Click for next page ( 8


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 7
CHAPTER 2 FINDINGS 2. 1 SUMMARY OF EXISTING CONDITIONS A comprehensive evaluation of the state-of-the-artin areas related to interchange design and traffic operations was conducted as part of this research. This evaluation consisted of a survey of practitioners and a review of existing traffic models. The focus of this evaluation was on issues underlying the design and operation of interchanges in urban or suburban areas. More specifically, the focus was on issues related to the signal-controlled ramp terminals and traffic flow along the cross street through these terminals. Consideration was also given to the relationship between the interchange ramp terminals and any adjacent, closely-spaced signalized intersections. 2.1.! Survey of Current Practice The intent of this survey was to gain insight into the current practices arid concerns of engineers who are responsible for interchange traffic operations. The survey was conducted in two stages. The first-stage survey was intended to obtain basic types of interchange-related infonnation such as common interchange types, traffic flow problems, and operational analysis techniques. The second-stage survey was designed to obtain more detailed information about interchange operations. This survey asked the respondent to select one interchange that they were familiar with and then respond to detailed questions about its operation and any steps taken to alleviate flow problems at this interchange. The respondent was also asked to describe the analysis techniques (or computer models) Hat they had successfi~ly used to evaluate interchange operations. The findings from these two surveys are summanzed in this section. A more detailed discussion of He survey findings is provided in Appendix A. Distribution. The first-stage survey was sent to more than 2,400 transportation engineers in the U.S. and abroad. The members of the American Association of State Highway and Transportation Officials' (AASHTO) subcommittees on traffic engineering, on design, and on transportation systems operation were specifically targeted. A large number of the Institute of Transportation Engineers ' (ITE) Urban Traffic Engineers Council and its Consultants Council were also included in the survey. In addition, several hundred surveys were sent to over selected members of ITE. After a review of each returned questionnaire, a total of 350 first-stage questionnaires were deemed completely responsive and valid for furler processing. Overall, there were 146 responses from the public sector which included state, city, and county highway agencies. Seventeen responses were received from outside of the United States. Responses were also received from ~ 87 consultants in 23 states. 2

OCR for page 7
The second-stage survey was sent to 1 90 individuals who responded to the first survey. A total of 3 1 completed surveys were returned, representing a 16 percent response rate. Of these surveys, 29 were deter~ninedto be valid responsesin the context that they addressed the interchange types and issues described in the survey. Overall, 21 states are represented among the 29 valid returned surveys. Findings. The first-stage questionnaire consisted of six questions that were primarily of the multiple-choice type. The second-stage questionnaire consisted of eleven questions, several of which had follow-up questions. In general, these questions inquired about the kinds of interchanges being used or constructed, the type of signal control used, the types of operational problems found at existing interchanges, and the methods used to evaluate and mitigate these problems. The responses to the questions on both questionnaires are summarized in the following paragraphs. The diamond interchange was found to be the most commonly used interchange configuration. This trendis likely due to the reduced right-of-way end construction costs associated with diamond interchanges relative to other configurations (e.g., partial cloverleaf). The distance between the diamond interchange ramp terminals can vary from 60 meters in densely-developed urban areas to 240 meters in suburban areas. In contrast, the distance between ramp terminals associated with a partial cloverleaf interchange generally range from 180 to 280 meters. Regardless of configuration, the interchanges that tend to experience operational problems are those with relatively short distances between the ramp terminals or between one terminal and an adjacent signalized intersection. These close spacings often lead to problems such as queue spillback, flow turbulence due to weaving, and left-turn bay overflow. Queue spillback represents the blockage of an upstream intersection by a traffic queue from a downstream intersection. The interchanges described by the survey respondents as having operational problems had ramp terminal distancesin the range of61 to 410 meters. The distance to the adjacent intersection for these same interchanges was in the range of 46 to 436 meters. The survey indicated that most interchanges have two semiactuated signal controllers, one controller for each ramp terminal. The two controllers are typically coordinated to facilitate progressed traffic flow along the arterial and minimum queuing on the street segment between the two terminals. Some interchanges have pretimed control with either one or two controllers. The few diamond interchanges that were pretimed and had one controller used four-phase-with-overlap phasing. Only a few interchanges had fi~ll-actuated, uncoordinated control. The distribution of operational problems found in interchange areas is shown in Figure 2. As this figure indicates, the operational problem that occurred most frequently is queue spillback at some junction on the cross street. This problem was generally related to the spilling back of a queue from a downstream ramp terminal or intersection into an upstream terminal or intersection. This spillback tended to significantly reduce the car acitY of the upstream junction. Also included in this ~Or i- - -a category is spillback stemming from a left-turn bay overflow. 2 - 2

