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.
57 6.0 Field Evaluation of Temporary Ramp Metering Many regions have deployed, sustained, and expanded ramp metering to improve daytime traffic operations on freeways. It is a proven and efficient tool to address traffic congestion and safety issues; however, it is a challenge to use under work zone conditions. Agency support and project costs also pose difficulties for state/local transportation agencies. There are very few state DOTs (Minnesota and Pennsylvania) that use ramp metering to improve the overall work zone mobility and safety and even then, it is rarely used. There is only one comprehensive study on the use of ramp metering in work zones and that evaluation was performed during off-peak conditions for the Missouri Department of Transportation. The focus of this report is to present results from the evaluation of ramp metering conducted during peak conditions in the United States and add to the body of knowledge on available strategies for improving mobility and safety in work zones. 6.1. Site Selection and Characteristics Through outreach efforts to state transportation agencies, the research team identified the following locations as test sites for evaluation: ⢠MN Route 52 Bridge Deck Replacement Project, Rochester, Minnesota. ⢠I-279 Parkway North Improvement Project, Ohio Township, Allegheny County, Pennsylvania. 6.1.1 MN Route 52 Bridge Deck Replacement Project, Rochester, Minnesota This project evaluated the effectiveness of ramp metering in a work zone setting on a bridge deck replacement project. The project involved replacing the existing concrete bridge decks on the US 52 bridges over US 63 in Rochester. Additionally, Minnesota Department of Transportation (MnDOT) upgraded the US 52 northbound lane additions near the end of the bridge. Construction narrowed US 52 traffic from two lanes in each direction to a single lane in each direction (Figure 18). MnDOT activated a ramp meter on the northbound US 63 entrance loop to northbound US 52 to help regulate traffic flow during construction. The ramp metering was in place from April 18 to July 1, 2016. The ramp meter operated only on weekdays between 7:30 to 8:30 a.m. and 4:00 to 5:00 p.m. MnDOT chose to use ramp metering because of the exceedingly high ramp volumes averaging more than 900 vehicle per hour (vph) during the morning peak period (ramp ADT approximately 12,600), with corresponding mainline morning peak volumes of close to 1,100 vph (mainline ADT approximately 17,500). Overall, the ramp contributes as much as 42 percent of total traffic volume at this location. The work zone posted speed limit was 55 mph.
58 The northbound US 63 entrance loop at the point of merging with Highway 52 was allowed to form two lanes when merging. However, metering allowed only one car per green phase to merge onto the mainline. In addition to the appropriate ramp metering signs, MnDOT placed advance SIGNAL AHEAD warning signs (W3-3) on the ramp. The signal was roadside- mounted at a height of 10 ft with an 8-in. green, yellow, red lens/housing assembly (Figure 19). The team collected data at four locations along the mainline (Route 52) near the interchange of MN Route 52/Route 63: ⢠Location 1, 2,600 ft upstream of on-ramp. ⢠Location 2, 800 ft upstream of on-ramp. ⢠Location 3, 400 ft downstream of on-ramp. ⢠Location 4, 3,000 ft downstream of on-ramp. The team also installed two cameras at the northbound entering ramp to monitor traffic operations with and without ramp metering. The first camera installation was approximately 100 ft east of the gore area to monitor vehicle merging and weaving. The other camera installation was at the mid-point of the metered ramp to capture driversâ ramp-metering compliance. Figure 19 shows the ramp geometrics and layout of the data collection locations.
59 Figure 18. Ramp-metering data collection locations on MN Route 52 and Route 63 loop b ramp, Rochester, Minnesota.
60 Figure 19. MnDOT ramp-control signal details. 6.1.2 I-279 Parkway North Improvement Project, Ohio Township, Allegheny County, Pennsylvania This project included concrete patching and overlay, preservation of 30 bridges and 49 overhead sign structures, repairs to 29 walls, repairs to ramps, improving lighting, repairing HOV lanes, updating signs, improving guardrail and drainage, and installing an anti-icing system on McKnight Road. In the northbound direction, work included reconfiguring the ramp from northbound I-579 to northbound I-279, lengthening the northbound Perrysville and Madison on-ramps, and paving on Route 28 between Anderson Street and Chestnut Street. In the southbound direction, traffic was crossed over into the northbound lanes at the Camp Horne (Exit 8) interchange. Both southbound lanes were shifted into the HOV lanes at the Perrysville Avenue (Exit 5) interchange before reentering mainline I-279 south of McKnight Road. The project continued through June 2019.
61 To evaluate the effectiveness of ramp metering in a work zone, Pennsylvania Department of Transportation (PennDOT) revised its approved MOT plans to include ramp metering at Union Avenue entrance ramp to southbound I-279. This was made possible through a combined effort of the state DOT, FHWA, and the research team. The Union Avenue entrance ramp to southbound I-279 (two lanes) was best suited for metering based on an operational assessment conducted by PennDOT. The assessment indicated that ramp metering may help to regulate traffic flow through the heavily congested corridor during construction. Ramp volumes peaked at 450 vph with mainline volumes at 1,200/1,750 vph for the right lane and left lane, respectively. The ramp metering was in place April 23âAugust 26, 2018. The ramp meter operated weekdays from 6:00â9:00 a.m. and 4:00â6:00 p.m.. Metering allowed only one car per green phase to merge onto the mainline. In addition to the appropriate ramp-metering signs, PennDOT placed advance SIGNAL AHEAD warning signs (W3-3) on the ramp. The signal support was roadside- mounted with the signal cantilevered over the road at a height of 14 ft with a green/red lens/housing assembly (Figure 20). The team collected data at four locations along the mainline (I-279) near the interchange of Union Avenue: 1. Location 1, 5,250 ft upstream from the southbound entering ramp (gore area). 2. Location 2, 1,300 ft upstream from the southbound entering ramp (gore area). 3. Location 3, 1,150 ft downstream from the southbound entering ramp (gore area). 4. Location 4, 5,250 ft downstream from the southbound entering ramp (gore area). The team also installed two cameras at the southbound entering ramp to monitor traffic operations with and without ramp metering. The first camera installation was approximately 600 ft south of the gore area to monitor vehicle merging and weaving. The second camera was installed at the mid-point of the metered ramp to capture driversâ compliance with the ramp metering. Figure 20 shows the ramp geometrics and layout of the data collection locations.
