Click for next page ( 190


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 189
APPENDIX H PILOT FIELD STUDY RESULTS INTRODUCTION Vehicle-actuated traffic signals have been used to control traffic at signalized intersections for the last six decades. The operational performance of these vehicle-actuated traffic signals is dependent on the traffic patterns at the intersection and the actuated controller settings. A well-designed vehicle- actuated control plan can reduce delays, emissions, and fuel consumption significantly, and a poorly- designed vehicle-actuated control plan can have the onnosite effect. Because average stormed delay .. . . .. . , ~.. 1 A ~ ~ ~ ~ per vehicle Is the measure or electiveness used to determine the level of service of signalized intersections all, one must be able to either measure or estimate delay to determine the level of service at a vehicle-actuated intersection. Currently, traffic simulation packages or the Highway Capaci~Manual (lICMp delay mode} [~] are used to estimate delay at vehicle-actuated intersections, even though one of the well-documented limitations of the HCM delay mode} (and some simulation packages) is its inadequate treatment of vehicle-actuated traffic signals. The HCM delay mode! was developed pnmar;ly for pretimed signals, and the delay at vehicle-actuated intersections is assumed to be 85 percent of the delay at pretimed intersections. One would expect properly designed vehicle-actuated signals to perform significantly better than pretimed signals at low and moderate volumes, but perhaps the same or only slightly better performance at high volumes. Properly designed short unit extensions are also expected to perform better than longer unit extensions due to snappy operation, and different unit extension and maximum green settings should result in different estimates of delay [21. The cycle length and green times presented in Appendix II of Chapter 9 of the HCM are based on the assumption that a vehicle-actuated controller wait maintain 95 percent saturation levels on the critical approach for the respective phases, an assumption that is not always correct. In addition, the HCM mode} does not take into consideration the traffic volumes or the different vehicle-actuated controller settings that are used (e.g., initial interval, vehicle extensions, minimum greens, and maximum greens). Thus, a delay mode} that is sensitive to both traffic volumes and vehicle-actuated controller settings is needed. Also of significance are procedures for determining average phase times and cycle lengths for vehicle-actuated traffic signals, since all analytical delay models rely on these parameter values as inputs; however, models for estimating vehicle-actuated traffic signal timings are not the focus of this paper. Proposed Delav Models Two alternative vehicle-actuated delay models (Fambro-Rouphai! and Akcelik-Chung) have recently been developed to address the need to more accurately estimate delay at vehicle-actuated traffic Appendix H: Page 1

OCR for page 189
signals. Both ofthe proposed models are modifications of the basic HCM delay model; i.e., the same basic models, but with additional parameters for vehicle-actuated control. As with the current HCM model, the estimated delay per vehicle is the sum of a uniform delay component (D1) and an incremental (random plus overflow) delay component (Day. The uniform delay component for both models use the HCM uniform delay term with slight adjustments for average signal timings and degrees of saturation. The overflow delay component for the Fambro-Roupha~! mode! uses a computational mode} developed as part of FIESTA Study 92-C- 0071, and the overflow delay component for the Akcelik-Chung mode! uses a computational mode! developed as part of NC~P 3-48. The overflow delay components in both models use smaller values of"k" for different vehicle-actuated controller settings and detector placements. Both delay models have companion models for estimating average green time, cycle length, and degree of saturation as a function of the controller settings, phasing patterns, traffic volumes, and geometric conditions tad; however, as previously mentioned, those models are not the focus ofthis paper. Their development and validation are documented in other reports. The two vehicle-actuated delay models are shown on the following pages. The Fambro-Rouphai} delay mode! is as follows: D = D,+D2 ( 1 (gaV/Cav)) Do = al01* Scan ~_;g /Cav~m,n(X,1.0) D2 = a2 p2 *90O TX n where: D D1 D2 al ~1 Cav gav X a2 Appendix H: Page 2 (1) (2) (X-1) + ,~ average stopped delay per vehicle, sec/veh; uniform stopped delay per vehicle, sec/veh; incremental stopped delay per vehicle, sec/veh; he , , ~. ~ (X- 1 )2 + 1 FIX ~ e 1 ~ (3) adjustment factor tor converting approach delay to stopped delay in the first delay term; adjustment factor for progression effects in the first delay term; average cycle length, see; average effective green time, see; degree of saturation for subject lane group; adjustment factor for converting approach delay to stopped delay in the second delay term;