OCR for page 7
Percent of Responses Spillback on Cross Street Other Spillback Poor Signal Timing Weaving-related Unbalanced Lane Volumes 0 10 20 30 40 50 Figure 2. Distribution of operational pro bZems in urban interchange areas. The reported flow problems related to queue spillback between the ramp terminals were generally associated with tight or compressed diamond interchanges. Flow problems related to queue spillback between a ramp terminal and adjacentintersection were more commonly associated with conventional (wide) diamond interchanges and partial cloverleaf interchanges. The wide spacing between ramp terminals for these interchanges tends to be associated with shorter distances between these terminals and the adjacent intersections. By design, the single point diamond configuration does not experience spillback between its terminals; however, it can experience spillback between it and the adjacent intersection during high-volume conditions. Other frequently cited problems at interchanges include unbalanced large volumes on the ramp terminal approaches, flow turbulence due to weaving, and a lack of effective signal coordination between the ramp terminals. The unbalancedlane volume problem stems from frequent driver propositioning for downstream turns in interchange areas. Drivers desiring to turn left (right) at a downstream intersection tend to move into the inside (outside) lane of a multilane lane group at the upstream intersection. This propositioning effectively reduces the capacity of the lane group by leaving some traffic lanes underutilized, even during high volume conditions. The weaving maneuver that is predominate in interchange areas is the off-ramp right-turn movement that weaves across the arterial to make a left-turn at the next downstream signalized intersection. This maneuver typically has a high volume associated with it such that considerable turbulence is created on the cross street. This turbulence results in significant speed reductions to the nonweaving traffic movements. 2 - 3

OCR for page 7
The lack of effective signal coordination along the cross street in interchange areas occurs for a variety of reasons. These reasons generally include incompatibility ofinterchar~ge phasing with the cross street system coordination plan and institutional barriers (i.e., the state operates He interchange and the city operates the adjacent intersection). The lack of efficient signal coordination can lead to increased delays and stops and precipitate the occurrence of spillback when the ramp terminals or intersections are closely spaced. A wide range of methods were described by the respondents for alleviating the aforementioned operationalproblems. Geometric improvements were most commonly cited. These improvements included adding a second left-turn lane or an additional through lane to the cross street. Many respondentsindicated that improved or updated signal timing and coordination helped mitigate some operationalproblems. These latterimprovements were often obtained through the use of existing software-based traffic analysis models. In general, software programs are more frequently used than manual methods for evaluating interchange traffic operations. The most commonly used software program is the signalized intersection analysts procedure included in the Highway Capacity Software (HCS). In general, this procedure was used to evaluate the individual ramp tenninals after appropriate calibration of the progression adjustment factors to account for nearby intersections. The popularity ofthis program may be due to its widespread acceptance by transportation engineers, its consistency with the methods described in Chapter 9 ofthe Highway Capacity Manual (HCM) (3j, and the relative ease with which it can be used. The most frequently cited strength of this program is that it is easier to use than multiple-intersectionsoftware programs (e.g., PASSER II, TRANSYT-7F,NETSIM, etch. Ofthe various software programs available, TRANSYT-7F was cited by nearly half of all the respondents as being used! for analyzing interchange operations. This finding may be due to the fact that TRANSYT-7F is sensitive to the proximity of adjacent ramp terminals or signalized intersections in its signal timing optimization routine. Another software model, PASSER- was also cited by many of the respondents as being a usefi~] too] to analyze arterial traffic flow through interchange ramp terminals. In the case of this latter model, the large response may be due to the fact that PASSERS optimizes signal phasing based on progression analysis. NETSIM was used by some of the respondents. This program was noted to be the only one that modeled queue spilIback and congested flow conditions. The respondents also noted that the existing software programs had some weaknesses that limited their ability to accurately mode} interchange traffic operations. The weaknesses cited for the HCS program (i.e., the HCM Chapter 9 procedure) were that it did not accurately model the effect of closely-spaced upstream intersections and that it did not yield queue length estimates. The weaknesses cited for PASSER Il were Hat it did not provide progression solutions for left-turn movements, did not consider queues when determining progression, did not allow the user to enter some types of interchange phasing, and did not fillly consider right-turn demand. NETSIM was noted to be very time consuming to use due to its microscopic simulation formulation. A couple of respondents nosed that none ofthe programs dealt explicitly with the coordination of a downstream 2 - 4

OCR for page 7
ramp meter with the ramp terminal. Further evaluations of these computer-based models, primarily for research applications in this project, watt be presented later in this chapter The survey found that the most commonly selected measure of effectiveness (MOE) for evaluating interchange traffic operations was traffic signal delay, followed by spilIback frequency, and volume-to-capacityratio. Delay was likely chosen by the practitioners because it represents the most tar~gible measure of effectiveness that is also comprehensible by the motoring public. After delay, queue spilIback frequency was the next most frequently cited MOE by the respondents. 2.~.2 Field Survey of Interchange Operations The research team studied over a dozen service interchanges dunug the field studies and spent marry hours observing traffic operations at the sites. Comparisons could be rapidly made among interchange types, types of operational problems observed, and the hypothesized probable cause of these problems. Our summary of these field sites having congested operations are noted below: Designlife of interchange probably exceeded, overall traffic demand exceeded interchange capacity during rush hours. Due to growth in suburban areas, older four-lane crossing arterials now need to be six lanes. The average daily traffic on many of We four-lane crossing arterials exceeded 30,000 ADT. Many "next" downstream signalized intersections along Me crossing arterial experience high access demands toffrom the Leeway (interchange) and are routinely too closely spaced to provide good operating conditions. Better access management, intersection spacing and design policies are needed. Traffic management of quelling and spilIback is difficult at interchanges due to high volumes and high percentages of turning traffic having typical lane distribution problems. Some approaches along the crossing arsenal and within the interchange can have almost constant demand within Me cycle, so queuing can not be mitigated using traditional signal coordination techniques. Four-quad parclos would seem to be more susceptible to constant demand conditions within the interchange because oftheir free flowing loop ramps. All parclo interchanges,including the four-quad AB that exits both left and right turns from the same side of a cross arterial approach, may experience high lane imbalances ofarrival flow on that side ofthe street, even at intersections along the crossing arterial upstream of the interchange. Many ofthe congested interchanges noted above had a predominant number of single-lane left turn bays within the interchange and/or have single lanes assigned on approach ramps 2 - 5