62 Figure 20. Ramp-metering data collection locations on I-279 and Union Avenue Ramp, Ohio Township, Pennsylvania.
63 6.2. Study Methodology 6.2.1 Data Collection Duration MN Route 52, Rochester, Minnesota. The team collected data for MN Route 52 westbound on weekdays during three periodsâmeter on (fixed-cycle length), meter off, and meter on (variable-cycle length). Data were collected with meter on (fixed-cycle length) from May 17 to May 20, and with the meter off from May 21 to May 29, 2016. The team then changed the meter flow rate algorithm and collected data again with meter on (variable-cycle length) from May 30 to June 3, 2016. The ramp meter was set for operation during peak hours (a.m. peak, 7:30â8:30 a.m.) and (p.m. peak, 4:00â5:00 p.m.). I-279, Ohio Township, Pennsylvania. The team collected data for I-279 southbound (both right lane and left lane) on weekdays during three periodsâmeter off, meter on (fixed-cycle length), meter on (variable-cycle length). Meter-off data were collected from April 23 to May13, 2018, and with Meter on (fixed-cycle length) from May 14 to June 4, 2018. The team collected meter on (variable-cycle length) data from June 5 to August 26, 2018. The ramp meter was set for operation during peak hours (6:00 a.m. to 9:00 a.m.) and (4:00 p.m. to 6:00 p.m.). Because the most congested peak period at both study areas was a.m. peak, the traffic analyses of this study focused on the a.m. peak period only. Ramp traffic volumes in p.m. peak period were insignificant. 6.2.2 Data Collection Procedures The team used Wavetronix sensors to collect vehicular data. Wavetronix sensors collect data by emitting a microwave radar beam. The sensors were trailer-mounted and stationed perpendicular to the roadway, outside the clear zone; as vehicles pass through the beam, the sensor detects the reflected microwave beam. The sensors can detect volume, vehicle classification, speed, 85th percentile speeds, and vehicle gaps across multiple lanes (up to 200 ft). All vehicular data were collected by direction and by lane. The team measured the following data per lane: volume, speed, vehicle classification, headway, and gap. The data were collected in 1-minute bins. The team screened all raw data to exclude missing data values and outliers such as vehicles traveling at very low or very high speeds. Data were analyzed separately for passenger cars and commercial vehicles. 6.2.3 Measures of Effectiveness The team evaluated the following operational MOEs:
64 ⢠Vehicle speeds along mainline with and without ramp metering. Because the intent is that ramp metering will control the flow rate of vehicles entering the main line, this treatment may increase vehicle operating speeds on the mainline. ⢠Travel time through the work zone with and without ramp metering. Travel time through a work zone is a measurement to determine the operational effect of implementing ramp metering. In this study, the team used locations 1 and 3 as the reference points to determine the travel time. As stated previously, because the intent is that ramp metering will control the flow rate of vehicles entering the main line, this treatment may reduce vehicle travel time through the corridor. ⢠Merging headways with and without ramp metering. Vehicle headway is a measure of the temporal space between two vehicles. As the average of vehicle headways is the reciprocal of flow rate, vehicle headways represent microscopic measures of flows passing a point. To some extent, the minimum acceptable mean headway determines the roadway capacity. ⢠Driver compliance rates. Visually process videos of ramp traffic. A vehicle is said to have complied with the ramp metering if it went through when the signal display is green. 6.2.4 Method for Statistical Test for Vehicle Speeds The method for statistical test for vehicle speeds is the same as described in Section 5.2.5. 6.2.5 Method for Statistical Test for Travel Time through the Work Zone Similar to the analysis of the vehicle speeds, the team used the t-test to compare the differences between the travel time through the work zone with and without the ramp metering. For the travel time analysis, the null and alternative hypotheses for testing the differences in two population travel time measures, μ1 and μ2, were: ⢠Null Hypothesis (H0): There has been no change in mean travel time as a result of ramp metering, or H0: μ1 â μ2 = 0. ⢠Alternative Hypothesis (Ha): There has been a change in mean travel time as a result of ramp metering, or Ha: μ1 â μ2 > 0. At the study site, a t-statistic was calculated during data collection periods. Equation 3 in Section 5.2.5 shows the equation used to apply independent two-sample t-statistics to test for the difference between two sample means.
65 6.2.6 Method for a Statistical Test for Frequency of Headway The statistical test for vehicle speeds is the same as described in Section 5.2.6. 6.2.7 Driver Compliance of Ramp Meter Signal This analysis is to determine the percentage of vehicles that complied with the ramp metering signal. The team used videos, which captured ramp driver behavior to determine the number of vehicles traveling through the ramp when the signal display was green compared to the total number of vehicles going through the ramp during the test periods. The team determined the compliance rate by the number of vehicles in compliance divided the total number of vehicles traveling through the ramp during the test periods. 6.3. Comparison of Results for Vehicle Speeds The team compared the vehicle speeds to evaluate the effect of ramp metering and to determine if ramp metering caused changes in travel characteristics. In this study, two ramp metering scenarios (Fixed-cycle length and Variable-cycle length) were compared to the without scenario. Implementing ramp metering was expected to have a positive effect on vehicle speeds on the mainline of a freeway at the vehicle-merging area. For this reason, the team analyzed vehicle speeds at the merge areas of the ramp and the mainline of the freeway. 6.3.1 MN Route 52, Rochester, Minnesota The a.m. peak (6:00 to 9:00 a.m.) is the most congested peak period at the study area; therefore, the traffic analyses of this study focused on a.m. peak hour only. The team conducted an analysis of vehicle speeds on the mainline of the freeway at the merging area (location 3) during the a.m. peak period. Figure 21 illustrates the vehicle speeds and traffic volumes on the main line for all scenarios. In general, it appears that ramp metering improved performance on the main line (i.e., vehicle speeds were higher for all meter-on scenarios (Options 1 and 2) during the a.m. peak hour period). What is also evident is that main line saturation was reached around 6:45 a.m., resulting in traffic flow breakdown prior to the ramp meter being turned on at 7:30 a.m. (to 8:30 a.m.). As expected, it took the ramp metering about 5â10 minutes to stabilize the traffic flow before any benefit could be realized, such as improvements in vehicle speed and travel time. However, greater benefits may have been realized if the ramp metering went into effect at 6:00 a.m., before the right lane flow reached its capacity of a combined 1,500â1,600 vehicles (with ramp volumes approaching 400â600 vehicles) per hour. This is much lower than the traditional traffic right lane volumes typically recommended for ramp metering at approximately 2,000 vph inclusive with ramp volumes approaching 400 vehicles per hour.
