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

Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

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

This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-14-058    Date:  April 2015
Publication Number: FHWA-HRT-14-058
Date: April 2015


Field Evaluation of Detection-Control System



This chapter uses the results and analyses from earlier chapters to formulate conclusions and recommendations based on results of the four studies. The goal of the traffic data collection was to determine propensity for red-light running and vehicles caught in the dilemma zone. The following conclusions are organized beginning with those related to the methodology and followed by conclusions related to field data collection elements of the research. The recommendations are the final section.


The objectives of this evaluation study were as follows:

  • Objective 1: Verify the D-CS design objectives through rigorous field instrumentation—at the moment of signal change from green to yellow, no truck should be in the dilemma zone, and no more than one passenger car should be in the dilemma zone.
  • Objective 2: Quantify the effectiveness of D-CS in improving safety and reducing dilemma-zone-related crashes and red-light violations at rural, high-speed, signalized intersections.
  • Objective 3: Identify the upper limit of traffic conditions under which the D-CS can operate safely and effectively when alternative signal timing strategies may start to fail.

To meet the three research objectives, TTI developed the following four studies:

  • Study 1: Performance Monitoring of Dilemma Zone Occupancy (addresses objective 1).
  • Study 2: Before-After Crash Data Study (addresses objective 2 in part).
  • Study 3: Before-After Crash Surrogate Study (addresses remainder of objective 2).
  • Study 4: Upper Limit Study (addresses objective 3).

TTI conducted studies 1 and 3 simultaneously by collecting field data at eight sites in four States (Florida, Illinois, Louisiana, and Texas). During study 4 (an extension of studies 1 and 3) data were collected at a high-volume site to determine whether an upper limit exists to the demands placed on the D-CS algorithm. MOEs for studies 1, 2, and 4 were the number of RLRs, the number of vehicles caught in the dilemma zone at the onset of yellow, and max-out frequency. Study 2 simply involved a comparison of the number and types of crashes from the before D-CS period (desirably 5 years) to the after D-CS period (desirably 2 years).

Related to Methodology

The methodology used for this research involved collecting a redundant set of data using different systems to ensure capture of the critical data. The industrial PC in the controller cabinet at each intersection recorded the data from the Wavetronix™ Advance detectors, the inductive loop actuations from the D-CS loops, and video detector outputs from detection zones placed just past the stop line at each main street approach. The methodology also included video cameras and DVRs for recording video of each main street approach. Data from the enhanced BIUs provided the controller state for post-processing of data. The major problem at the first site was that the CyberResearch™ industrial PCs was unable to handle the massive amounts of data being transferred for storage via its serial ports. The research team resorted to using two of its own Kontron™ industrial PCs, which worked flawlessly throughout the entire project.

TTI also used a cell modem in each controller cabinet to facilitate remote communication with each site. Attempting this project without such communication capabilities would have been unwise. The wireless modem allows the researchers to remotely access the site and download data. Otherwise, a researcher has to go to the site, manually download the data, and restart the data-logging process. Even with this capability in place, TTI still had to request local transportation department support occasionally to do a power cycle, install firmware upgrades, or complete other relatively simple tasks. In Florida, the sites experienced several power outages, prompting the local transportation department to install an “iBoot” device to be able to execute a power cycle remotely.

Each transportation department was willing to provide bucket-truck support for installation and removal of equipment. Attempts to pull cables through existing conduits were usually unsuccessful, so researchers resorted to running cables overhead and zip-tying them to existing cables during the few weeks the systems were in place. By using this methodology, researchers were able to reuse the same cables for all sites.

TTI attempted to collect at least 7 days of before data and 7 days of after data at each site. In a few cases, scheduling or other conflicts precluded collecting the full 7 days. This amount of data still far exceeded the amount actually needed and created challenges in manipulating files during the data analysis phase, especially when combined with all the different data elements being collected. For example, Microsoft® Excel (2007 and later) has a physical limit of one million lines of data. Many of the sites had so much data that one 24-h period exceeded this limit, requiring analysts to split files into multiple segments.

The primary system for identifying vehicles in the dilemma zone consisted of two Wavetronix Advance detectors—one for each main street approach. The detector tracks vehicles approaching the detector from 500 ft away until they reach about 100 ft away. TTI used the serial data stream to determine when vehicles were in the dilemma zone (defined as 2 s to 6 s travel time to the stop line). The output from the detectors provides the distance and speed of each vehicle it detects. Even though these sensors are accurate and worked very well for this purpose, their output (scanning each approach every few milliseconds) is probably what overwhelmed the CyberResearch™ computers. Another challenge from these detectors was the fact that they do not distinguish between trucks and cars. Because D-CS is designed to favor trucks, this research had to rely on the D-CS loops and their vehicle length determination to identify trucks (defined as vehicles more than 30 ft in length).

