Traffic Congestion and Reliability:
Trends and Advanced Strategies for Congestion Mitigation
A. Technologies for Advanced Traffic Monitoring
[Note: This Appendix was contributed by Mark Hallenbeck, Director, Washington State Transportation Center at the University of Washington.]
Reporting and using performance measures requires the collection of data that can be used to compute performance measures. In general, three different categories of data collection are required; data regarding
- Facility use and performance;
- Staff activities and resource use; and
- Events and incidents that disrupt "normal" freeway conditions.
Facility use and performance data provide an understanding of the mobility benefits provided by a freeway system. Staff activities and resource use data describe how available resources are being expended. Event and incident data describe the external forces that influence how a freeway operates. When combined, these three categories of data provide insight into what is happening on the freeway system, and what types of changes in mobility are occurring as a result of the combination of "events" and the application of various resources and policies to provide mobility in the face of those "events." Each of these subject areas is described below.
Facility Use and Performance
Facility use and performance can be collected either continuously (usually by capturing surveillance or control system data), or using samples, usually as part of special data collection studies.
Continuous Data Collection
A wide variety of technologies can be permanently mounted and used to provide continuous facility performance information. The available options can be broadly categorized as:
- Point detection;
- Beacon-based probe vehicle data; and
- Nontraditional probe vehicle performance.
Point detection involves placing surveillance equipment at a specific location and using the measures of traffic performance at that location to estimate traffic performance over a segment of roadway. The "roadway segment" described by a single point detector is normally between one-third to one mile long in most United States freeway surveillance systems.
Point detectors generally report data on vehicle volume and lane occupancy (which when combined can be used to estimate vehicle speed), and when deployed in a "dual loop" configuration can also directly measure and report vehicle speed and vehicle classification (by length.) These basic measures serve as the basis for a wide variety of the mobility (outcome) performance measures described elsewhere in this report. Vehicle volume estimates at different points on the roadway describe how heavily the roadway is being used. When combined with vehicle speed information, vehicle volume statistics describe how efficiently the roadway is operating. (That is, during what time periods does congestion effect vehicle throughput, and how far below "optimal throughput" is the freeway operating during those congestion periods?)
Speed measurements at consecutive points along a roadway segment can usually be converted into reasonable measures of travel time along freeway segments, thus providing performance measures describing the delays being experienced by travelers using that roadway.
The fact that data are collected continuously means that analysis of these data allow a review of the time-of-day, day-of-week, and geographic trends present in travel patterns. This allows agencies to understand when, where, and how frequently problems are occurring on their roadways, and how those trends change as new counter measures are implemented.
Continuous data collection means that "unusual" conditions are also measured, and the effect these conditions have on freeway performance can be determined and compared against "routine" congestion. This allows agencies to understand the relative importance of different "unusual" events, and gage the relative value of spending resources on responding more effectively to these events versus spending those resources on improving "routine" conditions. That is, how reliable is the freeway system? When and where does the freeway become unreliable? And how significant are the delays of these "unreliable" times compared to routine commute period congestion?
A wide variety of sensors can be used to provide these point statistics, with the most common being
- Inductance loops;
- Video detection (most commonly used to simulate inductance loops); and
- Microwave radar.
However, a number of other technologies, including infrared, sonic, and acoustic sensors can provide these same data inputs.
Selection of the data collection technology to be used is usually a function of conditions under which the system will operate. Each technology has strengths and weaknesses which lead different agencies to select different technologies.
Inductance loops are inexpensive to purchase, and are generally considered a robust, well known, reliable technology. However, inductance loops require lane closures for installation and for maintenance of the wire loop itself. In freeze/thaw climates, in pavements in poor condition, and if installation is poorly done, the wire can break, meaning that additional lane closures are required to replace the failed loop. In addition, because loops are physically "cut" into the pavement, they are not moveable, and thus must be replaced if lane lines are moved as a result of new construction activities or other geometric and operational changes.