OCR for page 7
at ramp terminals to serve left and/or right turning movements. Many approach ramps were single lane with only a modest flare to a two-lane approach at the ramp terminal. 6. e Traffic actuated operations on high-volume, single-lane movements appear to result in excessively long cycles that reduce the overall input capacity of the interchange. Protected- permissive left-turn operations, while reducing delays during moderate traffic, loses capacity during rush-hour conditions and, consequently, cannot be depended upon to provide significant capacity increases during these critical times. 7. Most traffic control strategies employed appear to be based on undersaturated flow conditions and may lose efficiency when oversaturated conditions arise. Management of queue spillback to mitigate the onset of congestion is needed together with the need to transition to downstream bottleneck control strategies once oversaturation has occurred. 2.2 SURVEY OF EXISTING TRAFFIC MODELS The first-round survey inquired about the types of analysis methods used to evaluate (not optimize timing) signalized interchange traffic operations. In general, software models were more frequently used than manual methods. The most commonly used software method is the Highway Capacity Software (HCS). PASSER II and TRANSYT-7F were also found to be frequently used in practical engineering applications. However, research applications usually require more complex computer simulation models than application-specific models like HCS and PASSER II. Computer simulationis a viable method with which to analyze situations which may occur at signalizedinterchanges,but for whatever reason are difficult to witness or collect date from field studies. This investigation was primarily based on literature and manuals for each model, and discussions withindividuals familiar with the models. Experience with each model is arguably the most informative method of discovering what a program can and cannot do. Time constraints always limit the depth with which each ofthese models can tee investigated. A list ofthe simulation models investigated is included in Table l. Simulation models can be described by their analysis approach, basis, objective and outcome. A model's analysis approach is either macroscopic or microscopic. A macroscopic simulation model is one in which the traffic stream is moved as one homogenous aggregate group, whereas a microscopic simulation model is vehicle specificin which each vehicle moves as its ownidentifiable entity. A simulationmodel's analysis basis is either empirical or analytical. The analysis basis refers to the algorithm on which the model is based. An empirical model is based on field observations or data and\or previous experience. Analytical models use mathematical formulas based on theoretical relationships. The analysis objective refers to the purpose of the simulation model. Models simulate traffic given certain geometric constraints, and/or optimize some specific traffic parameter. Lastly, a simulation model is described by its analysis outcome, which is either stochastic or deterministic. A stochastic model attempts to model human behavior by providing a 2 - 6

OCR for page 7
degree of randomness to its methodology. In this way the output is never the same given a set of inputs. Given the same Inputs, a detennin~stic model would have the same output every time the same data is input. Each model's analysis is also given in Table I. Table I. Simulation Models Examined Model FREFLO . . l _ Analysis i| Description ~Approach | Basis | Oboe ctive | Outcome ' 1 Freeway Macroscopic | Analytical | Sim ration Simulation Freeway and Microscopic Analytical Simulation or Surtace Street Optimization Network Model Freeway and Macroscopic Analytical Simulation Surface Street Network Model Freeway and Microscopic Analytical Simulation or Surface Street Optimization Network Model Urban Street Microscopic Analytical Simulation Network Model Signalized Macroscopic Analytical Optimization Diamonci Interchanges Signalized Multi- Macroscopic Analytical Optimization ~ Intersections | Signalized Multi- Macroscopic Analytical Simulation or Intersections Optimization Isolated Microscopic Analytical Simulation Intersection ~Analysis r COmplete Macroscopic Empirical Simulation Implementation of 1985 HCM Freeway Weaving Macroscopic Empirical Simulation Analysis , 1 Deterministic INTRAS 1 980 Stochastic CORFLO Deterministic INTEGRATION Version 1 5 Unknown 11 NETSIM Stochastic PASSER 111 1 990 Deterministic PASSER 11 1 990 TRANSYT-7F TEXAS Version 3.11 HCS Deterministic Deterministic Stochastic Deterministic FREWEV Version 1.1 Deterministic 2 - 7

OCR for page 7
2.2.1 Input and Output Obviously, each mode} has a required amount of input. Mar~y models have options that may or may not be important to ~is project, and therefore ~e input for some data is optional. An abbreviatedlist of model inputsis included~n Table 2. The table indicates~e inputs (bosh requ~red and optional) by each model. The list is not all-~nclusive. Model names were abbreviated in Table 2, but they are presented in ~e same order as they are listed in Table I. t Table 2. Mode! Inputs Input Model | FRE | INT | COE. | ITG | NET | PIII | PII | T-7F | TX | HCS | WEV lus Stop Delay 2apaciiy ~X ~ X ~X ~ X ~ X ~river & Vehicle X X 'haractenstics l l l l l l I :: ~rades ~X Horiz. Curve Data X Incident Data X ntersection Spacing l ~l ~I X | X ,ink Lengths | X | X | X | X | X | X | ,oad Factors Numb~er of X X X X Approaches Numb~er of I ~neS X X XX X X X X X X )-D Travel Pauerns l l l | X | X l l 'edestrianAcolation l l l | X X 'ercentageof l ~X l l l l l X |Vehicle Types Ramp Metering Rate X X Rte. Detouring Data X Saturation Flows X X X Signal & Sign ~| X l | X | X X ~ X X ~ X X Control Parameters Simulation X X X X X X X peed l l l l l l | average X X X X free flow X X X X X Through Volumes X X X X X X X X X X r~ng percentages ~ x ~ ~ x ~ ~ ~ ~ ~ ~ F ~Turning Volumes X X X X X X X X . Vertical Curve Data l | X l l l l l l I | X X X X X X 2 - 8

