Citation
An operational comparison of high-volume single point urban and tight urban diamond interchanges using transyt-7f and netsim

Material Information

Title:
An operational comparison of high-volume single point urban and tight urban diamond interchanges using transyt-7f and netsim
Creator:
Rens, Amy J
Publication Date:
Language:
English
Physical Description:
xi, 157 leaves : illustrations ; 28 cm

Subjects

Subjects / Keywords:
Roads -- Interchanges and intersections -- Mathematical models ( lcsh )
Roads -- Interchanges and intersections -- Mathematical models ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 155-157).
General Note:
Department of Civil Engineering
Statement of Responsibility:
by Amy J. Rens, P.E.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
44093806 ( OCLC )
ocm44093806
Classification:
LD1190.E53 1999m .R46 ( lcc )

Full Text
AN OPERATIONAL COMPARISON OF HIGH-VOLUME SINGLE POINT
URBAN AND TIGHT URBAN DIAMOND INTERCHANGES USING
TRANSYT-7F AND NETSIM
Amy J. Rens, P.E.
B.S., Iowa State University, 1988
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Science
Civil Engineering
1999
by


This thesis for the Master of Science
degree by
Amy J. Rens
has been approved
by
Bruce N. Janson
Sarosh I. Khan
Date


Rens, Amy J. (M.S., Civil Engineering)
An Operational Comparison of High-Volume Single Point Urban and Tight
Urban Diamond Interchanges Using TRANSYT-7F and NETSIM
Thesis directed by Associate Professor Bruce N. Janson
ABSTRACT
There has been an ongoing debate about whether a single point
urban interchange or a tight urban diamond interchange operates better at
high volumes. This thesis examines the operations of these two
interchange types and determines which one works better. Much of the
previous analysis was done without the advantage of actual measured field
data. This study incorporates field data for saturation flow rates and
clearance intervals and examines seventeen volume scenarios in both
TRANSYT-7F and NETSIM. The study found that the single point urban
interchange had a lower overall intersection average delay in all cases.
However, the overall intersection average delay is not significantly different
between the single point urban interchange and the tight urban diamond
interchange. The average delays for all movements are presented.
This abstract accurately represents the content of the candidates thesis,
recommend its publication.
Signei
ruce N. Janson


DEDICATION
To our unborn child, who accompanied me on this journey. May you find
the pursuit of knowledge as rewarding as your father and I do.


ACKNOWLEDGEMENTS
I would like to thank the following people for their support as I complete my
graduate degree:
Bruce Janson, for agreeing to be my major professor. You have offered
guidance and a willingness to work with my compressed schedule to
complete my degree.
Sarsoh Khan and Lynn Johnson, for being on my committee. I appreciate
the amount of effort it takes to be on a thesis committee. Thank you also
the classes Ive taken from you.
Carter & Burgess, for being a company that still thinks continued education
is valuable. And to Dave Stevenson and Marvinetta Hartwig, for approving
my requests for payment of these classes and for supporting my efforts.
Kevin, my husband, best friend, promoter, and critic; for being concerned
about my education; for pushing me when I needed it; for proofreading my
papers and thesis; for showing me the ropes of getting a thesis done; and
for taking on more of the household duties so that I could focus on
completing my thesis. Completing my masters degree fulfills one of the
goals we had when we moved to Colorado.
God, for providing me with the ability and strength to complete my masters
degree. May You be glorified in all that I do.


CONTENTS
Tables................................................................ix
Figures................................................................x
Chapter
1. Introduction.......................................................1
1.1 Key Variables and Definitions......................................2
1.2 Literature Review..................................................4
1.2.1 Introduction of the SPUI.........................................4
1.2.2 Operational Comparison Studies between the SPUI and the TUDI.....5
1.2.3 Operational Parameters of the SPUI and the TUDI.................13
1.3 SPUI/TUDI Interchange Selection Criteria........................16
1.4 Traffic Modeling Software.........................................19
1.5 Operational Issues at Ramp Terminals..............................22
1.6 Operations of a SPUI..............................................23
1.7 Operations of a TUDI..............................................26
2. Methodology.......................................................29
2.1 Study Approach....................................................29
2.2 Traffic Volume Scenarios.........................................30
2.3 Lane Geometry....................................................30
VI


