Traffic Control Devices and Measures for Deterring Wrong-Way Movements (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

24 death of the SAPD officer resulted in a significant response from law enforcement, the media, and the public. In May 2011, public transportation and law enforcement agencies in the San Antonio area created a WWD task force to share information and identify means to address and reduce WWD activity. The task force used various methods to doc­ ument WWD activity in San Antonio, with the purpose of identifying where WWD countermeasure deployment would be most meaningful and effective. After analyzing the vari­ ous WWD event data sources and information details avail­ able from each source, analysts determined that insufficient information existed to link WWD events with specific free­ way ramps where wrong­way drivers entered the freeway net­ work. Accordingly, there was no logical means that could be devised for prioritizing the treatment of one freeway ramp over another. Thus, the task force concluded that treatment of an entire freeway corridor was necessary in order to deter­ mine the effectiveness of WWD countermeasures. The task force selected the 15­mi US 281 corridor from I­35 (near downtown) to just north of Loop 1604 (the far north central side of San Antonio) as the Wrong­Way Driver Countermeasure Operational Test Corridor. Between March 2012 and June 2012, Texas DOT staff and contractors installed WRONG WAY signs with flashing red LEDs around the border at 26 exit ramps in the US 281 test corridor (see Figure 11 and Figure 12). The purpose of the flashing red LEDs was to increase the conspicuity of WRONG WAY signing at night. Originally, a single radar unit was installed to detect wrong­ way vehicles and activate the flashing red LEDs. However, sustained false alarm issues with the single radar setup led to deactivation of detection components. Therefore, the signs were set to flash under low ambient light conditions (i.e., at night and during some inclement weather events), whether or not a wrong­way vehicle was detected. Texas DOT thought that this operation was acceptable because it could potentially catch the attention of a wrong­way driver approaching on the frontage road, instead of waiting until the wrong­way vehicle Since the 1960s, transportation agencies have been imple­ menting and testing active detection and warning systems to deter wrong­way entries onto freeways (e.g., Tamburri and Theobald 1965; Tamburri 1965; Rinde 1978; Knight 1983; Moler 2002; Cooner et al. 2004; American Traffic Safety Ser­ vices Association 2014). More recently, agencies have been installing systems that activate flashing LEDs within the bor­ ders of WRONG WAY and DO NOT ENTER signs and are experimenting with red RFBs in conjunction with WRONG WAY signs to increase the sign’s conspicuity. This chapter pro­ vides a brief overview of applications of these two counter­ measures in Texas and Florida, and documents the research methodology, analysis, and results of evaluations conducted. LEDs within the Border of WRONG WAY Signs For the evaluation of the effectiveness of LED border­ illuminated WRONG WAY signs, the research team analyzed wrong­way driving (WWD) event datasets from San Antonio, Texas, as well as South Florida. The sections below provide a brief history of the LED border­illuminated WRONG WAY sign implementations in each state, document the findings of the analysis in each state, and provide some additional statis­ tics gleaned from both datasets. Texas Background At approximately 2:00 a.m. on March 15, 2011, a head­on collision occurred in San Antonio, Texas. This crash was caused by a driver who had entered the interstate going in the wrong direction. Both the wrong­way driver and right­ way driver (a San Antonio Police Department [SAPD] patrol officer) were killed. Although there had been multiple wrong­ way driver events in San Antonio previous to this crash, the C H A P T E R 3 Active Countermeasures for Freeways