OCR for page 7
The ou~ut available for each model investigated is ~ncluded ~n Table 3 The table does not include all output for every model. FREFLO, INTRAS, CORFLO, INTEGRATION, NETSIM, and TRANSYT-7F display most of its output on a link specific basis. TEXAS Model provides output by lane, approach, and for the ~ntersection as a whole. Table 3. Model Outputs OIl~Ut ~ _ Degree of Saturation | I r | | x | x | x | I I queue X X X X X smpped l l l l | X | l l | X Density I X _ L x I x ~ || uel Consumpd m | | X | l | | X | X | I I [ Graphical Simulation ll X T I x Level of Service L _ Il l X := Lane Changes l l l l l l l l l I 3-D Chart T | x I 1 X I I I ! I I I 11 Optimal Ti~ng X X X _ Person tmiPleSSes | X l | X ~1010 Queue Length l | ? I rX | X | | X | X | Saturation Flow l l l l l l | X | | X | ime mean X X X space ~nean | l l I ~ ~ ~ ~ 11 rime Space Diag. I I I I I I x I x I I I I Il rrave1 T v~eear~gveehm | X | X |~ 1 1 L 11 ~s XX X X X X ~ Volume 1 l I X TX I I I I x I I I ~ 2 - 9 x x

OCR for page 7
2.2.2 Summary of Mode! Capabilities An important decision in this project is what models should be used and how should they be used for research purposes. All the models investigated have some link to interchange and arterial operations. However, they may be used to develop relationships for situations where it would be difficult to collect field data. As a result, a list of geometric and operational characteristics, as well as other concerns, typical of interchange operations has been compiled in Table 4. Each model was then investigated as to its capability to model the stipulated geometr.c or operational characteristic. The results were shown in Table 4, and a brief discussion of the results follows. An interchange ramp terminal/frontage road operates differently from an arterial street due to the effect of the freeway and its ramps. For this reason, a model capable of simulating traffic on both arterial streets and freeways would be advantageous. INTRAS and CORFLO are the only two models investigated in this initial study capable of interacting freeway vehicles and arterial street vehicles. Because INTRAS is a microscopic model, a greater level of detail can be both input and extrapolated from INTRAS than from CORFLO. Weaving is another Important factor. A level of service can be assumed from FREFLO output (and CORFLO) for weaving areas such as an entrance ramp closely followed by an exit ramp. For INTRAS, entrance/exit ramp weaving is not specifically addressed in the manual; however, TTI has used INTRAS for freeway weaving analysis and has found the model to operas adequately. However, it is improbable that the logic used in FREFLO and INTRAS for a freeway weaving analysis can be applied to a Interchange ramp terminal weaving sections. Other weaving scenarios Involve the interaction of vehicles exiting the freeway and requiring a right turn at the ramp terminal intersection or vehicles turning out of a driveway and requiring a left turn at the first downstream intersection. These scenarios cannot be specifically modeled in INTRAS; however, INTRAS output does contain O-D charts which can quantify those maneuvers, and the output also quantifies the number of missed maneuvers. In other words, if a vehicle was destined to exit the freeway and turn right at the next intersection on the frontage road, but could not complete the maneuver, INTRAS includes this information in its output. NETSIM, on the other hand, is capable of traffic assignment parameters which could require a certain percentage of freeway exiting vehicles to turn right at the frontage road intersection. This process is, however, very complex and careful attention must be made to keep percentages of vehicle movements at each link equal to 100 percent. PASSERIII deals specifically with diamond interchanges at which such a weaving maneuver would take place, however, simulationof weaving in the vicinity of the intersection is beyond its scope. With interchanges being an integral part of freeway traffic management systems in some states, and with ramp metering becoming more prevalent, the issue of queue length could play an important role in freeway corridor operations. Queue length would aid in determining an adequate distance between a ramp exit or entrance and the interchange. Therefore, it would be desirable 2- 10