Video image detection technology was designed in part to deal directly with the limitations in loop technology. Because cameras are above ground, in many (but not all) instances, traffic lanes need not be closed to place, repair, maintain, or adjust the data collection devices. If lane lines are changed, detection zones in the camera image can often be "redrawn" without physically moving the camera system, thus allowing continued data collection without roadway closures or other significant disruptions to the facility or data collection system. Camera-based systems are thus generally considered a better choice than loops for point data collection on roadways where construction activities are expected to result in changing lane configurations or where pavement conditions make loop life problematic.
However, video image detection techniques also have limitations. Most of these problems stem from the fact that video systems can only measure "what they see." Thus video systems tend to work poorly in low-visibility weather conditions (e.g., heavy snow and thick fog.) Thus, they are often not recommended for implementation in climates where these conditions occur frequently.
In addition, video detection from beside the roadway can suffer from "occlusion"40 which degrades the accuracy of traffic volume counts. Cameras placed over top of roadways generally have less problem with occlusion, but often require lane closures when being installed, repaired, or maintained, because of safety rules governing work above active roadway lanes. Finally, cameras frequently require more routine maintenance than loop detectors, as dirt and water can reduce image clarity, thus degrading system performance.
Microwave radar technology was developed, in part, in response to the limitations in loop and video technology. The characteristics of microwave radar signals means that these systems are not affected by the weather problems experienced by video detection. Microwave radar does, however, have other minor limitations that generally result in slightly less accurate volume count information than obtained with loops and/or video detection. Like video detection, microwave radar can work from sensor positions either above the traffic lanes, or from beside the roadway. And also like video, the "above" locations provide more accurate data (less chance of occlusion) than the "side-fired" positions. Unfortunately, the "side fired" positions are usually less expensive to install, maintain, and repair because they do not require working within the constraints of moving traffic.
The primary limitation of point detectors is that they provide information about the performance about a single location, and that location may not be an accurate representation of the performance of the rest of the roadway segment to which those data are associated. This problem becomes less of a concern, the more closely spaced the point detectors. (That is, it is not much of an issue if the detectors are 400 feet apart, but it can be a problem if a single detector is used to represent traffic conditions along a mile-long roadway segment.)
Similarly, the fewer the geometric disruptions within a roadway segment, the better a single point detector is at estimating conditions within that segment. (For example, a single detector location will do a better job estimating conditions within a 1-mile roadway segment if there are no interchanges or major geometric features within that 1-mile segment than if there are interchanges or major geometric changes.) Even with fairly closely spaced detectors, the location of the detectors near specific traffic disruptions can effect how "representative" the data they produce are relative to the roadway segment they represent. For example, a detector placed just upstream of a ramp merge may underestimate roadway segment speed, as that specific section of road may see slightly slower speeds than the segment as a whole, as vehicles slow to allow vehicles entering the freeway from the ramp to merge.Such a sensor placement might also measure more "congestion" than the roadway segment as a whole experiences, because the merge point is likely to be the location within the segment which experiences congestion first (and stays congested longest) as a result of that merge.
As a result of these considerations, there is relatively little "simple" guidance on the deployment of point detectors. Instead, engineers designing point detection-based surveillance systems must tradeoff the cost, accuracy, and functionality of different sensor spacings and placements against the available budget and system requirements in order to settle on an appropriate design, and that final design effects how the resulting sensor data are converted into performance measures.
Beacon-based probe vehicle data collection is most commonly associated with electronic toll data collection systems. In these systems, a device (beacon) that uses Dedicated Short-Range Communication (DSRC) standards interrogates electronic vehicle tags as vehicles pass that reader location. The result is a data record that indicates when individual tag-equipped vehicles pass particular points on the roadway. (For toll collection systems, this allows automated billing of the owner of that vehicle, without forcing that vehicle to stop.)
By matching the time and location data associated with each vehicle as it passes from one beacon location to the next, it is possible to determine the travel time for that vehicle between two consecutive beacon locations. The result is an excellent data set that describe travel times for roadway segments defined by the location of the data collection beacons.