OCR for page 189
p2 = adjustment factor for progression elects in the second delay term; T = time period in hours, in which the model parameters are fixed; n = exponential factor for various types of intersections; k = delay parameter for given amval and service distributions; = variance of mean ratio of arrivals/cycle at a point; and c = capacity of the lane group, in veh/hr. For isolated intersections, the following defaults are used: al a2 ~1 ~2 T n k 0.76 for converting total to stopped delay; 0.76 for converting total to stopped delay; 1.0 for isolated intersections; 1.0 for isolated intersections; 0.25 hours; O; 0.5; and 1.0. Substituting these values in the Fambro-Rouphai! model, the equation can be simplified to the follow- ~ng form. This mode! (similar to the HCM delay model, but using k values between 0.04 and 0.20, depending on the unit extension values rather than a constant k value of 0.5) is used for estimating stopped delay at pretimed intersections. D = 0.3 C gav/Ca + 173 X_ + _ 2+ 32kX 1-(g,v/Cav~mln(X 1 0) ~ ( 1) ~ (X 1)~-~ ~ (4) The Ak~elik-Chung delay mode! is as follows: D = Do + D2 (5) D f *0.5r (1-u) for xxO (7) Appendix H: Page 3

OCR for page 189
D2 = o otherwise fd1 1 + 0.40 (sg)-0.3sy 0. kd = 0 40 (sg)07syl l (8) (9) x0 = 042 eh01G02 subject xo<0.95 (10) where: D D1 D2 C g = x qa s u y fdl fdl(x=l) c Tp k = d eh = Gem = average delay per vehicle, sec/veh; non-overflow delay; overflow delay; average cycle length, see; average effective green time, see; average effective red time, r = C -g, see; degree of saturation for subject lane group, x = (qaC)/(sg); arrival flow rate, veh/sec; saturation flow rate, veh/sec; green ratio, u = g/C; flow ratio, y = q/s; non-overflow term parameter; non-overflow term parameter at Id; capacity of the lane group, in vph; peak flow time period in hours, in which the mode} parameters are fixed overflow term parameter; gap setting as a headway value, see; and maximum green setting as a controller displayed value, sec. The differences in the two models are the different k variables, and the far in the uniform delay term and xO in the overflow delay term in the Ak,celik-Chung model, but not in the Fambro-Rouphai! model. Basically, the Fambro-Rouphai! mode} makes adjustments for different unit extension values (k) whereas the Ak~elik-Chung mode} adjusts for different unit extension and maximum green values (xO ). Also, the Fambro-Rouphai} mode} makes no adjustment to uniform delay, but does predict some overflow delay at low degrees of saturation. The Ak~elik-Chung mode! adjusts (slightly increases) uniform delay, but does not predict overflow delays at low degrees of saturation. Appendix H: Page 4

OCR for page 189
Obiectives The objective ofthe pilot study was to collect signalized intersection delay mode} input (geometry, signalization, and traffic data) and output (stopped delay) data at real-worId intersections, and to use this data to evaluate the two proposed vehicle-actuated delay models. A secondary objective of the pilot study was to identify and resolve problems with data collection procedures before sizeable resources for data collection were expended. It also should be noted that the pilot study served as training for the data collectors. Organization This appendix contains three sections. The first introduces the background and objectives of the pilot study. This introductory section also presents the proposed delay models and identifies all the var~- ables within the models to be measured in the field. The second discusses the pilot study site selection, describes the pilot study sites, discusses the techniques that were used in collecting data, and describes the procedures to be used in the reduction and analysis of field data for mode} validation. The third presents the results of the pilot study and recommendations for future field studies. STUDY PROCEDI~E The evaluation of alternative vehicle-actuated delay models requires that delay mode! inputs and outputs be collected under real-worId conditions and then the mode! outputs (measured delay) be compared to the mode! estimates ofthe outputs (estimated delay). Specifically, the study procedure involved collection of data that are used as inputs to the two delay models and the measurement of stopped delays during the same time penod. The input data collected included intersection geometry, turning movement counts on a cycle-by-cycle basis (aggregated to 15-minute intervals), actuated controller settings (minimum green, maximum green, and unit extensions), and signal timing data (green times and cycle lengths) on a cycle-by-cycle basis (averaged over ~ 5-m~nute intervals). The delay mode} inputs and outputs were collected by manual counts, video recordings, and elec- tron~c downioad~ng of controller data. The video records also provided redundancy for spot checking the accuracy of the manual and electronic data when needed. The mode! inputs from the field data were plugged into the proposed delay models to provide estimated delays for the conditions observed in the field. These estimated delays were then compared to measured delays from the field. The following sections describe the site selection and data collection procedures in more detail. Site Selection The calibration and validation ofthe proposed delay models requires the collection of field data from a diverse set of conditions to ensure that the mode! is applicable to the range of operating conditions Appendix H: Page 5