2.4 TRANSYT-7F Modeling...............................................33
2.5 NETSIM Modeling..................................................34
2.6 Assumptions......................................................37
2.6.1 Saturation Flow Rates...........................................37
2.6.2 Clearance Intervals.............................................38
2.6.3 Start Up Lost Time..............................................38
2.6.4 Network Speed/Free Flow Speed...................................39
2.6.5 Default Parameters in NETSIM....................................39
3. Results............................................................41
3.1 Optimal Signal Timing.............................................41
3.2 Calculation of Delay.............................................41
3.3 TRANSYT-7F Results................................................46
3.3.1 Overall Intersection Delay......................................46
3.3.2 Delay by Individual Movements...................................47
3.3.3 Cases with Spillback/Oversaturation.............................51
3.3.4 Comparison with Leisch and Fowler...............................54
3.4 NETSIM Results..................................................56
3.4.1 Overall Intersection Delay......................................57
3.4.2 Delay by Individual Movements...................................57
3.4.3 Cases with Spillback............................................69
3.4.4 TRAFVU Observations.............................................70
vii


3.5 Volume Limitations in NETSIM and TRANSYT
74
4. Conclusions...............................................78
Appendix
A. Sample Output File for SPUI TRANSYT-7F....................82
B. Sample Output File forTUDI TRANSYT-7F....................88
C. Sample Output File for SPUI NETSIM........................95
D. Sample Output File for TUDI NETSIM.......................110
E. Summary of Output Data...................................127
F. Sample Calculation of t test.............................153
References..................................................155
viii


TABLES
1.1 Comparison between SPUI and TUDI.............................17
2.1 Traffic Volume Scenarios...................................31
2.2 Default NETSIM Values......................................40
3.1 Optimum Signal Timing for SPUI from TRANSYT-7F.............42
3.2 Optimum Signal Timing for TUDI from TRANSYT-7F.............43
3.3 Overall Intersection Statistics Comparison from TRANSYT-7F...47
3.4 TRANSYT-7F: Average Delay by Movement........................49
3.5 Summary of Spillback Cases from TRANSYT-7F...................53
3.6 Comparison to Leisch Results: Cycle Length and Overall Delay.55
3.7 Comparison to Leisch Results: Saturation Flow Rates..........55
3.8 Overall Intersection Delay Comparison from NETSIM............58
3.9 NETSIM: Average Delay by Movement............................60
3.10 Summary of Spillback Cases from NETSIM......................69
IX


FIGURES
1.1 Single Point Urban Interchange...................................24
1.2 Signal Phasing: Single Point Urban Interchange...................25
1.3 Tight Urban Diamond Interchange..................................27
1.4 Signal Phasing: Tight Urban Diamond Interchange..................28
2.1 NETSIM Network: Layout of SPUI...................................35
2.2 NETSIM Network: Layout of TUDI...................................36
3.1 TUDI Movements that Travel through Both Intersections............45
3.2 Overall Intersection Delay in TRANSYT-7F.........................48
3.3 Volume to Capacity Ratio Comparison in TRANSYT-7F...............52
3.4 Overall Intersection Delay in NETSIM.............................59
3.5 Average Delay Comparison of Westbound Left Turns...............63
3.6 Average Delay Comparison of Eastbound Left Turns...............64
3.7 Average Delay Comparison of Northbound Left Turns..............65
3.8 Average Delay Comparison of Southbound Left Turns..............66
3.9 Average Delay Comparison of Westbound Through Movement.........67
3.10 Average Delay Comparison of Eastbound Through Movement.........68
3.11 Typical Spillback for the SPUI.................................71
3.12 Typical Spillback for the TUDI.................................72
x


3.13 Average Delay vs. Entering Volume-NETSIM.........75
3.14 Average Delay vs. V/C RatioTRANSYT..............77
XI


1.0 Introduction
The objective of this study is to determine whether a single point
urban interchange (SPUI) or a tight urban diamond interchange (TUDI) works
better operationally by modeling a number of high-volume traffic scenarios in
TRANSYT-7F and NETSIM. Since both of these interchanges will fit within
the right-of-way constraints of the urban area, the two interchange types are
often compared to each other. While other factors (e.g., cost, impact on the
surrounding street network, and pedestrian requirements) contribute to the
final selection, the question of which interchange type works better
operationally is a major factor in choosing the best interchange type for
particular conditions. This study compares only the traffic operations
between the SPUI and TUDI.
Previous studies have compared the operational efficiencies of these
two interchange types, but conflicting results have been recorded. This study
uses the results of two field studies that compared the saturation flow rate
and clearance intervals between the two interchange types. These field
studies were conducted after these previous studies. In light of this new
information, a comparison of the operational efficiency between the SPUI
and the TUDI is appropriate.
1