For toll roads with electronic toll collection systems, these data are essentially "free" for performance monitoring, since they exist for billing purposes. In cities where electronic tolling exists on some roads, it is also common (and fairly inexpensive) for agencies to place some additional readers on "free" roadways in order to capture vehicle time and location data on those facilities. In these cases, these data are not used for toll collection, but they do produce excellent travel time information on those selected road segments.
The result of these beacon-based systems is a very robust measurement system of travel times between tag readers. These data can then be aggregated and summarized to produce all of the travel time, delay, and trip reliability measures discussed elsewhere in this report. Travel times collected in this manner are more accurate than those estimated from point detectors, but they do not provide information about the geographic distribution of delays within the road segment being monitored.
For toll roads with closely spaced electronic toll collection points, the lack of spatial detail is not an issue. However, where tag reader spacings are fairly large, the lack of geographic detail can be a drawback, as the data collection system does not describe the location of any delays that are occurring. This can be a significant limitation on "free" roadways where multiple congestion locations (e.g., interchanges) lie within the measured roadway segment, as no data are gathered on the relative size and frequency of congestion caused by each of those locations.
Another limitation is that most toll tag-based data collection systems do not provide a measure of total facility use. Unless all vehicles using the facility must carry toll tags (this is the case on some roadways), the toll tag readers only monitor the performance of a sample of all vehicles. While this is generally good enough to provide an excellent measure of travel time, it does not provide an excellent measure of vehicle use on the facility. Thus, only half of the basic mobility performance information is available from most toll tag-based data collection systems.
As a result, vehicle volume use must be collected from other sources. For toll facilities, it can be obtained from the toll collection statistics. For "free" roadways taking advantage of the prior existence of tag-equipped vehicle fleets, additional data collection efforts are needed. These frequently involve a limited deployment of point detectors. The advantage of this "combined" approach to data collection is that the tag reader system provides very accurate and robust travel time information, and the point detectors can provide both vehicle volume information and estimates of the geographic distribution of delays. The downside of these systems is that they require two sets of data collection hardware and software.
Nontraditional probe vehicle performance systems are designed to provide travel time, speed and delay information without the beacon-based communications system. This allows these systems to cover much larger geographic areas (i.e., entire urban roadway networks) without the cost of building beacon-based communications infrastructure throughout those networks. None of these systems are actively used on a continuous basis in the United States, but they are in use in some parts of the world, and considerable effort is underway to complete testing and development of them in the United States.
While there are multiple companies/agencies working on this basic data collection approach, most of these efforts can be broken down into two general concepts.
- Cell phone tracking; and
- GPS-equipped vehicles with wireless data transmission.
A variety of different techniques are being promoted within each of these general approaches. Ongoing research is expected to provide more details in the near future about the costs and accuracy of estimates from these different approaches. Because the development process in this area is still underway, this subsection only introduces these data collection topics, and does not attempt to judge their relative merits.
Cell phone tracking techniques take advantage of the fact that it is possible to determine the approximate location of all cellular phones. By tracking the movement of cell phones it is possible to determine the speed of the cell phone. By restricting the analysis to those phones located on roadways, cell phone tracking provides a means to measure vehicle speeds on those roads.
Federal legislation intended to improve emergency response to cell phone users (E 911 requirements), and the commercial potential for "location-based services" associated with the location of cell phones have resulted in considerable effort to improve the accuracy and decrease the cost of collecting cell phone location information. Research is currently underway to determine the accuracy and cost of converting that information into roadway performance information.
The advantage of this technique is that the number of cell phone-equipped vehicles is quite high, and increasing. Thus, any road on which a cell phones is currently located becomes a data point on which vehicle speed can be obtained. This means that (potentially) entire roadway systems can be monitored without the need to install costly "roadway monitoring infrastructure." That is, the infrastructure will exist to meet E 911 needs, and roadway performance data can be obtained at the marginal cost of processing the existing cell phone data into roadway performance measures. Exactly what those "marginal costs" will be, and how accurate those performance statistics are (given the need to correctly assign specific cell phones to specific roads, and to remove from the data sets those phones not in vehicles without biasing the data being collected from very slow moving vehicles), is the subject of various ongoing field operational tests. However, significant potential exists for this technique.