25 was driving up the ramp. More recent implementations of these systems in Fort Worth activate the flashing LEDs only when a wrong­way vehicle is detected. The more recent sys­ tems include multiple radars and a camera to reduce false alarms and provide visual confirmation of the wrong­way vehicle, respectively. Where the length and design of the exit ramp allowed, WRONG WAY signs with flashing red LEDs around the border supplemented the existing static WRONG WAY signs. On shorter ramps, the WRONG WAY signs with flashing red LEDs around the border replaced the existing static WRONG WAY signing. The battery for the signs was encased in the sign pole and charged by a small solar array attached to the top of the sign support. Data Sources Even before the task force was created, SAPD and Texas DOT implemented several procedures with regard to responding to WWD events. In August 2010, SAPD began to use an emer­ gency call signal (i.e., E­tone) for its radio network when a wrong­way driver was reported to 911. In January 2011, SAPD implemented a code in its computer­aided dispatch system that specifically identified all wrong­way driver events. Simi­ larly, in March 2011, Texas DOT TransGuide traffic manage­ ment center (TMC) operators began logging all WWD events, not just those that resulted in a crash. Through their involvement in the San Antonio WWD task force, both SAPD and Texas DOT shared the WWD sub­ component of their data logs for the border­illuminated WRONG WAY signing evaluation. However, the Texas DOT TransGuide operator logs generally included only the events that occurred in its coverage area—about half of the freeways in the San Antonio region. Researchers also extracted WWD­involved crash informa­ tion from the Texas DOT CRIS. However, because CRIS only documents WWD­related crashes, rather than any event where WWD activity is observed, the number of records was small compared to the SAPD 911 WWD call logs and the Texas DOT TransGuide operator WWD logs. Overall, researchers primarily used the SAPD 911 call logs for the statistical analysis because these logs began in January 2011, continued without interruption through April 2016, and contained more data points. Results Researchers performed a before­after evaluation with a control group to determine whether or not a meaningful reduction in WWD events was observed along the US 281 test corridor (Gross et al. 2010). The control data were all of the WWD events in the remainder of the city of San Antonio (but not including the US 281 test corridor). The before period was 14 months long (January 2011 to February 2012). Researchers could not test the comparability of the treatment and con­ trol datasets before the LED border­illuminated WRONG WAY signs were installed because only one complete year of WWD event data were available (i.e., SAPD began coding all wrong­way driver events in January 2011). The WRONG WAY signs with flashing red LEDs around the border were installed between March 2012 and June 2012. Researchers did not include data from this time period in the analysis because the traffic control devices in the corridor were in flux. The after period was 46 months long (July 2012 to April 2016). Table 13 shows the before and after WWD event data for the treatment and control groups in San Antonio. From these data, researchers computed an event modification factor (EMF) of 0.68 with a 95 percent confidence interval of 0.45 to 0.91. Because the confidence interval does not include 1.0, it can be stated with 95 percent confidence that the treatment had an effect. In terms of the expected change in events, the EMF indicates a 32 percent reduction in the WWD events on the US 281 test corridor after the installation of WRONG WAY signs with flashing red LEDs around the border at all exit ramps in the corridor. In other words, the WWD events at the exit ramps with the border­illuminated WRONG WAY signs were reduced by about one­third. This percentage change was statistically significant at a 5 percent significance Source: Texas A&M Transportation Institute. Figure 11. WRONG WAY sign with flashing red LEDs around border.

26 level (α = 0.05), and the 95 percent confidence interval was −55 percent to −9 percent. In 2014, Finley et al. (2014) found similar results for WRONG WAY signs with flashing red LEDs around the border based on 14 months of before data and 22 months of after data. At that time, researchers calculated a 38 percent reduction in WWD events on the US 281 corridor after the installation of the signs. This percent change was statistically significant at a 5 percent significance level, and the 95 percent confidence interval was −63 percent to −13 percent. While the more recent percent reduction in WWD events is slightly less (32 percent), the expanded dataset findings show the long­ term effectiveness of these signs. Florida Background Florida’s Turnpike Enterprise (FTE) has installed WWD countermeasures along part of its roadway network in South Florida. These countermeasures consist of border­illuminated WRONG WAY signs, radar, cameras, and communication abilities (see Figure 13). When the radar detects a wrong­way vehicle, the red LED lights flash before the vehicle reaches the sign. Once the vehicle reaches the sign, a series of photos is taken from a side camera and sent to FTE’s TMC and Florida Highway Patrol’s (FHP’s) command center. FTE installed these technologies at six interchanges (12 ramps) on the Source: Texas A&M Transportation Institute and map data © 2017 Google. US 281 US 281 Figure 12. US 281 LED border-illuminated WRONG WAY sign pilot test corridor.