Cameras for video recording usually served a dual role—providing video to be recorded and serving as the count device for RLRs and total intersection approach counts. To detect RLRs, installers placed a detection zone just past the stop line. A detection occurring just past the onset of the red phase triggered a red-light running event, but the analysis process still verified these supposed events. For example, a vehicle stopping just beyond the stop line might trigger the detector but was not actually an RLR. Also, a vehicle’s headlight “sweep” at night from an opposing phase could trigger an unintended detection that might otherwise appear as an RLR.

Video recording was not continuous because of the enormous storage requirement. Researchers decided to purchase two DVRs with high-end features to limit the amount of video recorded. They chose a DVR with alarm capabilities so that the onset of red would trigger the recording to begin. Field personnel programmed each DVR to begin recording 5 s prior to the beginning of the red phase and 6 s after the beginning of the red phase. With this much lead-in video, post-processing of the video allowed viewers to watch vehicles on the approach several hundred feet before arriving at the stop line. There were several cases of RLRs entering the intersection very late in this recorded period, but such blatant violations were not counted against D-CS because no detection scheme would have likely prevented such results.

The data analysis methodology required the use of an exposure measure, so TTI needed to count the traffic passing straight through the intersection. Video detectors placed at the stop line provided such a count, but a weakness of video detectors with regard to counting was that they could not accurately distinguish between cars and trucks at night because of the nighttime algorithm’s use of headlights instead of the full vehicle length as in the daytime. The inductive loops at about 1,000ft from the intersection were more accurate, but counts at that location would include turning traffic. Therefore, TTI chose to use the video counts and not try to distinguish trucks from cars.

Matching data from multiple systems requires use of a common clock time. The most likely source for the components used to collect the data would be the PC in the cabinet, although not all devices in the cabinet were capable of using its clock. For example, the DVR had its own clock and was independent from the PC and related systems. However, researchers bypassed this issue by using VideoStamp devices, which caused the timestamp from the PC to be recorded on the video image, facilitating post-coordination with other data even if the DVR clock was significantly different from the PC clock. The VideoStamp device has an RS-232 port to receive the timestamp from the PC and an RCA video input to receive the video input recorded by the DVR. Without this device, the most appropriate approach was to synchronize all clocks either physically at the start of data collection and periodically afterward, or to find common points that could be associated with each system, such as the start of the red phase.

The methodology at the data collection sites was as consistent as researchers could feasibly make it. However, there were differences by site that affected the results and how the results should be interpreted. For example, some sites had no before dilemma-zone protection. In other words, some sites had no upstream detection whatsoever. All three Florida sites and the Louisiana site fit this category. Obviously, D-CS should reduce RLRs, vehicles caught in the dilemma zone, and the number of vehicle crashes at these sites compared with those with dilemma-zone protection prior to D-CS being installed. Better evaluation of D-CS would come from its comparison at sites with the more traditional dilemma-zone protection such as those in Texas and Illinois. In this study, the researchers evaluated before/after crash data at the evaluation sites and comparison sites. To assess effectiveness of D-CS in reducing dilemma zone-related red-light running and crashes, it is better to perform side-by-side comparison of D-CS with other dilemma zone protection technologies at the same location.

TTI followed instructions from the sponsor and began investigating red-light running, dilemma-zone encroachments, and max-outs for all days of both the before and after period. However, after many days of watching video to verify dilemma-zone encroachments and completing only a few days of actual data evaluation, the research team concluded that project resources were insufficient to complete the project this way. Continuing would provide hour-by-hour or day-by-day comparisons, but that became impractical. Therefore, researchers began using a different approach that required comparison based on the number of signal cycles at each intersection and used the count of vehicles as an exposure factor. The number of RLRs, the number of vehicles caught in the dilemma zone, and the number of max-outs were still the variables of interest in this procedure, but the procedure was not nearly as onerous as the previous labor-intensive approach. This procedure, which used a regression analysis methodology, also accounts for differences in traffic volume, site features, cycle length, and other known factors. It would have accounted for weather as well, but there were no weather conditions that were thought to affect the outcomes.

Related to Crash Surrogate Measures

Researchers began the data analysis using 24-h periods and developing before-after comparisons on that basis. Evaluating this quantity of data was not necessary, and resources were not available to continue and complete all sites on that basis. However, this document presents the limited partial results for information. Based on the partial analysis using 24-h data, TTI found that the number of RLRs always decreased with the use of D-CS compared with the before treatments.