The second technique takes advantage of the significant reductions in the cost of Global Positioning Satellite (GPS) technology. GPS devices report current location, heading, and speed information with a high degree of accuracy. When placed in vehicles and combined with electronic map information, GPS devices are the primary component of excellent vehicle location systems. Storage and analysis of the GPS location data allow for very accurate roadway performance measurement. The difficulty with GPS data is that it is the vehicle carrying the GPS device that has this performance information. To convert data from GPS-equipped vehicles into roadway performance information usable by a freeway operations agency, it is necessary to provide some communication mechanism to/from GPS-equipped vehicles in order to obtain that vehicle location and performance data. In addition, to provide reliable roadway performance estimates, a large enough number of vehicles must be equipped with GPS to provide an unbiased measure of roadway performance, and to provide the temporal and geographic diversity desired by the performance measurement system.
Cellular phone tracking has the advantage that a very large number of drivers/ passengers in vehicles now carry cellular phones. Thus, a large number of potential probes exist. GPS technology requires that GPS devices be installed in vehicles. While the number of GPS-equipped vehicles is increasing slowly (for example, all On-Star-equipped vehicles have GPS devices, even if the vehicle owner does not subscribe to the On-Star service) the number of GPS-equipped vehicles is still relatively small, and the majority of those vehicles do not provide for routine communication of their GPS data to outside sources.
Several ongoing research efforts are working to resolve the cost and device distribution issues associated with GPS technology. Dramatic changes in wireless communications technology have significantly lowered the cost of wireless communications, allowing more cost effective retrieval of data collected by vehicle probes. Significant private and governmental efforts are underway to promote both the Intelligent Vehicle Initiative and the Vehicle-Infrastructure Integration (VII) programs, which encourage adding technology like GPS to vehicles, and are aimed at providing the infrastructure necessary to communicate key pieces of information from that technology to the roadside to improve safety and operations.
As with cell phone tracking, a number of operational field tests are currently underway that are testing new developments in these areas and that have the potential for providing new, robust, data sets that can be used for roadway performance monitoring. For example, efforts are currently underway in Germany to incorporate GPS and cellular phone technology into heavy vehicle tax/fee collection systems. Adoption of similar systems in the United States would provide another source of vehicle (and thus roadway) performance information.
One significant drawback to probe vehicle-based performance monitoring (whether cell phone- or GPS-based) is that, like toll tag tracking, it does not provide information about the level of roadway use (vehicle volume.) It only provides information about the speeds and travel times being experienced. Thus, if probe vehicles are the primary source of performance information used, some supplemental data collection will be needed to supply the performance measures related to the level of use freeways are experiencing.
Special Study Data Collection
The previous discussion of data collection technologies focused on the types of technologies that operate continuously. That is, once placed in the field, they provide data regularly for reasonably long periods of time. Where these systems do not exist and agencies can not afford to implement them (or where supplemental data sets are required), special, short duration studies are often performed.
These special studies have the advantage of generally having lower costs. They have the disadvantage of (normally) being noncontinuous, and are thus less likely to be able to accurately collect performance data on the number, frequency, and severity of "unusual" events.
Special studies are generally focused on collecting specific pieces of information, not available through existing sources. Since these sources differ from region to region, the special studies needed in one region are different than needed in others.
In a region with significant freeway surveillance and incident response systems, special studies may only be needed to provide the vehicle occupancy and transit ridership information needed to convert vehicle volume information into estimates of person throughput, person hours of delay, and other performance measures that relate to key policy initiatives.
In other areas where continuous data collection systems do not exist, special floating car travel time runs may be performed in order to provide the baseline travel time statistics needed to judge routine freeway performance. Similarly, special traffic volume counts are often performed to provide key statistics on freeway use.