27 Homestead Extension of Florida’s Turnpike (HEFT SR 821) and five interchanges (five ramps) on the Sawgrass Express­ way (SR 869). The HEFT also contains mainline detectors around the test interchanges. These 17 ramps are listed in Table 14 and depicted in Figure 14 and Figure 15. Note that north bound and southbound ramps are equipped on SR 821, while only southbound ramps are equipped on SR 869 (Al­Deek et al. 2016). Data Sources The research team examined the number of crashes, police citations, 911 calls, and system alerts before and after the instal­ lation of the border­illuminated WRONG WAY sign warn­ ing system. The before and after periods were March 2013 to Time Period Treatment Group (n = 96) Control Group (n = 1073) Before 46 396 After 135 1667 Primary computations: EMF = (ObservedT,A/ExpectedT,A)/(1+(VarianceExpectedT,A/ExpectedT,A^2)) = 0.68 95 Percent Confidence Interval = EMF±1.96*EMF Standard Error = 0.45 to 0.91 Percent reduction = (1 – EMF)*100 = 32 Additional computations: Comparison ratio = ObservedC,A/ObservedC,B = 4.20 ExpectedT,A = ObservedT,B*comparison ratio = 193.64 VarianceExpectedT,A = ExpectedT,A^2*(1/ObservedT,B+1/ObservedC,B+1/ObservedC,A) = 932.34 VarianceEMF = (EMF^2*[(1/ObservedT,A)+(VarianceExpectedT,A/ExpectedT,A^2)]/ [1+(VarianceExpectedT,A/ExpectedT,A^2)]^2 = 0.01 EMF Standard Error = SQRT(VarianceEMF) = 0.12 EMF = Event Modification Factor ObservedT,A = Observed Treatment After ExpectedT,A = Expected Treatment After VarianceExpectedT,A = Variance of the Expected Treatment After ObservedC,A = Observed Control After ObservedC,B = Observed Control Before ObservedT,B = Observed Treatment Before VarianceEMF = Variance of the EMF SQRT = Square Root Table 13. San Antonio WWD event data and computations. Source: University of Central Florida. Figure 13. FTE border-illuminated WRONG WAY sign warning system. Ramp Number Ramp Location and Direction 1 821 NB OFF 29—NW 41 2 821 SB OFF 29—NW 41 3 821 NB OFF 31—NW 74 ST 4 821 SB OFF 31—NW 74 ST 5 821 NB OFF 34—NW 106 6 821 SB OFF 34—NW 106 7 821 NB OFF 35—US 27 8 821 SB OFF 35—US 27 9 821 NB OFF 43—NW 57 10 821 SB OFF 43—NW 57 11 821 NB OFF 47—NW 27 12 821 SB OFF 47—University 13 869 SB OFF 1—Sunrise 14 869 SB OFF 3—Oakland 15 869 SB OFF 5—Commerce 16 869 SB OFF 8—Atlantic 17 869 SB OFF 11—Sample Table 14. FTE border-illuminated WRONG WAY sign warning system pilot test locations.

28 Source: University of Central Florida. Figure 14. HEFT (SR 821) FTE pilot test interchanges.

29 September 2014 (19 months) and November 2014 to May 2016 (19 months), respectively. FTE installed the systems in October 2014. The analysis used a set of control interchanges without the countermeasures on the FTE system for com­ parison with the pilot test interchanges. There was a total of 29 control interchanges located on SR 821, SR 869, and SR 91 in Miami­Dade and Broward Counties. Table 15 contains a summary of the data sources for the before and after periods for the pilot test interchanges, a ran­ dom sample of the control interchanges, and all of the control interchanges. The WWD events shown in this table are events that occurred in the vicinity of either the test or control inter­ changes. This does not mean that all WWD events originated at these interchanges but that these interchanges were near­ est to the events and therefore the most likely points of ori­ gin. Table 15 shows that there was a small number of recorded crashes and citations during the study period. As with the Texas data, the research team focused on the WWD 911 calls that were reported to TMC for statistical analysis because the entire dataset contained all WWD events and many more data points. Source: University of Central Florida. Figure 15. Sawgrass Expressway (SR 869) FTE pilot test interchanges.