OCR for page 189
found In the field. Thus, the criteria used in identifying possible data collection locations should be based on the ability of the selected sites to provide a wide range of operating conditions. These con- ditions include geographic vanation, geometric vanation, traffic variation, and signal control param- eter variation, especially actuated control parameter variation (minimum green, maximum green, vehicle extensions, phase sequence, etc.); however, when searching for suitable study sites, selection is limited to those conditions that can be found in the field. Site Descriptions Based on previously stated criteria and the available intersections under vehicle-actuated control In the Bryan/College Station area, two sites were selected for this pilot study--FM 158 at Copperfield Dnve and University Drive at Spring Loop. These sites provided the desired geometnc, traffic, and signal timing variation; however, they by no means represent the entire range of vehicle-actuated operating conditions found in the field. Figures H-l and H-2 provide additional details for the two pilot study intersections. / \ 14 Copperfield Dr. ~ 14 F.M. 1 58 , -. ~3 Controller | Casnma Loca~on Bryan, Texas (semi-rural area) Geometry The main street (FM 158) runs north/south with one through-right lane (12-foot lane and 10 foot shoulder) and one leR-=ning lane in each direction. The minor street (Copperfield Drive) runs east/west and has two lanes (approximately 16 feet wide) in each direction. These lanes, however, are unmarked and there are no turn lanes in the east/west direction. Traffic Volumes The tragic volumes at this intersection are relatively low. Signal Con~o! Type The controller used at this intersection is a Kentron eight-phase actuated controller. The intersection is filly actuated with a single phase for the minor street movements, and Me ability to skip the leD-turn phase on the main street whenever left-turning vehicles are not present. Figure Ho-. FM 15X at Copperfield Drive Appendix H: Page 6

OCR for page 189
University Drive Spri - 1: Lit icon lg Loop Controller Cam era Location College Station, Texas (suburban area) Geometry The main sheet University Drive) runs east~rest with one ~rougb-ngbt lane, one through lane, and one left-turn lane in each direction. The minor street (Spring Loop) runs north/south with a through-nght lane and a left-tu~n lane in He southbound direction, and a through-left lane and a nght-turn lane in the northbound direction. Traff c Volumes The traffic volumes at this intersection are relatively high. Signal Con*o! Type The controller used at this intersection is an Eagle eight-phase actuated controller. The intersection is fillly actuated win split phasing on the minor sheet and protected/permitted lefc-turn phasing on the main street. Figure H-2. University Drive at Spring Loop Data Collection As mentioned several types of data were collected at the two pilot study sites. Video and electronic data were collected for three days at each site and used to evaluate the signal timing estimation models. Additionally' manual stopped delay data was collected for two hours at each site and was used to evaluate the vehicle-actuated delay models. The types of data collected and the methods of collection are summarized in Table H-] and described on the following pages. Appendix H: Page 7

OCR for page 189
Table H-~. Types of Data and Methods of Collection Type ofData | Method of Collection l Fixed Controller Parameters r Hard copy from City for the Period of Study Signalization Status on LED Display Data Using Photocells aSecond-by.-Second Basis | (laptop used to reduce date) Queue Lengths Manual On-site Counts Every 15 Seconds Continuous Tube Counters and Video Traffic Volume Counts I FM 158 and Copperfield Drive Data was collected at the FM 158 site on June 9 through 12, 1995. The following tasks were per- formed to collect the required data: Two weatherproof video cameras were installed at the top of the lum~naire located next to the controller. See Figure H-] for the location of the controller and the direction of the cameras. Continuous video was recorded with time-lapse VCRs from approximately 6:00 a.m. till approximately 9:00 p.m. each day. Tube counters were placed at appropriate distances on each approach to record the approach volumes for the entire data collection period. Traffic volumes were recorded in ~ 5-m~nute blocks. A manual stopped delay count was conducted from 7: ~ 5 to 8: ~ 5 a.m. and from 8:30 to 9:30 a.m. on Monday, June 12. Four people (one for each approach) counted the number of vehicles stopped at each approach every ~ 5 seconds. University Drive and Spring Loop Data was collected at the University Drive site June 12 through 15, 1995. The following tasks were performed to collect the required data: Two weather proofvideo cameras were installed at the top of two luminaries located diagonal to each other across the intersection. See Figure H-2 for the location of the controller and the direction of the cameras. Continuous video was recorded with time-lapse VCRs from approximately 6:00 a.m. till approximately 9:00 p.m. each day. Appendix H: Page 8