66 Figure 21. A.M. peak hour vehicle speed and traffic volumes at Location 3. 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 6: 00 :0 0 AM 6: 04 :0 0 AM 6: 08 :0 0 AM 6: 12 :0 0 AM 6: 16 :0 0 AM 6: 20 :0 0 AM 6: 24 :0 0 AM 6: 28 :0 0 AM 6: 32 :0 0 AM 6: 36 :0 0 AM 6: 40 :0 0 AM 6: 44 :0 0 AM 6: 48 :0 0 AM 6: 52 :0 0 AM 6: 56 :0 0 AM 7: 00 :0 0 AM 7: 04 :0 0 AM 7: 08 :0 0 AM 7: 12 :0 0 AM 7: 16 :0 0 AM 7: 20 :0 0 AM 7: 24 :0 0 AM 7: 28 :0 0 AM 7: 32 :0 0 AM 7: 36 :0 0 AM 7: 40 :0 0 AM 7: 44 :0 0 AM 7: 48 :0 0 AM 7: 52 :0 0 AM 7: 56 :0 0 AM 8: 00 :0 0 AM 8: 04 :0 0 AM 8: 08 :0 0 AM 8: 12 :0 0 AM 8: 16 :0 0 AM 8: 20 :0 0 AM 8: 24 :0 0 AM 8: 28 :0 0 AM 8: 32 :0 0 AM 8: 36 :0 0 AM 8: 40 :0 0 AM 8: 44 :0 0 AM 8: 48 :0 0 AM 8: 52 :0 0 AM 8: 56 :0 0 AM Vo lu m e (V eh icl e/ M in ) Sp ee d (M PH ) Axis Title A.M. Average Speed and Traffic Volumes (at Location 3, 1 Minute Bin). Option 1 (Fixed Cycle Length) - Speed Meter Off (Without Ramp Metering) - Speed Option 2 (Variable Cycle Length) - Speed Option 1 (Fixed Cycle Length) - Volume Meter Off (Without Ramp Metering) - Volume Option 2 (Variable Cycle Length) - Volume Poly. (Option 1 (Fixed Cycle Length) - Volume) Poly. (Meter Off (Without Ramp Metering) - Volume) Poly. (Option 2 (Variable Cycle Length) - Volume)
67 The changes in the mean speeds and 85th percentile speeds for vehicles with and without ramp- metering scenarios were calculated. Tables 24 and 25 show the comparison of mean speed and 85th percentile speed on the mainline and the statistical test results with and without ramp metering. The following section discusses both meter-on scenarios (fixed time vs. variable time). 6.3.1.1 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering As Table 24 shows, the mean speeds of vehicles on the mainline of the freeway increased for all time periods prior to 8:15 a.m. Table 24. Speed comparison, meter-off and fixed-cycle length ramp metering. Meter Off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 1 Meter off Option 1 Meter off Option 1 Meter off Option 1 Volume (Vehicles/Time Periods) 450 471 418 436 403 360 342 332 Mean Speed (mph) 25.61 33.09 22.74 34.31 29.74 45.74 47.28 46.59 85th Percentile (mph) 28.17 42.60 26.30 46.04 46.09 50.38 55.20 51.64 SD 6.43 8.38 3.39 10.27 11.44 5.64 8.24 5.82 Mean Speed tstatic -5.71 -7.48 -14.63 0.69 (Bold indicates significance at the 95% confidence level, α = .05) Similar to the mean speeds, the 85th percentile speeds on the mainline of the freeway also increased for time periods prior to 8:15 a.m. The t-test results indicated that increases in mean speed during the fixed-cycle length ramp metering scenario were statistically significant for the time periods from 7:30 to 8:15 a.m. 6.3.1.2 Meter-off Scenario vs. Variable-cycle Length Ramp Metering As Table 25 shows, the mean speeds of vehicles on the mainline of the freeway increased for all time periods from 7:30 to 8:30 a.m. Table 25. Speed comparison, meter-off, and variable-cycle length ramp metering. Meter off (Without Ramp Metering) and Option 2 (Variable-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 2 Meter off Option 2 Meter off Option 2 Meter off Option 2 Volume (Vehicles/Time Periods) 450 454 418 439 403 367 342 308 Mean Speed (mph) 25.61 27.20 22.74 26.37 29.74 44.82 47.28 48.52 85th Percentile (mph) 28.17 33.87 26.30 31.27 46.09 53.89 55.20 56.05 SD 6.43 6.51 3.39 5.51 11.44 9.47 8.24 7.61 Mean Speed tstatic -2.86 -9.20 -16.55 -1.77 (Bold indicates significance at the 95% confidence level, α = .05.)
68 Similar to the mean speeds, the 85th percentile speeds on the mainline of the freeway also increased for all time periods from 7:30 to 8:30 a.m. The t-test results indicated the increases in mean speed during the variable-cycle length ramp metering scenario were also statistically significant for all time periods. After implementing ramp metering, the speeds of vehicles on the mainline of the freeway increased in both scenarios. Therefore, it seems that implementing ramp metering has a positive effect, and although the changes varied, they were found to be statistically significant. What is also evident is that mainline saturation was reached around 6:30 a.m. and resulted in traffic flow breakdown prior to turning on the ramp meter. As expected, it took the ramp metering about 5â10 minutes to stabilize the traffic flow before any major benefit, such as improvements in vehicle speed, could be realized. The ramp meter was turned on after mainline saturation; however, even at high ramp volumes (~900 vph), the system was able to generate approximately 5â9 mph increase in mainline speeds. The fixed-cycle length ramp-metering scenario worked best at such high ramp volumes, with a 28% (8.6 mph) increase in speed compared with variable-cycle length ramp metering scenario at 16.1% (5.18 mph) increase in speed. 6.3.2 I-279, Ohio Township, Pennsylvania The most congested peak period at the study area was the a.m. peak; therefore, the traffic analyses of this study also focused on the a.m. peak hour only. The team conducted an analysis of vehicle speeds on the mainline (both right lane and left lane) of the freeway at the merging area (Location 3) during the a.m. peak period (5:30 to 10:00 a.m.). Figures 22 and 23 illustrate the vehicle speeds and traffic volumes on the mainline for all scenarios. In general, it appears that ramp metering improved performance of the main lines (i.e., vehicle speeds were higher for all meter-on scenarios [Options 1 and 2] during the a.m. peak hour period). Mainline saturation was reached around 5:45 a.m. and resulted in traffic flow breakdown. After that, the ramp metering stabilized the traffic flow and major benefits (such as improved vehicle speed and travel time) were realized. The team observed significant improvements in vehicle speed and volume between 6:45 a.m. and 8:15 a.m. (Figures 24).
69 Figure 22. A.M. peak hour vehicle speed and traffic volumes at Location 3 (right lane). 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 5: 30 :0 0 AM 5: 36 :0 0 AM 5: 42 :0 0 AM 5: 48 :0 0 AM 5: 54 :0 0 AM 6: 00 :0 0 AM 6: 06 :0 0 AM 6: 12 :0 0 AM 6: 18 :0 0 AM 6: 24 :0 0 AM 6: 30 :0 0 AM 6: 36 :0 0 AM 6: 42 :0 0 AM 6: 48 :0 0 AM 6: 54 :0 0 AM 7: 00 :0 0 AM 7: 06 :0 0 AM 7: 12 :0 0 AM 7: 18 :0 0 AM 7: 24 :0 0 AM 7: 30 :0 0 AM 7: 36 :0 0 AM 7: 42 :0 0 AM 7: 48 :0 0 AM 7: 54 :0 0 AM 8: 00 :0 0 AM 8: 06 :0 0 AM 8: 12 :0 0 AM 8: 18 :0 0 AM 8: 24 :0 0 AM 8: 30 :0 0 AM 8: 36 :0 0 AM 8: 42 :0 0 AM 8: 48 :0 0 AM 8: 54 :0 0 AM 9: 00 :0 0 AM 9: 06 :0 0 AM 9: 12 :0 0 AM 9: 18 :0 0 AM 9: 24 :0 0 AM 9: 30 :0 0 AM 9: 36 :0 0 AM 9: 42 :0 0 AM 9: 48 :0 0 AM 9: 54 :0 0 AM Vo lu m e (V eh icl e/ M in ) Sp ee d (M PH ) Axis Title AM Average Speed and Traffic Volumes (at Location 3, 1 Minute Bin) Right Lane Option 1 (Fixed Cycle Length) - Speed Meter Off (Without Ramp Metering) - Speed Option 2 (Variable Cycle Length) - Speed Option 1 (Fixed Cycle Length) - Volume Meter Off (Without Ramp Metering) - Volume Option 2 (Variable Cycle Length) - Volume Poly. (Option 1 (Fixed Cycle Length) - Volume) Poly. (Meter Off (Without Ramp Metering) - Volume) Poly. (Option 2 (Variable Cycle Length) - Volume)