30 Results To assess the effectiveness of the border­illuminated WRONG WAY signs in Florida, the research team performed a before­after evaluation with a control group to determine whether or not a meaningful reduction in WWD events was observed at the pilot test locations (Gross et al. 2010). The control group contained 11 similarly designed interchanges on the FTE system in South Florida near the test ramps (see Table 16). The research team conducted a test of comparabil­ ity for the treatment group and potential control group to assess the suitability of the candidate control group (Hauer 1997). Table 17 shows the test of comparability dataset for 4 years and odds ratio computations. The mean of the indi­ vidual odds ratios was 0.91, which is reasonably close to 1.0, and the standard deviation was 0.13. The 95 percent confi­ dence interval was 0.66 to 1.16. Because the mean was close to 1.0 and the 95 percent confidence interval included 1.0, the research team concluded that the control group was suitable. Table 18 shows the before and after WWD event data for the treatment and control groups in South Florida. From these data, researchers computed an EMF of 0.81 with a 95 percent confidence interval of 0.19 to 1.43. In terms of the expected change in events, the EMF indicates a 19 percent reduction in the WWD events at the treatment sites after the installa­ tion of WRONG WAY signs with flashing red LEDs around the border. However, because the confidence interval includes 1.0, it cannot be stated with 95 percent confidence that the treatment had an effect. WWD Event Statistics The research team also used the WWD event data on free­ ways in Texas and South Florida to examine some general characteristics of these incidents. In particular, the research team wanted to provide insight into the distance the wrong­ way driver traveled and the time span of the WWD events. In order to determine these two items of interest, the research team needed to identify WWD events with more than one call to 911. Out of the 2428 WWD events that occurred in San Antonio, Texas, between January 1, 2011, and April 30, 2016, 607 were multi­call events. In South Florida between January 1, 2006, and April 30, 2016, there were 283 multi­call events out of 1485 WWD events. Figure 16 shows the distance the wrong­way driver trav­ eled for the multi­call WWD events, and Figure 17 shows the time span (i.e., duration) of each multi­call WWD event. Since Miami-Dade and Broward Counties Before Period Crashes After Period Crashes Before Period Citations After Period Citations Before Period 911 Calls After Period 911 Calls Test Interchanges (n = 11) 1 1 3 5 16 24 Random Control Interchanges (n = 11) 1 2 0 4 15 22 Total Control Interchanges (n = 29) 4 5 10 14 37 51 Total (n = 40) 5 6 13 19 53 75 Table 15. Summary of SunGuide WWD events for test and control sites. Treatment Location Treatment Interchange Geometry Control Location Control Interchange Geometry Exit 29 on SR 821 Partial Cloverleaf Exit 5 on SR 821 Partial Cloverleaf Exit 31 on SR 821 Three-Leg Directional Exit 16 on SR 821 Split Diamond Exit 34 on SR 821 Three-Leg Directional Exit 17 on SR 821 Two-Leg Directional Exit 35 on SR 821 Partial Cloverleaf Exit 20 on SR 821 Partial Cloverleaf Exit 43 on SR 821 Partial Cloverleaf Exit 23 on SR 821 Partial Cloverleaf Exit 47 on SR 821 Partial Cloverleaf Exit 25 on SR 821 Partial Cloverleaf Exit 1 on SR 869 Full Diamond Exit 39 on SR 821 Two-Leg Directional Exit 3 on SR 869 Full Diamond Exit 14 on SR 869 Full Diamond Exit 5 on SR 869 Full Diamond Exit 15 on SR 869 Full Diamond Exit 8 on SR 869 Full Diamond Exit 19 on SR 869 Full Diamond Exit 11 on SR 869 Full Diamond Exit 53 on SR 91 Full Diamond Table 16. Treatment and control group ramps.