OCR for page 7
Table 4. Computer Simulation Model's Capabilities Computer Model Model Constraints ~ FREFLO ~INTRAS ~CORF~O ~ NETSIM ~PASSER III ~ PASSER II ~ TRANSY ~ TEXAS freeway Simulation ~Yes ~Yes ~Yes ~n/a ~n/a ~n ~n/a Frontage Road n/a Yes Yes Yes No No No Simulation nterchange Simulation ~No ~Yes ~Unknown ~Yes ~Yes ~Y s ~Yes driveways ~n/a ~Yes ~Yes ~Yes ~No ~it; ~No Type of Traffic Control No Ramp Stop, Yield, Stop, Yield, Stop, Yield, Pretimed or Signals Pre-timed Metering Fixed, Actuated Pretimed Signal Fixed Traffic- Signals or Control, 3 types Control, Some Control, Responsive Unsignali of Ramp Actuated Actuated Fixed Sequence zed Metering, Merge Control Control Signals and Diver e freeway Weaving ~LOS ~Yes ~ LOS Provided ~n/a ~n/a ~n' ~n/a Analysis _ Provided l rterial Weaving | n/a | Yes l No | nknown | n/a | n/ | n/a Analysis Caused by Two Closely Spaced Ramps arying Distance of | Yes | Yes l Yes | Yes | n/a | n/ | n/a ! I Weaving Area | rterial Weaving | n/a | O-D quip l No | Yes | No | N | No | Analysis Caused by Either Vehicle Exiting Freeway and Turning Right at Interchange, or Vehicle Turning From Driveway and Turning Left at Intersection . Varying Distance n/a Yes Yes Yes Yes No No Between Exit Ramp Terminal and Downstream Arterial | Inter sect ion U-Turn Area at n/a Yes Unknown Yes Yes No Unknown | Interchanges Exit Ramp Vehicles n/a Yes. All lanes Unknown Unknown No No No Able to Yield to Cross yield Arterial Traffic l I 'rediction of Queue | n/a | Unknow | Unknown | Yes | No | Ye | Yes Length at Intersections . sadistic Output At or ~No ~Yes ~Unknown ~ 1; Known ~No ~No t Unknown Near Capacity Levels ~ at V/C > 0.95 n/a No Yes No No Control, Stop, Yield, Pretimed, Semi actuated, or Full Actuated n/a n/a n/a Unknown Yes Yes No No Unknown 2- 11

OCR for page 7
2.6.3 Sensitivity Analysis Figure 27 illustrates the effect of arterial flow rate on weaving and arsenal maneuver speed. This figure shows the behavior of both models when all other factors are held constant. The values selected for these factors represent their respective average values as found in the database. The range of flow rates over which the two models are compared is larger than the corresponding range in the database. This extension was undertaken to show the overall behavior of each mode} when extrapolated to extreme (but realistic) values. The trends in Figure 27 show that both models predict an exponentially decreasing maneuver speed wad increasing arterial flow rate. This trend is somewhat consistent with the traditional speed-flow relationship for uninterrupted traffic streams in uncongested conditions. This figure also shows Mat the arterial maneuver speed is always higher than the weaving maneuver speed for the same flow rate. This trend is reasonable since the arterial vehicles enter the weaving section at speed while the weaving vehicles often must accelerate from a stopped (or sIowed) condition when departing the off-ramp. The trend toward convergence of the two models at higher flow rates is also reasonable as the weaving maneuver speed should approach the arterial maneuver speed as the capacity of the weaving section is neared. Maneuver Speed, m/s 10 8 6 4 o . 12 ~ ~~ ~61 M/ea~jn9 ~ - _ _ _ us = 13.4 m/s Nt -2 IL= 0 2 VW = 170 VPh , I , i I I O 500 1000 1500 2000 2500 3000 3500 Average Arterial Flow Rate, vph Figure 27. Effect of arterialflow rate on weaving and arterial maneuver speeds. 2 -60

OCR for page 7
2.7 RAMP VVEAVING CAPACITY MODEL The section of the cross arterial roadway between an interchange ramp terminal and a closely-spaced downstream intersection generally experiences operational problems, reduced capacity, and deteriorated Levels of Service (LOS) when the ramp-to-intersection weaving is heavy and difficult to perform. The more difficult traffic maneuver to perform usually is the off-ramp right turn trying to cross and then turn left at the next downstream intersection. When the downstream intersection is signalized, additional queuing in the left turn lane shortens the elective weaving length, resulting in increased operational problems. An additional operational constraint is the physical capacity of the ramp-arterial crossing maneuver. This maneuver usually operates like a freeway merge operation dunng rush-hour conditions because even free right-turn maneuvers are usually performed from a stopped position in queue. The Highway Capacity Manual (3) does not address arterial weaving. This section will present a method for estimating arterial crossing capacity based on NETSIM traffic simulation studies. Bow random and progressed flow conditions along the arsenal can be evaluated. Models to predict operating speeds in arsenal weaving sections are presented in Appendix E. 2.7.1 Study Methodology The expenmentaltestbed shown in Figure 28 was coded In TRAF-NETSIM to simulate Me study conditions. An arterial free speed of 60 km/in was assumed. The distance between the ramp terminal and the downstreamintersection was 200 meters. The ramp traffic, on yield control, made a right turn onto the arterial and then made a left turn at the downstream intersection. The arterial traffic went through the downstream intersection without making any turns. The strategy was to heavily Toad the cross weave with abundar~t demand, i.e., maintain a standing off-ramp queue so that the maximum ramp crossing volume could be observed for different operating conditions. l 111 l Diamond Interchanged Figure 28. Arterial testbedfor ramp-to-arterial weaving analysis. 2 - 61 1 1 1 1 1 1 1 1 ~ t .] Area of Study 1 1 1 1 ~1 1 1 1 1 1 1 1 1 ~1 ~ ~ '