70 Figure 23. A.M. peak hour vehicle speed and traffic volumes at Location 3 (left lane). 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 5: 30 :0 0 AM 5: 36 :0 0 AM 5: 42 :0 0 AM 5: 48 :0 0 AM 5: 54 :0 0 AM 6: 00 :0 0 AM 6: 06 :0 0 AM 6: 12 :0 0 AM 6: 18 :0 0 AM 6: 24 :0 0 AM 6: 30 :0 0 AM 6: 36 :0 0 AM 6: 42 :0 0 AM 6: 48 :0 0 AM 6: 54 :0 0 AM 7: 00 :0 0 AM 7: 06 :0 0 AM 7: 12 :0 0 AM 7: 18 :0 0 AM 7: 24 :0 0 AM 7: 30 :0 0 AM 7: 36 :0 0 AM 7: 42 :0 0 AM 7: 48 :0 0 AM 7: 54 :0 0 AM 8: 00 :0 0 AM 8: 06 :0 0 AM 8: 12 :0 0 AM 8: 18 :0 0 AM 8: 24 :0 0 AM 8: 30 :0 0 AM 8: 36 :0 0 AM 8: 42 :0 0 AM 8: 48 :0 0 AM 8: 54 :0 0 AM 9: 00 :0 0 AM 9: 06 :0 0 AM 9: 12 :0 0 AM 9: 18 :0 0 AM 9: 24 :0 0 AM 9: 30 :0 0 AM 9: 36 :0 0 AM 9: 42 :0 0 AM 9: 48 :0 0 AM 9: 54 :0 0 AM Vo lu m e (V eh icl e/ M in ) Sp ee d (M PH ) Axis Title AM Average Speed and Traffic Volumes (at Location 3, 1 Minute Bin) Left Lane Option 1 (Fixed Cycle Length) - Speed Meter Off (Without Ramp Metering) - Speed Option 2 (Variable Cycle Length) - Speed Option 1 (Fixed Cycle Length) - Volume Meter Off (Without Ramp Metering) - Volume Option 2 (Variable Cycle Length) - Volume Poly. (Option 1 (Fixed Cycle Length) - Volume) Poly. (Meter Off (Without Ramp Metering) - Volume) Poly. (Option 2 (Variable Cycle Length) - Volume)
71 Figure 24. Hourly volumeâLocation 3, after the merge area. The team calculated changes in the mean speeds and 85th percentile speeds for vehicles in the with and without ramp-metering scenarios. Tables 24 and 25 show the comparison of mean speed and 85th percentile speed of vehicles on the mainline (both right lane and left lane) of the freeway and the statistical test results with and without ramp metering. The following sections discuss both meter-on scenarios (fixed time vs. variable time). 6.3.2.1 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering As Tables 26 and 27 show, the mean speeds of vehicles on the mainline of the freeway increased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. 0 500 1000 1500 2000 2500 4:48:00 AM 6:00:00 AM 7:12:00 AM 8:24:00 AM 9:36:00 AM Hourly Volume (Location: After the Merge Area) R - Meter off R - Meter on (Fixed) R - Meter on (Varied) L - Meter off L - Meter on (Fixed) L - Meter on (varied)
72 Table 26. Speed comparison, meter-off scenario and fixed-cycle length ramp metering (right lane). Meter off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter Off Option 1 Meter Off Option 1 Meter Off Option 1 Meter Off Option 1 Volume (vehicles/time period) 200 255 190 240 195 245 252 266 Mean Speed (mph) 17.87 22.69 18.87 22.87 17.87 22.31 24.56 30.91 85th Percentile (mph) 24.80 30.00 26.40 32.00 24.80 30.20 33.00 48.00 SD 7.47 6.63 7.47 8.17 7.47 7.82 9.81 13.32 Mean Speed tstatic -3.24 -2.42 -2.76 -2.58 (Bold indicates significance at the 95% confidence level, α = .05.) Table 27. Speed comparison, meter-off scenario and fixed-cycle length ramp metering (left lane). Meter Off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter Off Option 1 Meter Off Option 1 Meter Off Option 1 Meter Off Option 1 Volume (vehicles/time period) 289 379 299 262 304 353 343 378 Mean Speed (mph) 17.49 23.31 18.42 23.33 18.87 23.67 26.69 33.02 85th Percentile (mph) 25.40 33.40 27.00 31.00 29.00 32.60 39.00 51.00 SD 8.29 7.75 7.93 8.64 8.51 8.90 11.91 14.75 Mean Speed tstatic -3.44 -2.81 -2.62 -2.24 (Bold indicates significance at the 95% confidence level, α = .05.) Similar to the mean speeds, the 85th percentile speeds on the mainline of the freeway also increased for time periods from 7:30 to 8:30 a.m. The t-test results indicated the increases in mean speed during the fixed-cycle length ramp metering scenario were statistically significant for the time periods from 7:30 to 8:30 a.m. Overall, speeds increased in both right and left lanes, under the fixed-cycle length ramp metering scenario by 24% (4.8 mph) and 41 percent (8.34 mph), respectively. A larger increase in the left lane was expected as fewer vehicles try to merge across to the left lane. 6.3.2.2 Meter-off Scenario vs. Variable-cycle Length Ramp Metering As Tables 28 and 29 show, the mean speeds of vehicles on the mainline of the freeway increased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane.
73 Table 28. Speed comparison, meter-off scenario and variable-cycle length ramp metering (right lane). Meter Off (Without Ramp Metering) and Option 2 (Variable-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter Off Option 2 Meter Off Option 2 Meter Off Option 2 Meter Off Option 2 Volume (vehicles/time period) 200 293 190 287 195 255 252 297 Mean Speed (mph) 17.87 35.64 18.87 30.27 17.87 29.14 24.56 30.49 85th Percentile (mph) 24.80 56.40 26.40 45.20 24.80 43.85 33.00 45.80 SD 7.47 19.06 7.47 12.56 7.47 13.44 9.81 12.63 Mean Speed tstatic -5.83 -5.23 -4.79 -2.49 (Bold indicates significance at the 95% confidence level, α = .05.) Table 29. Speed comparison, meter-off scenario and variable-cycle length ramp metering (left lane). Meter Off (Without Ramp Metering) and Option 2 (Variable-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter Off Option 2 Meter Off Option 2 Meter Off Option 2 Meter Off Option 2 Volume (vehicles/time period) 289 389 299 365 304 333 343 378 Mean Speed (mph) 17.49 38.60 18.42 33.56 18.87 32.00 26.69 33.33 85th Percentile (mph) 25.40 61.40 27.00 51.80 29.00 47.85 39.00 51.00 SD 8.29 21.49 7.93 14.26 8.51 15.08 11.91 13.54 Mean Speed tstatic -6.15 -6.22 -4.96 -2.47 (Bold indicates significance at the 95% confidence level, X = .05.) Similar to the mean speeds, the 85th percentile speeds on the mainline of the freeway also increased for all time periods from 7:30 to 8:30 a.m. The t-test results indicated the increases in mean speed during the variable-cycle length ramp metering scenario were also statistically significant for all time periods. After implementing ramp metering, the speeds of vehicles on the mainline of the freeway increased in both scenarios (Figure 25). Therefore, it seems that implementing ramp metering had a positive effect and although the changes varied, the changes were found to be statistically significant.
74 Figure 25. Vehicle speedâLocation 3: After the merge area.
75 Overall, speeds increased in both right and left lanes, under the variable-cycle length ramp metering scenario by 57.2% (11.45 mph) and 69.8% (14.17 mph), respectively. The left lane was expected to see a larger increase. However, it is clear that when the ramp volume exceeds 650â 750 vph and mainline volume reaches 1,600 vph, fixed-cycle length ramp metering scenario appears to perform better. 6.4. Comparison of Travel Time The comparison of travel times through work zones is a good measurement of effectiveness to determine the effect of implementing ramp metering. In this study, the team used locations 1 and 3 (defined in Section 6.1) as the reference points to determine the travel time. Ramp metering controlled the number of vehicles released from the ramp to the mainline of the freeway and it is expected to have positive effects on vehicle merging near the ramp, which should result in shorter travel time on the mainline. 6.4.1 MN Route 52, Rochester, Minnesota The team conducted an analysis of vehicle travel time of on the mainline of the freeway from Location 1 to Location 3 in the morning peak period (6:00 to 8:30 a.m.) to determine the effects of implementing ramp metering on both ramp-metering scenarios. Figure 26 illustrates travel time from Location 1 to Location 3 at the different scenarios. It shows that meter-off scenario had a longer travel time than both Meter-on scenarios during the a.m. peak hour period.