As with continuous data collection, a wide variety of data collection techniques exist for collecting freeway performance information on a "special study" basis. The following discussion describes only a few of the more common techniques.
Traffic volume counts on high-volume freeways can be very difficult to perform, as traditional axle sensor-based traffic counters can not be safely deployed on high-volume roadways. The result is that many state agencies have been working with vendors of non-intrusive data collection technologies to develop portable versions of these devices.41 Most commonly, these devices (usually using microwave radar, video, or acoustic sensor technologies) are placed, along with a power source such as a solar panel or batteries) on an extendable pole attached to a trailer. The trailer is then placed beside the roadway, behind a guardrail or concrete barrier, and the data collection device views traffic from a "side-fired" orientation.
The other common approach is for an agency to place convention road tube-based counters on all ramps within a corridor and use those ramp counts to estimate volumes on the freeway mainline.
Travel time and delay information is most commonly collected using floating car studies. However, as with other data collection efforts, a wide variety of other techniques can be used to collect this information.42 The floating cars themselves can be paid data collection consultants, or volunteers recruited for the task. Travel times can also be collected using various license plate (or other vehicle) matching techniques.
In general, floating car studies provide better geographic information relative to the trip being monitored (i.e., where are the delays taking place and how big are those delays) than license plate matching approaches, but the license plate matching techniques provide a much bigger sample of the travel times being experienced by different vehicles using that roadway during the study time period.
All short-duration travel time studies have limitations in collecting data relative to temporal differences in travel time (whether those are time-of-day or day-to-day temporal differences), simply because of the cost of collecting those data. Thus, short-duration travel time studies usually only collect data for a sample of days and time periods. If the sample design is done well, a special study will produce an unbiased estimate of the mean condition and a reasonable estimate of trip reliability. However, limitations in the number of hours of travel being monitored make it difficult for short-duration studies to provide accurate measures of travel reliability relative to both time-of-day and day-to-day variations in traffic conditions.
Other congestion measures, especially the geographic extent of congestion, have been collected by a number of regions using aerial surveillance. The most common form of this is when agencies hire consultants who fly planes43 and take photographs of traffic conditions of specific roadway segments over multiple hours and/or days. The photographic images are then analyzed to provide estimates of volume, delays, and the geographic distribution of congestion by time of day for the roadways being studied.
Estimating person use of freeways corridors almost always involves short-duration data collection efforts to collect the vehicle occupancy data needed to estimate average vehicle occupancy (AVO). Vehicle occupancy counts may include transit ridership estimates, although in many cases transit ridership can be obtained directly from transit authorities (who perform this task with some combination of manual counts and automated passenger counters located on buses).44
Most vehicle occupancy counts are done manually, although, some vendors of image detection software are starting to market systems that they claim can count passengers in vehicles. Vehicle occupancy counting on freeways is quite difficult, and can only be done when lighting is good, from locations where the viewing angle into the passing vehicles allows a clear view into the passenger compartment, and where the data collection personnel can stand (or sit) safely. These limitations, along with the cost of manual data collection generally limit AVO counts to a few sample locations, on a few sample days, along each freeway corridor of interest. These AVO estimates are then applied throughout the corridor.
- Occlusion occurs when one vehicle "hides" another vehicle within the video image. This commonly happens when a truck passes a camera placed beside the roadway. A car on the far side of the truck from the camera can not be seen by the camera, and is thus not counted by the video detection software.
- More on these efforts can be learned by visiting the National Detector Clearinghouse at http://www.nmsu.edu/~traffic/. Another good resource is Summary of Vehicle Detection and Surveillance Technologies, https://www.fhwa.dot.gov/ohim/tvtw/vdstits.htm.
- A good resource for learning more about travel time data collection is the Travel Time Data Collection Handbook available on-line at https://www.fhwa.dot.gov/ohim/timedata.htm.
- Other techniques exist to perform this same basic data collection task.
- If transit ridership can be collected directly from the transit authority, AVO counts are normally restricted to non-transit vehicles. Person volume is then computed as (AVO * volume) for non-transit vehicles + transit ridership.