31 Group Year 1 (October 2010– September 2011) Year 2 (October 2011– September 2012) Year 3 (October 2012– September 2013) Year 4 (October 2013– September 2014) Treatment 11 12 8 11 Control 8 9 6 11 Primary computations: Years 1 & 2 Odds Ratio = ((TreatmentB*ControlA)/(TreatmentA*ControlB))/(1+ (1/TreatmentA)+(1/ControlB)) = 0.85 Years 2 & 3 Odds Ratio = 0.81 Years 3 & 4 Odds Ratio = 1.06 Mean Odds Ratio = 0.91 Standard Deviation = 0.13 95 Percent Confidence Interval = Mean Odds Ratio±1.96*Standard Deviation = 0.66 to 1.16 TreatmentB = Total WWD events for the treatment group in year i TreatmentA = Total WWD events for treatment group in year j ControlB = Total WWD events for control group in year i ControlA = Total WWD events for control group in year j Table 17. WWD events before treatment was installed. Time Period Treatment Group (n = 43) Control Group (n = 37) Before 18 15 After 25 22 Primary computations: EMF = (ObservedT,A/ExpectedT,A)/(1+(VarianceExpectedT,A/ExpectedT,A^2)) = 0.81 95 Percent Confidence Interval = EMF±1.96*EMF Standard Error = 0.19 to 1.43 Percent reduction = (1 – EMF)*100 = 19 Additional computations: Comparison ratio = ObservedC,A/ObservedC,B = 1.47 ExpectedT,A = ObservedT,B*comparison ratio = 26.4 VarianceExpectedT,A = ExpectedT,A^2*(1/ObservedT,B+1/ObservedC,B+1/ObservedC,A) = 116.86 VarianceEMF = (EMF^2*[(1/ObservedT,A)+(VarianceExpectedT,A/ExpectedT,A^2)]/ [1+(VarianceExpectedT,A/ExpectedT,A^2)]^2 = 0.10 EMF Standard Error = SQRT(VarianceEMF) = 0.32 EMF = Event Modification Factor ObservedT,A = Observed Treatment After ExpectedT,A = Expected Treatment After VarianceExpectedT,A = Variance of the Expected Treatment After ObservedC,A = Observed control after ObservedC,B = Observed control before ObservedT,B = Observed treatment before VarianceEMF = Variance of the EMF SQRT = Square Root Table 18. FTE WWD event data and computations. these distances and times are based on calls to 911, the actual distance traveled by the wrong­way driver and the duration of the event could be longer. However, the research team believes these data provide critical insight into the character­ istics of WWD events. Figure 16 does not include the multi­ call WWD events where the reported location of each call was the same (25 percent and 65 percent for Texas and South Florida, respectively). Figure 17 does not include the multi­ call WWD events where the reported time of each call was the same (11 percent and 63 percent for Texas and South Florida, respectively). According to Figure 16, the distance the wrong­way driver traveled for the multi­call WWD events was similar between the two datasets. The only exceptions appear to be that there were fewer wrong­way drivers that traveled 1 mi or less and more wrong­way drivers that traveled 15 mi or more in South Florida. Overall, approximately two­thirds of the wrong­way drivers traveled 3 mi or less. However, this figure does show

32 Figure 16. 911 multi-call event distance. Figure 17. 911 multi-call event time span.

33 that wrong­way drivers can travel 10 mi or more (5 percent). The maximum distance in the dataset was 34 mi in South Florida. On average, the wrong­way driver traveled 3.6 mi. Figure 17 shows that the durations of the wrong­way events were also similar between the two datasets. Once again, the main differences appear at the extremities. Overall, 57 percent of the multi­call events occurred within a 3­minute time period. In addition, only 5 percent of the multi­call WWD events lasted 15 minutes or longer. The maximum duration in the dataset was 53 minutes in South Florida. The average duration of a WWD event based on multiple calls to 911 was about 5 minutes. Overall, these data show that WWD events occur quickly and over relatively short distances, limiting the amount of time public agency personnel and law enforcement have to respond to the event. The research team also used the multi­call subset of both datasets to identify trends regarding the percent of WWD events by month, day of the week, and hour of the day. Figure 18 shows the percent of multi­call WWD events by month. It appears that the percent of events is relatively consistent throughout the year. Figure 19 shows the percent of multi­call WWD events by day of the week. While WWD events occur throughout the week, the percent of events increases on Friday and peaks on Saturday and Sunday (weekend). The events over these 3 days of the week account for 60 percent of the multi­call WWD events. Figure 20 shows that most of the WWD events are reported between midnight and 5:00 a.m. (64 percent). The peak around the 2:00 a.m. hour (the typical time for establishments that serve alcohol to close) is more pronounced in Texas than in South Florida. Red RFBs Above and Below WRONG WAY Signs The research team evaluated the effectiveness of red RFBs above and below WRONG WAY signs at freeway ramps using WWD detection data from the Central Florida Expressway Authority (CFX) in Florida. Red RFBs were implemented at five initial ramps starting in February 2015. The limited num­ ber of sites (five) and recent implementation (2015) made it difficult to analyze the effectiveness of these devices using 911 calls, citations, and crash data. Therefore, researchers were able to perform only a preliminary analysis on the WWD detec­ tions captured by these devices. A detailed statistical analysis can be conducted once enough data are collected from these and additional implementation sites. The sections below pro­ vide a brief history of the red RFB implementations in the greater Orlando, Florida, region and document the findings of the analysis. Figure 18. 911 multi-call events by month.