76 Figure 26. A.M. peak hour travel time from Location 1 to Location 3 (distance: 2,800 ft) 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 6: 00 :0 0 AM 6: 03 :0 0 AM 6: 06 :0 0 AM 6: 09 :0 0 AM 6: 12 :0 0 AM 6: 15 :0 0 AM 6: 18 :0 0 AM 6: 21 :0 0 AM 6: 24 :0 0 AM 6: 27 :0 0 AM 6: 30 :0 0 AM 6: 33 :0 0 AM 6: 36 :0 0 AM 6: 39 :0 0 AM 6: 42 :0 0 AM 6: 45 :0 0 AM 6: 48 :0 0 AM 6: 51 :0 0 AM 6: 54 :0 0 AM 6: 57 :0 0 AM 7: 00 :0 0 AM 7: 03 :0 0 AM 7: 06 :0 0 AM 7: 09 :0 0 AM 7: 12 :0 0 AM 7: 15 :0 0 AM 7: 18 :0 0 AM 7: 21 :0 0 AM 7: 24 :0 0 AM 7: 27 :0 0 AM 7: 30 :0 0 AM 7: 33 :0 0 AM 7: 36 :0 0 AM 7: 39 :0 0 AM 7: 42 :0 0 AM 7: 45 :0 0 AM 7: 48 :0 0 AM 7: 51 :0 0 AM 7: 54 :0 0 AM 7: 57 :0 0 AM 8: 00 :0 0 AM 8: 03 :0 0 AM 8: 06 :0 0 AM 8: 09 :0 0 AM 8: 12 :0 0 AM 8: 15 :0 0 AM 8: 18 :0 0 AM 8: 21 :0 0 AM 8: 24 :0 0 AM 8: 27 :0 0 AM Tr av el T im e (S ec on d) Time AM Travel Time from Location 1 to Location 3 Option 1 (Fixed Cycle Length) Meter Off (Without Ramp Metering) Option 2 (Variable Cycle Length) D2 D1 D1: Variation Between Meter Off and Option 1 D2: Variation Between Meter Off and Option 2
77 The team used a comparison of travel times to evaluate the effect of ramp metering and determine if ramp metering caused changes in travel characteristics. The team used the t- statistic to evaluate the effect of different ramp-metering scenarios. The team calculated changes in the average travel time and the 85th percentile travel time for vehicles traveling from Location 1 to Location 3 between the with and without ramp metering scenarios. Tables 28 and 29 show the comparison of average travel time, the 85th percentile travel time, and the statistical test results with and without ramp metering.
78 6.4.1.1 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering As Table 30 shows, the average travel time from Location 1 to Location 3 decreased for all time periods before 8:15 a.m. Travel time savings per vehicle ranged from 28% in the early stages of the ramp metering being turned on to over 60% when fully regulating traffic flows. Table 30. Travel time comparison, meter-off scenario and fixed-cycle length ramp metering. Meter off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 1 Meter off Option 1 Meter off Option 1 Meter off Option 1 Volume (vehicles/time period) 450 471 418 436 403 360 342 332 Average Travel Time/vehicle (seconds) 95.60 68.62 125.53 50.15 91.73 35.05 34.57 34.60 85th Percentile Travel Time (seconds) 129.29 97.11 133.30 63.36 111.98 36.06 35.66 35.23 SD 34.98 23.74 8.88 15.24 28.63 1.55 1.73 1.51 Mean Speed tstatic 2.47 16.55 7.66 -0.06 Similarly, the 85th percentile travel time from Location 1 to Location 3 also decreased for all time periods from 7:30 to 8:30 a.m. Overall, the average travel time per vehicle was reduced from 88.2 seconds to 49.0 seconds per vehicle (44.7% decrease). The t-test results indicated statistically significant decreases in travel time during fixed-cycle length ramp-metering scenario. 6.4.1.2 Meter-off Scenario vs. Variable-cycle Length Ramp Metering As Table 31 shows, the average travel time from location 1 to location 3 decreased for all time periods prior to 8:15 a.m. Travel time savings per vehicle were almost identical to Meter On (fixed time) and ranged from 33 percent in the early stages of the ramp metering being turned on to over 60 percent when fully regulating traffic flows. Similar to the mean speeds, the 85th percentile travel time from Location 1 to Location 3 also decreased for all time periods prior to 8:15 a.m. The t-test results indicated statistically significant decreases in travel time during variable-cycle length ramp metering scenario for the time periods from 7:30 to 8:15 a.m. After implementing ramp metering, the travel time from Location 1 to Location 3 decreased in both ramp-metering scenarios. Overall, the average travel time per vehicle was reduced from 88.2 seconds to 52.3 seconds per vehicle (41% decrease).The travel time savings were similar and statistically significant. It can be reasonably concluded that the improved travel time (42%
79 reduction) is the result of implementing ramp metering. Therefore, there seems to be a positive effect on travel time associated with implementing ramp metering. Table 31. Travel time comparison, meter-off scenario and variable-cycle length ramp metering. Meter off (Without Ramp Metering) and Option 2 (Varied-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 2 Meter off Option 2 Meter off Option 2 Meter off Option 2 Volume (vehicles/time period) 450 454 418 439 403 367 342 308 Average Travel Time/vehicle (seconds) 95.60 64.03 125.53 66.50 91.73 35.62 34.57 34.66 85th Percentile Travel Time (seconds) 129.29 71.90 133.30 81.16 111.98 36.57 35.66 36.11 SD 34.98 9.22 8.88 11.95 28.63 1.05 1.73 1.36 Mean Speed tstatic 3.38 15.36 7.59 -0.17 6.4.2 I-279, Ohio Township, Pennsylvania The team also conducted an analysis of travel time of vehicles on the mainline (both right and left lanes) of the freeway from Location 1 to Location 3 in the morning peak period (6:00 to 9:00 a.m.) to determine the effects of implementing ramp metering on both ramp-metering scenarios. Figures 27 and 28 illustrate travel time from Location 1 to Location 3 at the different scenarios. It shows that meter-off scenario had a longer travel time than both Meter-on scenarios during the a.m. peak hour period for both right and left lanes.