OCR for page 189
Tube counters were placed at appropriate distances for each approach to record the approach volumes for the entire data collection period. Traffic volumes were recorded in 15-m~nute blocks. A manual stopped delay count was conducted, from 3: ~ 5 to 4: ~ 5 p.m. and from 4:30 to 5:30 p.m. on Wednesday, June 14. Four people (one for each approach) counted the number of vehicles stopped at each approach every 15 seconds. Data Reduction The data reduction process included four steps: (~) determining the fixed traffic signal control parameters; (2) summarizing the variable vehicle-actuated control parameters (green times and cycle lengths) from the electronic data files and/or video tapes; (3) counting turning movements (from videos In time periods that the stopped delay data was collected; and (4) summarizing the queue data obtained in the field. Data AnalYsis To est~rnate delays using the proposed vehicle-actuated delay models, the 1 5-minute data collected In the field (saturation Dow rates, turning movement counts, average green times, cycle lengths, etc.) were used as inputs to the vehicle-actuated delay models. The estimated delays from the models were then compared to the measured delays Tom the field to evaluate how well the models did or did not match reality. Specifically, the estimated delays were plotted on the years against measured delays on the x-ax~s to compare how wed the estimated delays matched the measured delay. The plot also includes a 45- degree line (i.e., the line on which both measured and estimated delays are equal) to aid in this comparison. Data points above this line represent estimated delays higher than measured delays, and data points below this line represent estimated delays less than measured delays. Finally, regression analysis (Ho pl-l.O) was used to determ ne whether estimated and measured delays were significantly different. STUDY RESULTS The vehicle-actuated delay mode! inputs were measured in the field and then plugged into the two models to provide delay estimates for the conditions observed in the field. The one exception to measured inputs was saturation flow rate which was estimated to be ~ 900 vphgpI (vehicles per hour green per lane) based on studies at similar intersections in the area. Measured stopped delay values were manually calculated using traffic volume counts from the video tapes and queue counts per- fonned In the field. The resultant measured delays were then compared to the estimated delays from the Fambro-Rouphai! and Ak,celik-Chung models. The following sections summarize the results of those comparisons. Appendix H: Page 9

OCR for page 189
FM 158 at Copperfield Drive The average cycle lengths and green times for the intersection were collected using a lap top com- puter arid an interface unit that monitored the controller's status. The estimated and measured delays by approach are shown In Table H-2 and Figure H-3 for the Fambro-Rouphai} delay mode! and Table H-3 arid Figure H-4 for the Ak~elik-Chung model. Note that both proposed delay models produce similar results; i.e., estimated delays are close to measured delays for the north/ south approaches, and estimated delays are higher than measured delays for the east/west approaches. Statistically, the regression models' slope and mean square error were i.7 and 55.5 for the Fambro-Rouphai! model, and I.9 and 66.4 for the Ak~elik-Chung model. A single phase and a single exclusive lane for through traffic are provided on the major street (FM 158~. The north/south approaches to the intersection also have relatively low traffic volumes and long green times. As a result, the degree of saturation on these approaches ranged Dom 0.08 to 0.29, and both the estimated and the measured stopped delays are less than 5.0 sec/veh. Note that for both approaches and both delay models, the estimated delays are very close to the measured delays Dom the field. Single phase and single shared lanes for less, throughs, and rights are provided for the minor street (Copperfield Dnve). The east/west approaches to the intersection have relatively low traffic volumes and short green times. The green times provided for these movements depend on the westbound approach volumes because it is the critical movement In the chase (westbound approach volumes are noticeably higher than eastbound approach volumes). Table H-2. Stopped Delay (sec/veh) at ~ 158 and Copperfield Drive (Fambro-Rouphail) Time | Northbound | Southbound | Eastbound | Westbound l eriod ~ Measured Estimated Measured Estimated Measured Estimated Measured 7:15-7:30 2.3 3.0 0.7 2.9 3.0 19.9 7:30-7:45 2.1 3.8 2.4 1.6 14.3 16.6 7 :45-8 :00 3.2 3.0 2.1 3.1 10.0 21.2 8:00-8:15 1.1 2.8 1.1 2.7 5.6 26.2 8:30-8:45 3.2 2.4 1.9 1.1 18.0 27.9 8:45-9:00 0.9 2.5 1.3 0.9 18.8 28.6 9:00-9: 15 2.7 4.7 1.4 0.9 15.0 30.9 9:15-9:30 2.4 3.7 2.6 0.4 7.5 31.3 Estimated 13.1 14.0 18.9 10.6 21.1 18.1 22.6 29.2 24.2 10.0 20.0 12.1 29.2 30.0 32.1 32.2 Appendix H: Page 10