34 Figure 19. 911 multi-call events by day of week. Figure 20. 911 multi-call events by hour of day.

35 Background The first WWD study, conducted by the University of Cen­ tral Florida for CFX in 2013, showed that there was a need to detect and deter WWD on CFX’s roadways (Al­Deek et al. 2013). To achieve this objective, University of Central Florida researchers suggested a variety of WWD countermeasures with a hierarchy of technology and differing levels of inter­ vention. One of these countermeasures was the concept of using WRONG WAY signs equipped with red rectangular rapid flashing beacons (RRFBs). Based on this suggestion, CFX submitted a request to experiment with red RRFBs as a WWD countermeasure to FHWA. This request to experi­ ment was approved by FHWA in October 2014. While early documentation of the system referred to the devices as red RRFBs, further investigation into the actual red flashing pattern revealed that it was different than the approved pedestrian RRFB indications (i.e., 800­millisecond flash cycle length with a 2–5 or wig­wag plus simultaneous flash [WW+S] pattern). The red flashing pattern cycle length is 20 seconds with the following wig­wag pattern (also see Table 19): • The top left side and bottom right side beacons are on for 500 milliseconds. • The top right side and bottom left side beacons are on for 500 milliseconds. Thus, the duty cycle is 1 second and each beacon is on for 50 percent of its duty cycle. Because the red flashing pattern did not contain a rapid flash component, the research team refers to the WWD countermeasure as red RFBs in this report. CFX chose to pilot test red RFBs at the following five ramps based on the results of the first WWD study (Al­Deek et al. 2013, 2015): • SR 528 and SR 520—one off ramp (eastbound off ramp). • SR 408 and Hiawassee Road—two off ramps. • SR 408 and Kirkman Road—two off ramps. Figure 21 shows the general location of the sites. Figure 22 shows the SR 528 eastbound off ramp pilot site, with the ramp highlighted in red. CFX and its contractors installed and tested the first sets of WRONG WAY signs with RFBs at this site. The implementation date for this site was Febru­ ary 21, 2015. Figure 23 shows the four test sites on SR 408, with the ramps highlighted in red. The implementation date for these four sites was June 14, 2015. Additionally, CFX decided to improve its WWD detection capabilities to more accurately determine the frequency of WWD on its roadways. The RFB WRONG WAY sign setup consists of one pair of signs (one sign on each side of the road) near the ramp inter­ section with the frontage road (first set of signs) and one pair of signs closer to the mainline (second set of signs). Figure 24 shows the components of the RFB sign system. All four sign assemblies at each site have a WRONG WAY sign with one red RFB above the sign and one below the sign. In addition, the right­hand sign (from the wrong­way driver’s point of view) contains two radar detectors, two cameras, and a cellu­ lar modem with antenna. One radar and camera face forward (toward the approaching wrong­way driver), and the other radar and camera face toward the side (to detect if the vehicle passes the first set of signs). The cellular modem provides text alerts and email notifications to the regional TMC through the manufacturer’s software application for any wrong­way vehicles that pass the first set of signs, allowing the regional TMC operators to confirm the wrong­way vehicle and notify law enforcement (Al­Deek et al. 2015). For the RFBs to be activated, a wrong­way vehicle must enter the range of the forward­facing radar (which is typi­ cally 100 ft). Once a wrong­way vehicle is detected by this Cumulative Time in Milliseconds (ms) Top Left Top Right Bottom Left Bottom Right 100 X X 200 X X 300 X X 400 X X 500 X X 600 X X 700 X X 800 X X 900 X X 1000 (1 second) X X On time (ms) 500 500 500 500 Percentage of cycle beacon is on 50% 50% 50% 50% Note: X = beacon is on for 100 ms Table 19. Red RFB flash pattern.

36 Source: University of Central Florida. Figure 21. Red RFB pilot study sites. Source: University of Central Florida. Figure 22. SR 528 pilot study site details.