80 Figure 27. A.M. peak hour travel time from Location 1 to Location 3, right lane (distance: 5,280 ft). 0.0 100.0 200.0 300.0 400.0 500.0 600.0 6: 00 :0 0 AM 6: 15 :0 0 AM 6: 30 :0 0 AM 6: 45 :0 0 AM 7: 00 :0 0 AM 7: 15 :0 0 AM 7: 30 :0 0 AM 7: 45 :0 0 AM 8: 00 :0 0 AM 8: 15 :0 0 AM 8: 30 :0 0 AM 8: 45 :0 0 AM Tr av el T im e (S ec on d) Time AM Travel Time from Location 1 to Location 3 (Right Lane) Option 1 (Fixed Cycle Length) Meter Off (Without Ramp Metering) Option 2 (Varied Cycle Length) D1 D2 D1: Variation Between Meter Off and Option 1 D2: Variation Between Meter Off and Option 2
81 Figure 28. A.M. peak hour travel time from Location 1 to Location 3, left lane (distance: 5,280 ft). 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 400.0 450.0 6: 00 :0 0 AM 6: 15 :0 0 AM 6: 30 :0 0 AM 6: 45 :0 0 AM 7: 00 :0 0 AM 7: 15 :0 0 AM 7: 30 :0 0 AM 7: 45 :0 0 AM 8: 00 :0 0 AM 8: 15 :0 0 AM 8: 30 :0 0 AM 8: 45 :0 0 AM Tr av el T im e (S ec on d) Time AM Travel Time from Location 1 to Location 3 (Left Lane) Option 1 (Fixed Cycle Length) Meter Off (Without Ramp Metering) Option 2 (Varied Cycle Length) D1 D2 D1: Variation Between Meter Off and Option 1 D2: Variation Between Meter Off and Option 2
82 The team used a comparison of travel times to evaluate the effect of ramp metering and determine if ramp metering caused changes in travel characteristics. The t-statistic was used to evaluate the effect of different ramp metering scenarios. The team calculated changes in average travel time and the 85th percentile travel time for vehicles traveling from Location 1 to Location 3 between the with and without ramp metering scenarios for both right lane and left lane. Tables 32 and 33 show the comparison of average travel time and the 85th percentile travel time along with the statistical test results with and without ramp metering. 6.4.2.1 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering As Tables 32 and 33 show, the average travel time from Location 1 to Location 3 decreased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. Travel time savings per vehicle ranged from 16% in the early stages of the ramp metering being turned on to over 48% when fully regulating traffic flows for the right lane. For the left lane, travel time savings per vehicle ranged from 4% in the early stages of the ramp metering being turned on to more than 35% when fully regulating traffic flows for the right lane. Table 32. Travel time comparison, meter-off scenario and fixed-cycle length ramp metering (right lane). Meter off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 1 Meter off Option 1 Meter off Option 1 Meter off Option 1 Volume (vehicles/time period) 200 255 190 240 195 245 252 266 Average Travel Time/vehicle (seconds) 383.03 321.11 334.17 273.92 311.25 161.79 156.96 113.99 85th Percentile Travel Time (seconds) 444.26 369.49 410.42 307.38 344.38 199.33 209.13 122.67 SD 60.19 83.16 57.31 50.97 65.22 50.09 67.54 9.78 Mean Speed tstatic 2.34 3.04 7.04 2.44 (Bold indicates significance at the 95% confidence level, α = .05.)
83 Table 33. Travel time comparison, meter-off scenario and fixed-cycle length ramp metering (left lane). Meter off (Without Ramp Metering) and Option 1 (Fixed-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 1 Meter off Option 1 Meter off Option 1 Meter off Option 1 Volume (vehicles/time period) 289 379 299 262 304 353 343 378 Average Travel Time/vehicle (seconds) 268.28 256.77 249.59 198.74 215.04 139.07 131.51 100.75 85th Percentile Travel Time (seconds) 307.93 295.89 266.31 223.00 260.47 168.74 165.44 109.19 SD 36.49 59.71 58.62 27.82 37.17 35.83 43.62 7.10 Mean Speed tstatic 0.64 3.04 5.70 2.70 (Bold indicates significance at the 95% confidence level, α = .05.) Similarly, the 85th percentile travel time from Location 1 to Location 3 also decreased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. Overall, travel time per vehicle was reduced from 287 seconds to 214.2 seconds per vehicle (25.4% decrease) and 209.6 seconds to 191.7 seconds (8.6% decrease) in the right lane and left lane, respectively. The t-test results indicated the decreases in travel time during fixed-cycle length ramp-metering scenario were statistically significant for all time periods from 7:30 to 8:30 a.m. for the right lane. The t- test results also indicated the decreases in travel time during fixed-cycle length ramp-metering scenario were statistically significant for time periods from 7:45 to 8:30 a.m. for the left lane. 6.4.2.2 Meter-off Scenario vs. Variable-cycle Length Ramp Metering As Tables 34 and 35 show, the average travel time from Location 1 to Location 3 decreased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. Travel time savings per vehicle ranged from 40% to 75% for the right lane. For the left lane, travel time savings per vehicle ranged from 31% to 65%.
84 Table 34. Travel time comparison, meter-off scenario and variable-cycle length ramp metering (right lane). Meter off (Without Ramp Metering) and Option 2 (Variable-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 2 Meter off Option 2 Meter off Option 2 Meter off Option 2 Volume (vehicles/time period) 200 293 190 287 195 255 252 297 Average Travel Time/vehicle (seconds) 383.03 95.32 334.17 100.54 311.25 93.66 156.96 94.16 85th Percentile Travel Time (seconds) 444.26 107.45 410.42 112.63 344.38 99.23 209.13 106.17 SD 60.19 13.80 57.31 9.78 65.22 5.54 67.54 9.52 Mean Speed tstatic 18.05 15.56 12.87 3.57 (Bold indicates significance at the 95% confidence level, α = .05.) Table 35. Travel time comparison, meter-off scenario and variable-cycle length ramp metering (left lane). Meter off (Without Ramp Metering) and Option 2 (Variable-cycle Length) 07:30 to 07:45 07:45 to 08:00 08:00 to 08:15 08:15 to 08:30 Meter off Option 2 Meter off Option 2 Meter off Option 2 Meter off Option 2 Volume (vehicles/time period) 289 389 299 365 304 333 343 378 Average Travel Time/vehicle (seconds) 268.28 92.42 249.59 93.18 215.04 90.52 131.51 90.30 85th Percentile Travel Time (seconds) 307.93 113.75 266.31 101.58 260.47 96.24 165.44 96.75 SD 36.49 14.05 58.62 10.10 37.17 8.42 43.62 7.36 Mean Speed tstatic 17.42 10.18 12.66 3.61 (Bold indicates significance at the 95% confidence level, α = .05.) Similar to the mean speeds, the 85th percentile travel time from Location 1 to Location 3 also decreased for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. Overall, travel time per vehicle was reduced from 287 seconds to 96 seconds (66.5% decrease) and 209.7 seconds to 90.9 seconds (56.6% decrease) in the right lane and left lane, respectively. The t-test results indicated the decreases in travel time during variable-cycle length ramp-metering scenario were statistically significant for all time periods from 7:30 to 8:30 a.m. for both right lane and left lane. After implementing ramp metering, the travel time from Location 1 to Location 3 decreased in both ramp metering scenarios. The travel time savings were similar and statistically significant.
85 It can be reasonably concluded that the improved travel time (20% for fixed and 60% for variable) is the result of implementing ramp metering. Therefore, there seems to be a positive effect on travel time associated with implementing ramp metering. Variable-cycle length ramp- metering scenario seems to have a greater benefit when the network is not operating at saturation. 6.5. Comparison of Results for Frequency of Headway Headway is a good measure of congestion and lack of passing opportunities created by the traffic mix; it is also a good measure of safety as lane changing and frequent passing generally lead to conflicts and the likelihood of crashes. In general, a longer headway accepted by a merging vehicle is safer than a shorter headway. The team examined the headways accepted by following vehicles to see if there were any differences between the with and without implementation of ramp metering. 6.5.1 MN Route 52, Rochester, Minnesota The team conducted an analysis of headways of vehicles on the mainline of the freeway at Location 3 in the morning peak hour (7:30 to 8:30 a.m.) to determine the average values and distribution for both ramp-metering scenarios. As mentioned earlier, the team used the K-S test to judge how faithfully a distribution fits the sample data. The K-S test was adopted to determine the goodness-of-fit in the work zone traffic condition. Table 36. K-S test results for the meter-on scenarios. Scenarios Meter Off Meter On (Fixed-cycle Length) Meter-On (Variable-cycle Length) Volume (Vehicle/Hour) 1614 1600 1568 Mean Headway (seconds) 2.29 2.42 2.44 Median Headway (seconds) 2.17 2.25 2.23 Maximum Difference (D) 0.167 0.183 Significant Difference No No Table 36 summarizes the K-S test results of the meter-on scenarios. The following section discusses both Meter-on scenarios.
