cathy McGhee
Virginia Department of Transportation

Controlling Traffic Networks

Dr. Cassandras addressed the challenges of controlling traffic networks and described some of his efforts to improve the accuracy of data to overcome these challenges. Dr. Cassandras explained that, even though the achievable optimum of a network can be calculated, it is still very hard to control traffic networks. He noted that one reason for this is that there are not enough controls in the system (e.g., traffic signals and tolls) and this leaves little opportunity to provide feedback. Dr. Cassandras mentioned that another reason is that drivers do not know what the other drivers are doing. This can lead to poor decisionmaking in which drivers act in ways that will benefit themselves, which is in conflict with those actions that will optimally benefit the system.

Dr. Cassandras highlighted that innovative sensor technology and the data it produces have the potential to overcome some of these challenges by creating smart transportation systems. He noted that sensors can be infrastructure-based, such as inductive loops, and camera- or imaging-based sensors, which are commonly used today. Dr. Cassandras suggested that another method is to use GPS and accelerometers in mobile phones and vehicles as sensors without any requirement for additional infrastructure. These types of sensors can provide status information such as vehicle position, velocity, destination, fuel level, and battery level. Dr. Cassandras noted that the accuracy of data is important and essential to creating smart systems and implementing successful real-time traffic control. He provided an example from his research to show how a smart-parking system using cameras and loop detectors, which are a low-cost solution, can produce accurate data with minimal delay (shown in figure 3).(2)

Dr. Cassandras mentioned that a challenge facing researchers in this field is to find new mechanisms to control traffic based on data collected through sensors. He highlighted some research projects he is working on to achieve this, as follows:

A screenshot shows a photo of a parking garage. Two green squares highlight vacant spaces in the garage, two red squares mark the occupied spaces.

Figure 3. Wireless loop detectors used for a smart-parking system.

Using Sensor Data

Richard Denney discussed how to use sensor data to create new standards that are useful for next generation control systems. He noted that the purpose of today's traffic sensors is simply to call and extend the green phase of the signal. Control systems have been built around reducing complaints from drivers about waiting at signals too long, waiting at queues that back up too far, or stopping at too many traffic signals. Denney highlighted that calling and extending green phases do not require distinguishing one car from another; therefore, these detectors are not useful for counting cars. He noted that the purpose for counting cars is to feed traffic-simulation models; however, that application for the data has not proven sufficiently valuable to motivate industry to develop the capability for sensors to accurately count cars.

Denney suggested that there is a need to establish how data taken from sensors can be made more actionable. For example, the data could be used to optimize traffic in a way that allows queues only in areas where they will cause the least performance degradation. Denney highlighted that, to take advantage of new detection technology, new use cases must be developed. He noted that part of developing the use case is understanding what needs to be optimized. These use cases must be clearly defined and helpful for today's uses.

Using Reidentification Data

Dr. Young presented on the importance of reidentification of data and how it can be used to produce useful data. Dr. Young noted that Bluetooth™ and other reidentification data provide broad-based travel-time data that do not have to be modeled to obtain a good sample. The resulting travel-time distributions allow for a detailed interpretation of intersection performance. Dr. Young highlighted that the data are comparable in accuracy to outsourced probe data. He suggested that travel time is particularly relevant because it is what the user experiences directly.

Dr. Young then showed workshop participants a cumulative distribution function (CDF) of sampled travel-time data, which can be used to compare information such as before-and-after signal timing and phasing of signals (shown in figure 4). He noted that travel time is the standard metric for assessing performance and highlighted that the desired way of measuring travel time is with a CDF because it facilitates comparison of before-and-after scenarios, different signal-timing approaches, and different facilities. In addition, it also shows degradation of performance over time. Dr. Young mentioned that a challenge for researchers is quantifying good or poor performance based on the CDF.

Dr. Young noted that the methods outlined in his presentation work well for highways, but industry does not yet know how to model an arterial management system and has not come to a consensus on effective performance measures. He suggested that travel time needs to be expressed in a way that can be understood by those who make decisions about funding.

Dr. Young also described a recent project that aims to validate outsourced probe data by comparing them to a reference dataset. The project is focused on the Interstate-95 corridor in Pennsylvania, and there are three vendors that are online and collecting data. Dr. Young mentioned that the datasets from each vendor are very similar to one another and to the reference dataset. He stated that this means that outsourced probe data are effective at capturing vehicle congestion on interstates.

Two charts show the cumulative distribution function from sampled travel-time data. The chart on the left compares travel time in minutes to the hour of day. The chart on the right compares cumulative probability to travel time in minutes. As the hour of day increases to 4 PM the travel time increases on the left chart. This is matched by a sharp increase in cumulative probability at the same time on the right chart.

Figure 4. Cumulative distribution function from sampled travel-time data.
Note: black data points on the left chart highlight the 4–5 PM peak hour, which has
the greatest delay. This corresponds to the curve highlighted in the chart on the right.

 

 

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101