OCR for page 7
Preliminary testing revealed that the weaving capacity from the ramp terminal follows the pattern of a negative exponential function with increasing arsenal volume. Thus, negative exponential regression analysis was performed to mode! the weaving capacity. The basic form of the exponential regression equation for the predicting ramp capacity is shown below. OR ~-e~PQ. where: OR Q. a T 1 C He ramp crossing/weaving volume (vph); arterial through volume (vph)' coefficient ofthe model = Tc / 3600; coefficient of the model = Hs / 3600, critical gap of ramp weave, see, and minimum follow-up headway, sec. The coefficients of the exponential equation, a and ,B, for random flow were determined on the basis ofthe simulations for various arsenal through volume conditions. The coefficients a and ,B were computed by inputting the simulated arsenal and the ramp crossing volumes into SAS, a statistical software analysis package (49, and perfonning the desired regression analysis. For the random flow conditions, the arterial traffic was vaned from 100 vph to 2000 Ash. Weaving across one, two, and three arterial lanes was studied for the volume conditions noted. Also, the effect of the change in decile gap acceptance distribution in NETSIM was studied. For progressed flow conditions, the arterial traffic was vaned from a v/c of 0.2 (500 vph) to a v/c of 0.8 (2000 vph) for a three lane arterial. A cycle length of 100 seconds and a clearance interval offour seconds per phase were also assumed. Various PF ranging from 0.l to ~ .8 were also simulated by varying Me percent vehicles arriving on green (PINGS at the upstream intersection. 2.7.2 Study Results The next section consists ofthe results obtained in the various cases involving random flow conditions along the arterial. Also, the computed coefficients for determining the ramp crossing volumes for different arterial flow conditions are presented. Changes in the gap acceptance distnbution were observed to affect the ramp crossing vol~ne. The second section presents the calibration coefficients for the proposed negative exponential equation for computing the ramp crossing volume for different arterial through volumes. The third section covers the results of simulations involving several volume conditions and different progression factors. The effect of progression on the ramp crossing volume is discussedin detail in this section. The development of the final mode} fonn and the methodology used to predict the ramp capacitor across the arterial weaving section for various progression factors are presented In Chapter 3 and Appendix E. 2 - 62

OCR for page 7
Random Flow. Initial ramp capacity studies were conducted with NETSIM assuming that the cross arterial had no progression end random flow. Moreover, preliminary testing ofthe mode! assessed the sensitivity of capacity to the default gap acceptance function provided in the model. The ~ 994 HCM states that the critical gap for a right turn from a Yield sign onto a major street could be taken as 5.5 seconds (3~. TRAF-NETSIM assumes a decile distribution wherein the default median value is taken as 6.4 seconds. In order to simulate the HCM recommended distribution, Card Type 145 in TRAF-NETSIM was coded to produce a decile distribution having a median value of 5.5 seconds. Hence the data file with the new decile distribution and an upstream link length of 365 meters was simulated for random flow. The effect of changing the decile gap distribution for three lanes can be seen _ . . ~. . . . . . in Figure 29 Due to the lower (better) gap acceptance cnter~a, more ramp vehicles can make a right turn onto the arterial. Though the trend is similar' the ramp crossing volume for the HCM decile distribution is slightly higher thar1 the TRAF-NETSIM default decile distribution. Follow ing a review of the gap acceptance study results shown in Figure 29, it was arbitranly decided to continue using the NETSIM default distribution in subsequent mode} building. 1 1800 1 1 1600- % 1400 - "a L 1200 _ i ~ 1 ~1000 ~ 800 I 1. | - - - - - Using default NETSIM median gap value I Using HCM median gap value 600 400 200 100 500 1000 Arterial Volume (vph) Figure 29. Elect of NETSIMdecile gap distribution for three-lane arterial. ~ 500 2000 The effect of the number of lanes on ramp crossing volumes is illustrated in Figure 30. The drop in the ramp crossing volume is sharper u ith an increase in the number of vehicles on the one- lane arterial because all the vehicles have to use the single lane so the number of acceptable gaps available is reduced. For the two and the three lane cases, the same number of vehicles are distributed over two or three lanes, as the case may be, and there is a lesser effect on the ramp crossing vehicles. The net increase in the vehicles per hour per lane for the one lane arterial case is largest and hence its ramp capacity is affected the most. 2 - 63

OCR for page 7
as E ! ~ _ O as . . 1 ~ Boo 1600 ~L ~. 1-lane ~2-lane 3-lane 1 ~ 400 _ 1200 1 , 000 'I Boo 1 600 1 Too 1 200 + O ! 100 500 1000 1500 2000 , _ _ = 500 1000 1500 Arte ria I Vo lu m e (vph ) Figure 30. Elect of number of lanes on maximum ramp volume. Observations ofthe simulation results of Figure 30 suggest~at an exponentialmode} would reasonably fit the interchange ramp capacity results generated by NETSIM. The values of ramp capacity were obtained by simulation of the desired conditions and the coefficients of the model were determined using SAS, a statistical analysis software package f49. Figure 3 1 shows how well We model fits the traffic simulation program values. The points indicate the average of ten simulation runs while Me lines indicate Me trend using the calibrated exponential model. 200 - 1' 1800 1600: ~ S/m 2/ane 1400 i. ~I ~ Sim. 3-/ane 1200 5` '\ &\ ~- Reg. 1-/ane 1 000 _ \ . . W Q 600 . ~ ` ` __ _ ^ 400 ~-- ~ - -_. ~ 100 500 1 000 1500 2000 ~ .__ ___ .___ .___ ____ Arterial Volume (vphJ 1l Figure 3 1. Comparisons of ramp capacityfor simulation and exponential regression mode! results. 2 - 64