37 radar, the RFBs on all four signs are activated and flash in a wig­wag pattern to get the wrong­way driver’s attention. At the same time, the forward­facing camera (first camera) captures a series of images of the wrong­way vehicle. If the driver continues driving the wrong way past the first set of signs, the side radar detection zone is entered. Once this happens, the side camera (second camera) takes a series of images to confirm that the vehicle is still traveling the wrong way. The regional TMC is also alerted at this time via a text alert (Al­Deek et al. 2015). Initial field testing of the red RFB WRONG WAY sign system and mainline detection in January 2015 found that there were some detection issues with the red RFB WRONG WAY sign system. The second camera had difficulty captur­ ing images of small vehicles traveling at high speeds. UCF researchers suggested that modifications be made to the cam­ eras to improve their reliability in capturing images of these vehicles. Eventually, additional tests in June 2015, fall 2015, and spring 2016 improved the accuracy for both the sign activation/first camera and second camera to 100 percent Source: University of Central Florida. Figure 23. SR 408 pilot study site details. Source: Central Florida Expressway Authority. Figure 24. Red RFB WRONG WAY sign assembly components.

38 during field tests. Communication between the field devices and the regional TMC was also good, at an average of 7 sec­ onds between the WWD detection and the regional TMC receiving the WWD alert. However, UCF researchers noted that this time can be shortened or eliminated when the devices are integrated within SunGuide® (statewide Florida traffic management software) (Al­Deek et al. 2015). Toll road customers, law enforcement officers, and Florida DOT officials were also surveyed about the red RFB devices and asked if they preferred the red LED border­illuminated signs implemented in South Florida or the red RFB signs. Nine hundred FTE toll road drivers, 247 FHP troopers, and five Florida DOT officials were shown two short videos, one of the red LED border­illuminated signs and one of the red RFB signs. These videos showed the devices in action as they would appear to a wrong­way driver. Over 70 percent of both surveyed groups (toll road drivers and law enforcement/ transportation professionals) preferred the red RFBs over the red LEDs due to the wig­wag flashing pattern and the extra set of red RFB signs installed by CFX (Al­Deek et al. 2015). Data Source The research team obtained information on the total num­ ber of WWD detections and whether the wrong­way vehi­ cle turned around from February 2015 through July 2016 (18 months) for the five test sites previously described. The WWD detections were classified into two main categories (i.e., WWD acts and WWD events) and three turnaround types (i.e., confirmed, probable, and unknown). A WWD act occurs when a passenger vehicle is driving the wrong way on the exit ramp. It does not include vehicles reversing on the exit ramp; bicyclists traveling the wrong way on the exit ramp; or lawn mowers, utility vehicles, or emergency vehicles driving the wrong way on the exit ramp. The definitions of true WWD events, confirmed turnarounds, probable turn­ arounds, and unknown turnarounds, as well as how the total number of WWD acts is related to the three different types of turnarounds, are as follows: • True WWD events trigger both oncoming and side cam­ eras, whereas WWD acts only have to trigger the oncoming camera. Every WWD event is a WWD act, but not every WWD act results in a WWD event because some WWD acts only trigger the oncoming camera. • Confirmed turnarounds are WWD acts that have images that clearly show the wrong­way driver starting to turnaround. • Probable turnarounds are WWD acts that do not trigger the side camera. In these cases, the images do not clearly show the driver turning around. However, because the vehi­ cle does not trigger the side camera, it is likely the driver turned around. • Unknown turnarounds are WWD events where the wrong­way vehicle triggers the side camera but does not seem to be slowing down or starting to turnaround. As none of these unknown turnarounds resulted in a crash or the driver being apprehended by law enforcement, it is assumed that the driver had turned around and corrected his or her WWD act at some point on the CFX system. • The sum of confirmed turnarounds + probable turn­ arounds + unknown turnarounds is equal to the total amount of WWD acts. Results Table 20 contains a list of the WWD detections at the five sites with RFBs from February 21, 2015, through July 31, 2016. In total, 23 WWD detections occurred over this 527­day period. The research team found that only six detections (26 percent) were deemed to be true WWD events (i.e., both oncoming and side cameras triggered). For the remaining 17 WWD acts (i.e., only the oncoming camera was triggered), five (29 percent) were confirmed turnarounds and 12 (71 per­ cent) were probable turnarounds. Most of the detections (65 percent) occurred at night. In addition, 87 percent involved a passenger car. Summary This chapter documented the research team’s evaluation of the effectiveness of LED border­illuminated WRONG WAY signs and a red RFB WRONG WAY sign system at freeway exit ramps. In terms of the expected change in WWD events, the Texas data indicated a 32 percent reduction after the imple­ mentation of LED border­illuminated WRONG WAY signs. This finding was statistically significant at a 95 percent con­ fidence level and shows the long­term effectiveness of these signs. In Florida, a 19 percent reduction in WWD events was found after the implementation of LED border­illuminated WRONG WAY signs; however, this finding was not statisti­ cally significant (most likely due to the small sample size). Based on the 911 multi­call dataset, researchers found that, on average, WWD events lasted 5 minutes and the wrong­ way driver traveled 3.6 mi. The multi­call dataset also con­ firmed that WWD events occur more on weekends and in the early morning hours (i.e., midnight to 5:00 a.m.). CFX implemented the first red RFBs in conjunction with WRONG WAY signs and detection systems in February 2015 at one ramp. In June 2015, CFX installed RFBs at four addi­ tional ramps. Based on the available dataset, it appears that at least three­quarters of the WWD detections resulted in the errant driver self­correcting before reaching the main lanes. These promising findings led CFX to install red RFBs