OCR for page 189
~ u 1 ;; ' ' '';' ' '; ;;;;;;;;;;;;;;;;;;;; ;; ;; ;;;;;;;;;;;;;; o ONorth XSouth ~ E a s t ~ W e s t 0 20 40 60 Measured Stopped Delay (sec/veh) Figure H-3. Stopped Delay at HI 158 and Copperfield Drive Table H-3. Stopped Delay (sec/veh) at FM 158 and Copperfield Drive (Ak~elik-Chung) Time ~ NortI bound ~Southbound ~Eastbound ~Westbound Period | Mc~umd Evened Measured Evened Mo~u~t End Metered EARLY& 7:15-7:30 2.3 3.2 0.7 3.1 3.0 22.3 13.1 23.8 7:30-7:45 2.1 4.2 2.4 1.7 14.3 18.6 14.0 20.5 7:45-8:00 8:00-8: 15 3.2 4.2 3.2 2.1 8:30-8:45 3.2 8:45-9:00 0.9 9:00-9:15 2.7 9:15-9:30 2.4 3.4 2.9 2.6 1.9 2.7 1.3 5.1 1.4 14.3 10.0 5.6 18.6 14.0 23.4 18.9 28.6 10.6 25.2 32.4 __ . . . 1.1 18.0 30.6 24.2 32.6 1.0 18.8 31.4 10.0 33.5 1.0 15.0 33.5 20.0 35.6 5 7.5 34.1 12.1 35.8 Appendix H: Page 11

OCR for page 189
- - ~ 4C - Q o in ._ In o , 2., 0 20 40 60 Measured Stopped Delay (seciveh) Figure E-4. Stopped Delay at FM 1SS and Copperfield Drive As noted, estimated delays on the east/west approaches to the intersection were higher than the measured delays. This discrepancy is a result of the high right-turn volume (50 percent of total traffics, the 16-foot wide approach lanes, and the fact that delays estimated on both approaches included right-turn-on-red and right-turn free-queue vehicles in the traffic volume counts. In other words, many vehicles in the field were not actually delayed as long as predicted by the models (i.e., they had more capacity and a shorter effective red). Thus, the estimated delays predicted by the models should have been higher than the actual delays measured in the field. These results provide further support for both models' credibility. University Drive and Suring Loon The average cycle lengths and green times were determined using a stop watch and the video tapes of the intersection. The estimated and measured delays by approach are shown in Table HE and Figure H-5 for the Fambro-Rouphai} mode} and Table H-5 and Figure H-6 for the Ak~elik-Chung model. Again, note that both models produce similar results; i.e., estimated delays are higher than Appendix H: Page 12