OCR for page 7
Applying the exponential regression analysis in SAS, the average of the observed volume data for each case was used to estimate the Known coefficients cat and ,8 in the exponential equation. a multiplied by 3600,denotedas Tc, is the average critical gap time of the corresponding lane configuration, while ~ multiplied by 3600' denoted as Hi, is the minimum headway of the e e r_4 ' , ~ ~ _ my_ e artery weaving section. ~ ~ i, , ~ la '~e 11 s lows t :le coen~c~ents ax and ~ of the exponential model computed for one, two and three lane arterials. The coefficients in the proposed exponential equation are accurate estimations of the TRAF-NETSIM simulated operations in terms of standard errors and their variances. Table ~ ~ illustrates the coefficients of the mode! on a per lane basis. For the per large analysis, the results of the one, two and three lane cases were pooled and regressed. It cart be observed that the values of TC arid Hs are close to that of the one lane case. Table 17. Coefficients of the exponential regression mode' I,anes ~ -lane 2-lane 3-lane Coefficients a ~ ~a ~a ~ _ Exponential 0.00195 0.000657 0.001 I S 0.000574 0.00088 0.000565 R2 Value 0.9977 0.9995 0.9989 Conversion of ~Tc ~Tc ~ Its ~Tc ~ E Coefficients Values (sec.) 7.02 2.36 4.26 2.06 3.17 2.03 Table IS. Coefficients of the exponential regression mode! on a per lane basis Coefficients Exponential R2 Value Conversion of Values Coefficients (sec.) a ~ 0.002091 ~ 0.9970 ~ Tc ~ 7.52 ~ 0.000583 Hs 2.10 Progressed Flow. The NETSIM simulations were used to determine ramp crossing volumes for progressed arterial flow. Different progression factors were analyzed, ranging from PFs of 0.1 to Its. A PF value of I.0 is essentially uncoordinated, uniformly distributed flow. Progression factors from unity reflect the degree of platooning ofthe dominant flow. Volume-to-capacity ratios of 0.2, 0.4, 0.6, 0.7 and 0.8 on the upstream feeding movements were studied for a three-lane arterial. operating in two chases to create two platoons flowing downstream ~ _' In order to simulate various PF, vehicles were emitted from the upstream intersection ~ such that one platoon aIrives , ~ . . _ on red and the other platoon arrives on green. At the merge point, the notion of red and green only characterizes the degree of platooning in the arterial flow, as there is no signal at the merge point. 2 - 65

OCR for page 7
Figure 32 summarizes the experimental results. Polynomial regression equations for these plots were determined using SAS. Dunug low volume conditions on the arterial, little change occurred in ramp crossing capacity for different PFs. As the volume on the arterial increased, the ramp crossing volume decreased significantly due to fewer acceptable gaps for the weaving maneuver. For higher volumes, the change in the ramp crossing volumes for venous PFs becomes more significant. A PFof ~ .0 is considered random flow and the ramp crossing volume is the least for a PFof I.0 when compared to other PFs between O.1 to 1.~. The flow graph takes the shape of a parabola which has its minimum at a PF of 1 .0. Figure 32 clearly indicates the trend of ramp crossing volumes for v/c ratios of 0.2, 0.4, 0.6, 0.7 and 0.S, respectively. Me difference in ramp crossing volume for a PFof O.] to that of a PFof I.0 Increases with an increase in arsenal volume. In other words, We difference between the ramp crossing volumes for a PF of 0. ~ to that of a PF of ~ .0 increases with an increase in the v/c ratios on the arterial. Since the green ratio (green t~me/cycle length) is the same along both the approaches of the arsenal, Me PEG for a PF less than I.0 corresponds to the PER for a PF greater than ~ .0 and vice-versa. Plots of ramp crossing volumes for PF less than I.0 are a mirror image of plots for PF greater than ~ .0 about the axis of PF of ~ .0. 1200 1 ~ ~ 1100 > 1 ~ 1000 ._ tan en O0 900 ! ~ `~: 800 .~For\dc=04 ~For~c=0.6 ~For~c=0.7 ~For~c=0.8~ v m- ,^~%,, a., ~_7 ~m Ha_ _7x ~3< X ~ INK 700 1 ~, . , , , 1 o.o 0.2 0.4 0.6 0.8 1.0 1.2 1. ~1.6 1.8 1l PF I Figure 32. Elect of PF on ramp crossing volume for various v/c. Adjustment Factors for Progression. In order to further simplify the simulation results, the regression equations from the graphs for various v/c ratios were used to determine individual values of ramp crossing volume. A PFof I.0, also considered es random flow, was used as the basis for development ofthe adjustment factors. The factors for over PF were computed by determining the ratio of the value at PF of ~ .0 to that of another PF. Because the coves were parabolic and the values on one side of the curve were mirror images of the over, adjustment factors for PF from 0. 2 - 66