39 at 30 other ramps in 2016 and 2017, resulting in a total of 35 ramps equipped with red RFB WRONG WAY signs. These additional ramps will provide more data that can be used in future analysis. The pilot study of red RFB WRONG WAY signs provided some important lessons learned that can be applied to future applications of these devices. The second set of signs with the second camera are important to help determine whether wrong­way vehicles continue traveling or stop before reach­ ing the second set of signs. To understand even better whether Site Date Time Time of Day Vehicle Type Detection Type Turned Around? SR 528 EB Exit 31 2/21/2015 11:59:29 PM Night Car WWD Act & WWD Event Unknown SR 408 EB Exit 4 6/24/2015 6:33:58 AM Day Car WWD Act Confirmed SR 408 EB Exit 4 6/28/2015 3:31:24 AM Night Car WWD Act Probable SR 408 EB Exit 4 8/16/2015 7:52:08 PM Day Car WWD Act Probable SR 408 EB Exit 5 8/31/2015 12:18:07 AM Night Car WWD Act Probable SR 408 WB Exit 5 9/18/2015 1:17:29 AM Night Car WWD Act & WWD Event Unknown SR 408 EB Exit 5 9/26/2015 8:21:01 PM Night Car WWD Act Confirmed SR 408 WB Exit 5 10/4/2015 3:11:06 PM Day Car WWD Act Confirmed SR 408 EB Exit 5 10/29/2015 12:22:53 PM Day Car WWD Act Probable SR 408 WB Exit 5 1/15/2015 9:28:23 PM Night Car WWD Act Probable SR 408 WB Exit 4 11/22/2015 12:16:42 PM Night Car WWD Act Probable SR 408 EB Exit 5 1/24/2016 7:43:52 PM Night Car WWD Act Probable SR 528 EB Exit 31 1/31/2016 4:08:27 AM Night Car WWD Act Probable SR 408 WB Exit 4 1/31/2016 6:48:38 AM Night Car WWD Act & WWD Event Unknown SR 528 EB Exit 31 2/18/2016 2:35:53 AM Night Car WWD Act Probable SR 408 WB Exit 4 3/1/2016 1:07:03 AM Night Car WWD Act & WWD Event Unknown SR 408 WB Exit 4 3/6/2016 12:51:25 PM Day Car WWD Act & WWD Event Unknown SR 408 EB Exit 4 3/27/2016 2:40:21 AM Night Car WWD Act & WWD Event Unknown SR 408 WB Exit 4 5/7/2016 6:30:18 PM Day Car WWD Act Confirmed SR 408 WB Exit 4 6/10/2016 9:30:09 AM Day Truck WWD Act Probable SR 408 WB Exit 4 6/24/2016 5:14:10 PM Day Car WWD Act Probable SR 408 WB Exit 4 7/8/2016 6:00:50 AM Night Truck WWD Act Probable SR 408 EB Exit 4 7/8/2016 6:03:02 AM Night Truck WWD Act Confirmed Table 20. List of all WWD detections since RFBs installed at five CFX sites (2/21/15–7/31/16). these devices are stopping wrong­way vehicles from enter­ ing the highway, a third camera could be installed on the highway near the exit gore to see if any detected wrong­way vehicles continue onto the highway. The placement of the signs is also important to ensure that the radar detectors can detect wrong­way vehicles and activate the RFBs before the vehicle reaches the sign. Finally, it is important to have minimal communication delay between the devices and the regional TMC to reduce the response time to detected WWD events.