OCR for page 189
Table H-4. Stopped Delay (sec/veh) at University Dr. and Spring Loop ~ambro-Rouphail) Northbound ~Southbound ~Eastbound ~Westbound Time Period ~Measured F.ctima~d Measured Fct~mated Measured Estimated Measured Fstirnated 3:15-3:30 31.2 28.7 40.5 28.7 3.6 14.7 4.4 15.0 3:30-3:45 40.7 25.6 33.1 30.1 5.1 15.6 4.5 16.0 3:45-4:00 58.6 45.7 50.6 44.4 3.3 12.6 6.9 12.4 4:00-4:15 62.7 43.1 45.3 42.4 3.3 14.2 6.0 15.1 4:30-4:45 69.4 43.7 52.8 41.8 4.2 15.1 7.6 14.5 4:45-5 :00 57.1 40.7 60.6 43.0 4.0 16.3 6.1 16.4 5:00-5:15 50.2 44.7 53.6 45.5 9.4 20.6 22.5 18.6 I ':15-5:30 70.0 41.5 53.7 42.s 8.8 20.3 14.4 19.5 80 60 . . ~ .. . . . . . .. .. . . ..... ......... . . . . . ~ .. ... . ~ . . ~ ..... . ONorth XSouth ~ East Newest . o 0 20 40 M easu red Delay (sec/veh) 60 80 Figure lI-5. Stopped Delay at University Drive and Spring Loop Appendix H: Page 13

OCR for page 189
Table H-5. Stopped Delay (sec/veh) at University Dr. and Spring Loop (Ak~elik-Chung) Northbound ~Southbound ~Eastbound ~Westbound Time e 4 ~ MGS=~ Ed need M=~umd Evened M=ed ED Mood ED 3:15-3:30 31.2 31.6 40.5 31.4 3.6 15.6 4.4 16.1 3:30-3:45 40.7 28.5 33.1 33.1 5.1 16.8 4.5 17.1 3:45-4:00 58.6 49.9 50.6 48.4 3.3 13.4 6.9 13.2 4:30-4:45 69.4 48.0 52.8 45.4 4.2 16.0 7.6 15.5 4:45-5:00 57.1 45.0 60.6 46.6 4.0 17.3 6.1 17.4 5:00-5:15 50.2 47.9 53.6 47.9 9.4 21.8 22.5 19.7 5:15-5:30 70.0 45.6 s3.7 4s.7 8.8 21.6 14.4 20.8 80 ONorth XSouth AEast fewest 0 20 40 60 80 M e a s u re d S to p p e d D e la y Is e c/v e h ~ Figure H-6. Stopped Delay at University Drive and Spring Loop Appendix H: Page 14

OCR for page 189
measured delays for the east/west approaches, and estimated delays are lower than measured delays for the north/south approaches. Statistically, the regression models' slope and mean square error were 0.77 and 79.0 for the Fambro-Rouphai! model, arid 0.84 and 87.6 for the Ak~elik-Chung model. Exclusive/permissive left-turn phasing, and one through and one through-right lane are provided on the major street (IJniversity Drive). The east/west approaches to the intersection have moderate traffic volumes and relatively long green times. Estimated delays are higher than measured delays because the intersection is adjacent to a coordinated system and derives some benefits Tom progression (lower delays), especially in the eastbound direction. In other words, some platoon~ng and progression effects are present in the eastbound traffic and delays are being overestimated. When adjustments are made to account for these effects (calculating appropriate progression adjustment factors) eastbound estimated delays decreased significantly and more closely matched the measured delays (Table H-6 and Figure H-7 for the Fambro-Rouphai! Model and Table H-7 and Figure H-8 for the Ak~elik-Chung Model). Split phasing, and one exclusive turn lane and one shared through-turn lane in each direction are provided for the minor street (Spnng Loop). The north/south approaches to the intersection have moderate traffic volumes and relatively short green times. Estimated delays are higher than measured delays because saturation flow rates were overestimated (degrees of saturation were underest~rnated). This problem became evident when long delays and cycle failures (oversaturation) were noted on the video tapes and yet the estimated degrees of saturation were less than 0.7. These conditions are not consistent with one another. Table H-6. Stopped Delay (sec/veh) at University Dr. and Spring Loop (E:astbound Estimated Delay Adjusted for Progression) (Fambro -Rouphail) 4.4 4.5 6.9 6.0 | ime | Northb 3und | Southbound | Eastbound | Westbound ~ Period | Measured Es~na~d Measured End Mound Es~na~d Meowed 3:15-3:30 31.2 28.7 40.5 28.7 3.6 4.3 3:30-3:45 40.7 25.6 33.1 30.1 5.1 4.7 3:45-4:00 58.6 45.7 50.6 44.4 3.3 4.4 t:00-4: 15 62.7 43.1 45.3 42.4 3.3 5.0 4:30-4:45 69.4 43.7 52.8 41.8 4.2 5.1 4:45-5:00 57.1 40.7 60.6 43.0 4.0 6.0 5:00-5:15 50.2 44.7 53.6 45.5 9.4 12.3 5:15-5:30 70.0 41.5 53.7 42.5 8.8 9.5 Fatima 15.0 16.0 12.4 15.1 7.6 6.1 22.s 14.4 14.5 16.4 18.6 19.5 Appendix H: Page 15