86 6.5.1.2 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering Figure 29 presents a visual performance comparison of headway distributions through a cumulative distribution function in the morning peak period at Location 3 for the meter-Off scenario and the fixed-cycle length ramp metering. The team observed a slight shift in the headway distribution toward longer headways resulting from ramp metering. The Meter-on scenario had a longer headway in approximately 50% of samples (cumulative percentage 45% to 95%) with a maximum headway difference of 0.4 seconds. The mean value of headway was 2.42 seconds with ramp metering as opposed to 2.29 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of K-S test for the maximum difference between the cumulative distributions, D, was 0.25. The results of the K-S test for (meter-off scenario vs. fixed-cycle length ramp metering) shows a value of D of 0.167 (less than the critical value of 0.25), which suggests the differences in the two cumulative distributions are not statistically significant. Figure 29. Cumulative headway distribution plot, meter-off scenario vs. fixed-cycle length ramp metering (7:30 to 8:30 a.m.). 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Cu m ul at iv e pe rc en ta ge Headway (Second) Meter Off Meter on-Fixed Cycle Length
87 6.5.1.3 Meter-off Scenario vs. Variable-cycle Length Ramp Metering Figure 30 presents a visual performance comparison of headway distributions through a cumulative distribution function in the morning peak period at Location 3 for the meter-off scenario and variable-cycle length ramp metering. The team observed a slight shift in the headway distribution toward longer headways as a result of ramp metering. The meter-on scenario has a longer headway in more than 90% of the samples (cumulative percentage 5% to 97%) with a maximum headway difference of 0.4 sec. The mean value of headway was 2.44 seconds with ramp metering as opposed to 2.29 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of K-S test for the maximum difference between the cumulative distributions, D, was 0.25. The results of the K-S test for (meter-off scenario vs. variable-cycle length ramp metering) shows a value of D of 0.183 (less than the critical value of 0.25), which suggests the differences in the two cumulative distributions are not statistically significant. Figure 30. Cumulative headway distribution plot, meter-off scenario vs. variable-cycle length ramp metering (7:30 to 8:30 a.m.). 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Cu m ul at iv e pe rc en ta ge Headway (Second) Meter Off Meter on-Variable Cycle Length
88 The headways of vehicles on the mainline of the freeway at Location 3 showed an increase in both ramp-metering scenarios from meter-off scenario (without ramp metering). The result of the K-S test indicated that the differences in the two cumulative distributions were not statistically significant. It can be reasonably concluded that although the headway increased slightly, a positive effect, this was not as a result of implementing ramp metering. The converse is that by implementing ramp metering, the headway remains unchanged, and assuming that headway is a safety surrogate, then safety remained unchanged. 6.5.2 I-279, Ohio Township, Pennsylvania The team conducted an analysis of headways of vehicles on the mainline (both right lane and left lane) of the freeway at Location 3 in the morning peak hour (7:30 to 8:30 a.m.) to determine the average values and distribution for both ramp metering scenarios. As noted, above, the team used the K-S test to judge how faithfully a distribution fits the sample data. The K-S test was adopted to determine the goodness-of-fit in the work zone traffic condition. 6.5.2.1 Right Lane Table 37 summarizes the K-S test results of the meter-on scenarios for the right lane. The following section discusses both meter-on scenarios. Table 37. K-S test results for the meter-on scenarios (right lane). Right Lane Scenarios Meter Off Meter On (Fixed-cycle Length) Meter On (Varied-cycle Length) Volume (Vehicle/Hour) 837 1005 1,133 Mean Headway (seconds) 4.01 3.74 3.22 Median Headway (seconds) 3.75 3.53 2.86 Maximum Difference (D) 0.10 0.28 Significance No Yes
89 Meter-off Scenario vs. Fixed-cycle Length Ramp Metering (Right Lane) Figure 31 presents a comparison of cumulative headway distributions in the morning peak period at Location 3 for meter-off scenario and fixed-cycle length ramp metering. The team observed a shift in the headway distribution toward shorter headway resulting from ramp metering. The mean value of headway was 3.74 seconds with ramp metering as opposed to 4.01 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of K-S test for the maximum difference between the cumulative distributions, D, is 0.16. The results of the K-S test (meter-off scenario vs. fixed-cycle length ramp metering) shows a value of D of 0.10 (less than the critical value of 0.16), which suggests the differences in the two cumulative distributions are not statistically significant. Figure 31. Cumulative headway distribution plot, meter-off scenario vs. fixed-cycle length ramp metering, right lane (7:30 to 8:30 a.m.). 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 1 2 3 4 5 6 7 8 9 Cu m ul at iv e Pe rc en ta ge Headway (Second) Right Lane Meter Off 7:30-8:30 AM Right Lane Meter On (Fixed Time) 7:30-8:30 AM
90 Meter-off Scenario vs. Variable-cycle Length Ramp Metering (Right Lane) Figure 32 presents a visual performance comparison of headway distributions through a cumulative distribution function in the morning peak period at Location 3 for meter-off scenario and variable-cycle length ramp metering. The team observed a shift in the headway distribution toward shorter headway as a result of ramp metering. The mean value of headway was 3.22 seconds with ramp metering as opposed to 4.01 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of the K-S test for the maximum difference between the cumulative distributions, D, is 0.16. The results of the K-S test (meter-off scenario vs. variable-cycle length ramp metering) shows a value of D of 0.28 (greater than the critical value of 0.16), which suggests the differences in the two cumulative distributions are statistically significant. Figure 32. Cumulative headway distribution plot, meter-off scenario vs. variable-cycle length ramp metering, right lane (7:30â8:30 a.m.). 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 1 2 3 4 5 6 7 8 9 Cu m ul at iv e Pe rc en ta ge Headway (Second) Right Lane Meter Off 7:30-8:30 AM Right Lane Meter On (Varied Time) 7:30-8:30 AM
91 6.5.2.2 Left Lane Table 38 summarizes the K-S test results of the meter-on scenarios for the left lane. The following discusses both Meter-on scenarios. Table 38. K-S test results for the meter-on scenarios (left lane). Left Lane Scenarios Meter Off Meter On (Fixed-cycle Length) Meter On (Varied-cycle Length) Volume (Vehicle/Hour) 1235 1472 1465 Mean Headway (seconds) 3.24 2.66 2.62 Median Headway (seconds) 2.86 2.40 2.40 Maximum Difference (D) 0.22 0.24 Significance Yes Yes Meter-off Scenario vs. Fixed-cycle Length Ramp Metering (Left Lane) Figure 33 presents a visual performance comparison of headway distributions through a cumulative distribution function in the morning peak period at Location 3 for the meter-off scenario and fixed-cycle length ramp metering. The team observed a shift in the headway distribution toward shorter headway resulting from ramp metering. The mean value of headway was 2.66 seconds with ramp metering as opposed to 3.24 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of the K-S test for the maximum difference between the cumulative distributions, D, was 0.15. The results of K-S test (meter-off scenario vs. fixed-cycle length ramp metering) shows a value of D of 0.22 (greater than the critical value of 0.15), which suggests the differences in the two cumulative distributions are statistically significant.