Table 48 summarizes these results for weekdays only; weekend results might be different. Results in Illinois are especially important because the two sites there were the only ones in this evaluation group with reasonably adequate dilemma-zone protection before D-CS was installed. The other sites had no dilemma-zone protection before D-CS. Dilemma-zone results improved for cars with the use of D-CS as well, but results for trucks were different in this sample. The number of trucks in the dilemma zone were the same in before to after periods at the Cummings Lane site, and they increased from one in the before to three in the after condition at the Main Street site.

Table 48. Summary of partial 24-h operations data.




Dilemma Zone





U.S. 24/Cummings Ln.











U.S. 24/Main St.











LA 3235/LA 3162







U.S. 27/Griffin Rd.







—No data.

Because of resource constraints, researchers stopped the 24-h data analysis and resorted to a methodology using regression analysis. Results of the regression analysis indicated that D-CS decreased red-light running, the number of vehicles in the dilemma zone, and the number of max-outs. Findings from 28 1-h periods indicate an 82-percent reduction in RLRs, a 73-percent reduction in vehicles caught in the dilemma zone, and a 51-percent reduction in max-outs.

Related to Upper Limit Study

The statistical analysis used for the Upper Limit Study using data from U.S. 281/Borgfeld Drive near San Antonio, Texas, indicated that the max-out frequency decreased with increasing maximum green duration from 75 to 85 to 95 s. However, this trend was not statistically significant. The field effort related to the Upper Limit Study was originally planned as a precursor to simulation. It would have served the role of model calibration as well as determining the effect of increasing maximum green. However, project resources were insufficient to do both.

Related to Crash Data Analysis

Chapter 1 provides crash results based on an earlier evaluation of D-CS at five sites in Texas. There was a 39-percent reduction in severe crashes on the two approaches controlled by D-CS. The data suggest that 9 severe crashes (and about 18 property-damage-only crashes) were prevented during the time that D-CS was operating. If just those crashes that are influenced by D-CS are considered (i.e., rear-end, left-turn opposed, and sideswipe), then D-CS installation accounted for a 50-percent reduction in severe “influenced” crashes.[1]

The more recent crash data analysis using comparison sites suggests that D-CS had no effect on TOT and FI crashes and produces a reduction of 9 percent for angle plus RE crashes. The standard deviation of this estimate of average safety effect is 15 percent, so at a 95-percent confidence level, the result is not significant. This result can be attributed to the small sample size. Achieving a significant result at the 5 percent level would require a larger number of treated sites, a larger period of crash data, or both.


TTI recommends that D-CS be viewed as a viable solution to improving intersection safety at high-speed, isolated intersections. Its emphasis on trucks is a salient feature that makes it unique in comparison with other types of dilemma-zone protection. Agencies that would not consider above-ground detection for dilemma zones should have an option available to them such as D‑CS (because D-CS has relied on inductive loops). However, there are reasons to investigate non-loop options for D-CS, including wireless communications and above-ground detectors such as side-fire radar detectors, which can provide speed and length. Some agencies are already minimizing the installation of detectors in the pavement in favor of non-intrusive options, so this approach would improve the chances of D-CS becoming more universally applicable and perhaps less expensive.

The TTI research team encountered a fair amount of concern within State and local transportation agencies about installing Naztec controllers in cases where the local agency had no experience with this controller. The Government is in the process of approaching other manufacturers to encourage implementation of the D-CS algorithm. TTI engineers had contacted other manufacturers earlier when they contacted Naztec, but at that time, the other manufacturers decided against D-CS. With new evidence that D-CS improves safety at high-speed intersections, the manufacturers might now respond differently.

To integrate the D-CS algorithm, other controller manufacturers must have a significant incentive to do so, and they will probably need support from programmers who are familiar with the D-CS algorithm. A positive response from the controller manufacturers today might also occur simply because of above-ground detectors that are either available today or will be available soon to provide dilemma-zone protection. The current dilemma zone detectors do not have all the same features as D-CS (e.g., the emphasis on trucks and speed and length measurement accuracy), and the only way to know how well they work is to test them scientifically in a side-by-side comparison with D-CS.

Application Considerations

D-CS is intended for use at isolated, full-actuated intersections on high-speed roadways where the major road approach has an 85th-percentile speed (or posted speed limit) of 45 mi/h or higher. A left-turn bay is required for each major road approach, and a right-turn bay (or full-width shoulder) is desirable. For existing intersections with multiple advance detectors, decisionmakers should consider replacing the existing system with D-CS when the existing system’s design life is finished.

Simulation and field study have shown that the system’s performance degrades with frequent turning activity from the major road approaches. Performance has been acceptable when the turn percentage is less than 40 percent. The following conditions make D-CS even more desirable:

  • Higher than normal truck traffic.
  • Locations where approach speeds vary significantly.
  • Locations with high crash rates (e.g., angle plus RE crashes).

[1]The D-CS enhanced controller is manufactured by Naztec (as part of Trafficware).


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