OCR for page 189
20 0 20 40 Measured Delay (sec/veh) 60 80 Figure lI-7. Stopped Delay at University Drive and Spring Hoop Table H:-7. Stopped Delay (sec/veh) at University Dr. and Spring Loop (Eastbound Estimated Delay Adjusted for Progression) (Ak~elik-Chung) Tune ~ Northbound ~Southbound T Eastbound ~Westbound ~ Penod ~ Measured Emoted Measured Estimated Measured Estimated Measured Estimate 3:15-3:30 31.2 31.6 40.5 31.4 3.6 4.5 4.4 16.} 3:30-3:45 40.7 28.5 33.1 33.1 5.1 4.9 4.5 17.1 3:45-4:00 58.6 49.9 50.6 48.4 3.3 4.6 6.9 4:00-4:15 62.7 47.5 45.3 46.2 3.3 13.2 _ _ 5.3 6.0 16.0 4:30-4:45 69.4 48.0 52.845.4 42 5.3 7.6 15.5 4:45-5:00 57.1 45.0 60.646.6 4.0 6.3 6.1 17.4 5:00-5:15 50.2 47.9 53.647.9 9.4 12.8 22.5 19.7 5:15-5:30 70.0 45.6 53.745.7 8.8 9.9 14.4 20.8 Appendix H: Page 16

OCR for page 189
80 - ~ 60 to - ~ 4 0 a oh ,, 2 0 .. ................................... ............... . .;;;;;; ; 0 20 40 60 80 M easu red Sto p Fed Delay (sec/veh) Figure H-8. Stopped Delay at University Drive and Spring Loop This inconsistency is a result of the north/south approaches to the intersection narrowing to a single lane about 100 feet upstream ofthe stop bar. Thus, even though there were two lanes at the stop bar, queues of more than four or five vehicles blocked entry to one of the lanes and reduced the saturation flow rate and capacity of the downstream signal. During these blockages, the effective degree of saturation in the field was much higher than estimated by the models. If this adjustment was made, the e~echve capacity would have been reduced, and estimated delays would have increased and more closely matched the measured delays in the field. Appendix H: Page 17

OCR for page 189
OBSERVATIONS AND CONCLUSIONS The pilot study evaluated two vehicle-actuated delay models and tested a data collection procedure for collecting delay mode} inputs and outputs. The following observations and conclusins from this study are offered 1. Both vehicle-actuated delay models produced similar results when compared to field data. Statistically, the Fambro-Rouphai! mode] performed slightly better than the Ak,celik-Chung model; however, the difference was not significant. Thus, there is no basis to recommend one mode! over the other at this point in tune. 2. The data collection procedure (video, electronic, and manual) worked well. It was relatively simple to set up and execute, and in the case of an equipment failure, the redundancy provided by the video tapes allowed any missing data to be recovered. 3. 4. Right-turns-on-red and free-queue right turns increase a signal's elective capacity, thus reducing delay in the field. Since current delay models do not address these conditions and neither of them is an objective of this study, selection of fixture study sites should strive to minimize their occurrence. Progression effects have a significant impact on delay in the field. Thus, they must be taken into account for studies of this type. As these effects exist at almost all signalized intersections in urban areas, fixture field studies should incorporate a mechanism to quantify these effects before comparing estimated to measured delays. APPENDW :H REFERENCES Highway Capacity Manual. TRB Special Report 209, Transportation Research Board, National Research Council, Washington, D.C., 1985. Tarnoff, P. I., Selecting Traffic Signal Control at Individual Intersections. National Cooperative Highway Research Program Report 233, Transportation Research Board, National Research Council, Washington, D. C., June ~ 98 I. Capacity Analysis of Traffic-Actuated intersections, Draft Intenm Report, NCHRP 3-4S, Transportation Research Board, National Research Council, Washington, D. C., January 1995. Appendix H: Page 18