92 Figure 33. Cumulative headway distribution plot, meter-off scenario vs. fixed-cycle length ramp metering, left lane (7:30 to 8:30 a.m.). 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 1 2 3 4 5 6 7 8 9 Cu m ul at iv e Pe rc en ta ge Headway (Second) Left Lane Meter Off 7:30-8:30 AM Left Lane Meter On (Fixed Time) 7:30-8:30 AM
93 Meter-off Scenario vs. Variable-cycle Length Ramp Metering (Left Lane) Figure 34 presents a visual performance comparison of headway distributions through a cumulative distribution function in the morning peak period at location 3 for the meter-off scenario and variable-cycle length ramp metering. The team observed a shift in the headway distribution toward shorter headway as a result of ramp metering. The mean value of headway was 2.62 seconds with ramp metering as opposed to 3.24 seconds without ramp metering. With the significance level α of 0.05, the critical statistic of the K-S test for the maximum difference between the cumulative distributions, D, was 0.15. The results of the K-S test for (meter-off scenario vs. variable-cycle length ramp metering) shows a value of D of 0.24 (greater than the critical value of 0.15), which suggests the differences in the two cumulative distributions are statistically significant. Figure 34. Cumulative headway distribution plot, meter-off scenario vs. variable-cycle length ramp metering, left lane (7:30â8:30 a.m.). The headways of vehicles on the mainline of the freeway at Location 3 exhibited a decrease in both ramp metering scenarios from the meter-off scenario (without ramp metering). The result of the K-S test indicated that the differences in the two cumulative distributions were 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 1 2 3 4 5 6 7 8 9 Cu m ul at iv e Pe rc en ta ge Headway (Second) Left Lane Meter Off 7:30-8:30 AM Left Lane Meter On (Varied Time) 7:30-8:30 AM
94 statistically significant for three scenario comparisons (right lane, meter-off vs. variable-cycle length; left lane, meter-off vs. fixed-cycle length; left lane, meter-off vs. variable-cycle length) and not statistically significant for one scenario-comparison (right lane, meter-off vs. fixed- cycle). It can be reasonably concluded that the decrease in headways, a negative effect, was the result of implementing ramp metering. However, this can be expected as the mainline traffic volumes increased as ramp platoons were controlled.
95 6.6. Network Summary Figure 35 provides a visual volume and speed graphic of the effect of ramp metering under meter off, fixed-cycle length, and variable-cycle length scenarios taken at the gore (mainline and ramp). It is clearly visible that traffic volumes increase in both left and right lanes as demonstrated by a greater concentration of lighter colors for both fixed-cycle and variable-cycle length scenarios. Similarly, the same observation can be demonstrated for vehicle speeds. *Nâmeter-off scenario, Fâfixed-cycle length scenario, Vâvariable-cycle length ramp metering Figure 35. Volume vs Speed charts (left/right lane) at gore.
96 Figure 36 illustrates the positive effect along the corridor before/after metering by operations. While both fixed-cycle length and variable-cycle length scenarios yielded positive results, the variable-cycle length scenario showed improved results with respect to vehicle speeds. All data collection segments (1 through 4) are shown in Figure 36 with location 1 and 4 being furthest upstream and downstream, respectively. A visual comparison of each segment under respective modes - meter off, fixed-cycle length, and variable-cycle length scenarios reveal less lower speed (greater concentration of lighter colors), respectively. Figure 36. Network effect.
97 6.7. Driver Compliance Rates The primary interest in evaluating the effectiveness of ramp metering is the driversâ compliance rate. The compliance rate will determine whether ramp metering works well for the study area. 6.7.1 MN Route 52, Rochester, Minnesota The design of the freeway entrance ramp allows for the use of three-section heads (MUTCD 2009, Section 4I). Based on the observed ramp flows, the team developed two signalization schemes (1) Fixed-cycle lengthâgreen, yellow, and red times are consistent for each cycle. (2) Variable-cycle lengthâgreen, yellow, and red times vary based on traffic flow on the ramp. The duration for green, yellow, and red times were set for 1.0, 0.5, and 3.5 seconds, respectively, for fixed-cycle length ramp metering. For variable-cycle length ramp metering, the green time varied from 1 to 1.5 seconds, the yellow time varied from 0.5 to 1.5 seconds, and the red time varied from 3.5 to 4.5 seconds. The team reviewed 1 hour of video on-ramp for both ramp metering scenarios during the a.m. peak hour and with no enforcement present. The following sample sizes and compliance rates were obtained for the ramp metering: ⢠Fixed-cycle length ramp meteringâsample size: 445, 63.1% compliance. ⢠Variable-cycle length ramp meteringâsample size: 376, 76.3% compliance. 6.7.2 I-279, Ohio Township, Pennsylvania The team developed two signalization schemes (1) Fixed-cycle lengthâgreen and red times were consistent for each cycle. (2) Variable-cycle lengthâgreen and red times varied based on traffic flow on the ramp. The duration for green and red times were set for 0.5 and 3.5 seconds, respectively, for Fixed-cycle length ramp metering. For Variable-cycle length ramp metering, the red time varied from 3.5 to 5 seconds. The team reviewed 1 hour of video on-ramp for both ramp-metering scenarios during the a.m. peak hour, with no enforcement present. The following sample sizes and compliance rates were obtained for the ramp metering: ⢠Fixed-cycle length ramp meteringâsample size: 283, 92.2 percent compliance. ⢠Variable-cycle length ramp meteringâsample size: 247, 93.5 percent compliance. In general, driver compliance was greater than expected given that no enforcement was present prior to or during the study period. It appears that the two-signal head commands greater compliance than a three-signal head.
98 6.8. Work Zone Crash Modification Factor for Ramp Metering This section discusses the CMF calculation for ramp metering in work zones. Table 39 shows the expected and actual crash results for the deployment of the ramp meter. The Total Hours column indicates the number of hours of data analyzed. Table 39. Expected and actual crash results for ramp metering. Treatment Total Hours Expected Crashes Actual Crashes Percent Change Ramp Meter Deployment 5,880 8.2 7 -15 For the ramp meter condition, there was a 15 percent decrease from expected to actual crashes. In order to determine the proportional effects of the treatments on the numbers of crashes, an odds ratio analysis was undertaken according to the following equations: Where: CMFD = crash modification factor = proportional effect of a deployment on crashes: TAD = total actual crashes during a deployment (equal to 7 in this case); TED = total expected crashes during a deployment (equal to 8.2 in this case); TAND = total actual crashes when nothing was deployed (equal to 5 in this case); TEND = total expected crashes when nothing was deployed (equal to 7.2 in this case); and SD (CMFD) = standard error. Table 40 shows the results from the CMF calculation. The calculated CMF for the deployment of ramp meter is less than 1, indicating that this treatment had some effect on reducing the number of crashes, without taking standard error into account.
99 Table 40. CMF results for ramp meter. Treatment CMFD SE(CMFD) ADT Ramp Meter Deployment 0.847 0.544 Up to 100,000 Vehicles The CMF calculation was limited because of the few test sites. Agencies should use this only as a guide, monitor all work zones, and take appropriate action to mitigate any increase in crashes (i.e., severity and number).