OCR for page 7
to 1.0 were computed. Figure 33 presents adjustment factors for various PFrarging from 0.1 to 1.0 for volume-to-capacity (v/c) ratios of 0.4, 0.6, 0.7 and 0.~. 1.25 1. 1= JO 1.05 L. ~ 1.00 1.10 0.95 1 20 - 1 r I -~-v/c= 0.4 ___. vlc= 0.6 _____v /c = 0.7 vlc=0.8 1. i_ 0.90 , , , I 0.1 0.3 o.s 0.7 0.9 PF Figure 33. Capacity adjustment factors for various progression factors. 1 l ll Table 19 shows the actual (average) capacity adjustment factors simulated for various progression factors ranging from 0.1 to 1.0. Table 20 provides related capacity adjustment factors obtained using the exponential equation (Equation 31) shown below as developed from Table 19 average results using SAS. Note that average arterial lane volumes V are used in Equation 3 1 to provide a more convenient data input format. fPF = 1 + ,0l5*eLo-oo44*v-3os*pFl where: ~_ JPF PF ramp weaving capacity adjustment factor; progression factor, and V = arsenal volume per hour per lane (vphpl). (32) From a comparison ofTables 19 and 20, it can be seen that the above exponential equation follows a close fit of the actual average adjustment factors. The sum of squares error (SSE) was determined to be 0.00217. 2 - 67

OCR for page 7
2.7.3 Determination of Ramp Capacity during Random Flow The ramp capacity during random flow, denoted by OR can be determined by inputting the total arterial volume in Equation 3 1. The computed coefficients a, ,8 shown in Table 1 7 have been computed on the basis of number of lanes in the arterial section and can be used in the exponential equation. The results of one, two and three lane arterials were pooled and regressed to compute the common coefficients for single and multilane arterials. Depending on the decree of accuracy required oy tne user, me a~rerent coe~nc~ents could ne used for predicting ramp capacity. Table 19. Actual adjustment factors for PF of 0.1 to 1.0 Progression Factors, PF v/c 0~4 0~6 0~7 0~8 0~1 1 046 1098 1150 1205 0~2 1 0~3 1 0~4 1 0~5 1 0~6 1 0~7 1 0~8 1 036 T 1~028 11.020 11~014 11 009 T 1 005 11 002 077 T 1 059 11 043 11~030 11 009 T 1 011 11~005 1.118 T 1 090 11.066 11046 11 019 T 1 017 1 1 007 1.162 T 1 124 11.091 11063 11 030 T 1 023 1 1.010 0~9 1 1~0 1 001 1.001 1 002 1.000 1.000 1.000 1.003 1 eOOO Table 20. Computed adjustment factors for PF of 0.1 to 1.0 Progression Factors, PF v/c 0~4 0~6 0~7 0~8 0~1 1 048 1.100 .142 1208 0~2 1 0~3 1 0~4 1 0~5 1 0~6 1 0~7 1 0~8 1 035 1 026 1 019 1 014 1~010 1 008 1 006 1 074 1 054 1 040 1029 1 021 1 016 1 012 . 1~105 1 077 1 057 1042 1.031 1 023 1 017 . 1153 1 113 1 083 1061 1 045 1 033 1 025 0~9 1 1~0 1 004 1 009 1 012 1 018 1 003 1 006 1 009 1 013 2~7~4 Adjustment for Sneakers Comparison ofthe simulation results between random flow and progressed flow at a PF of 1.0 revealed that the ramp crossing volume for a PF of 1.0 was higher. The difference between the ramp crossing volumes between random and progressed flow increased with an increase in arsenal volumes. This difference in ramp crossing volumes was attributed to the sneakers crossing during the two phase change intervals i.e., sneakers (S,9. In over words, the ramp vehicles completed the weaving maneuver by making use of the large gap available to the ramp vehicles during the two phase change intervals of four seconds each at the upstream intersection. The random flow conditions had a situation wherein the upstream intersection had 100 percent green on the arsenal movement and hence the effect of sneakers was not observed. On the average, approximately three vehicles were completing the weaving maneuver during each phase change interval. The effect of 2 - 68

OCR for page 7
sneakers was confirmed visually by observing the animation of the simulation for the required conditions in GTRAF. Thus, the ramp crossing capacity, adjusted for sneakers, would be QR Qua Sm where: Q R QR sm ramp capacity adjusted for sneakers (vph); ramp capacity for random flow (vph); and sneaker volume (vph). 2.7.5 Application ofAdjusiment Factors (33) In order to obtain the ramp capacity for different progression factors, the adjustment factors for progression,fpF, needs to be multiplied to the ramp capacity which has been adjusted for sneakers as follows: = where: QPF Q R fPF QPF QR fPF ramp capacity adjusted for progression (vph); ramp capacity for random flow (vph); and adjustment factor for progression. (34) The application methodology of this formulation of arterial weaving wall be presented in Chapter 3. Moreover, field studies of arterial weaving operations were conducted and are described in detail in Appendix E (129. Several empirical models of maneuver speeds and delays as related to local conditions are provided. These studies were extremely tedious and time consuming. Other initial NETSIM simulation studies of arteriaVramp weaving operations were conducted end reported (139. All of these studies showed the benefit of increased signal separation between the interchange and the next downstream signal together with the benefit of arterial signal coordination during undersaturated conditions. 2 - 69

OCR for page 7