This chapter includes the following items:
1) a summary of key variables and definitions in signalized intersection
operations,
2) a summary of previous studies and articles on the operations of the
SPUI and theTUDI,
3) a discussion of all of the factors involved in selecting a SPUI orTUDI,
4) a summary of traffic modeling issues,
5) an overview of operational issues at ramp terminals,
6) an introduction to the operations of a single point urban interchange,
and
7) an introduction to the operations of a tight diamond urban
interchange.
1.1 Key Variables and Definitions
The following key variables and definitions are essential in
understanding traffic operations at signalized intersections. Ramp terminals
in high volume urban areas are usually signalized.
Saturation flow rate is the number of vehicles that could enter the
intersection in a single lane if the signal were always green for that lane, and
vehicles never stopped (McShane and Ross, 1990, p. 58). Saturation flow
rate is measured in vehicles per hour green per lane (vphgpl).
2


Discharge headway is the time or headway between vehicles moving
in a lane. A constant headway is achieved in a stable stream of vehicles is
called the saturation headway. Headway is measured in seconds/vehicle.
The relationship between the saturation flow rate and saturation headway is
defined as follows:
S = 3600/h (1.1)
where S = saturation flow rate (vphgpl)
h = saturation headway (sec/veh)
(McShane and Ross, 1990, p.58)
Start up lost time is the accumulation of time at the beginning a green
interval when the first few vehicles take more time than the saturation
headway. Usually, the saturation headway is achieved by the sixth or
seventh vehicle in the queue (McShane and Ross, 1990, p.58).
Clearance lost time is the time between the last vehicle from one
approach entering the intersection and the start of the green signal for the
opposing movement (McShane and Ross, 1990, p.58).
Clearance interval is the amount of time allocated to the yellow and
all-red signal interval to allow for clearing the intersection before the next
movement enters the intersection (McShane and Ross, 1990, p.59).
Capacity is defined as the actual number of vehicles that are serviced
by an lane at a signalized intersection. The capacity is based on the
3


saturation flow rate, start up and clearance lost times. It is measured in
vehicles per hour per lane (vphpl) (McShane and Ross, 1990, p.59).
Spillback is defined as extensive queuing that occurs during
congested times and blocks the previous intersection. The demand on the
intersection approach exceeds the capacity of the approach.
1.2 Literature Review
This literature review examines the introduction of the SPUI, the
operational comparison studies between the SPUI and TUDI, and the studies
on the operational parameters of the SPUI and TUDI.
1.2.1 Introduction of the SPUI
Greiner Engineering (n.d. circa1981) introduced The Urban
Interchange in a 60-page glossy book that included layouts, photos, and a
simplified capacity analysis promoting the SPUI as the superior interchange
for locations of limited right-of-way in urban areas. The capacity comparison
related a theoretical urban interchange (SPUI) to the measured capacity at an
existing diamond interchange. The theoretical capacity was based on an
assumed capacity per cycle length multiplied by the number of cycles per
hour. The results showed that the SPUI more than doubled the capacity of a
4


diamond interchange. This comparison appears to be the basis of later claims
that the SPUI is capable of handling more traffic than the TUDI.
In 1985, the SPUI was included in an article in the ITE Journal
discussing innovations in diamond interchanges (Leisch, 1985). By the late
1980s, the single point urban interchange was being promoted in industry
trade journals as the way to move more traffic in less space. Hawkes and
Falini (1987) stated that the SPUI could accommodate twice the maximum
volume as the conventional diamond interchange in most cases. Urban
Interchanges Moves More Traffic in Same Space (1989) states that the SPUI
can handle 6500 to 7100 vehicles in the peak hour. These statements are
unsubstantiated in both articles.
1.2.2 Operational Comparison Studies between
the SPUI and TUDI
Many studies have compared the operating characteristics of the
SPUI and the TUDI. However, the two most popular are the Leisch, Urbanik
and Oxley (1989) study and the Fowler (1993) study.
1.2.2.1 Leisch, Urbanik, and Oxley Study
Leisch et al. (1989) looked at 5 existing interchanges and their traffic
volumes to determine whether a SPUI or a CDI (compressed diamond
5


interchange) would be more appropriate. The study gave traffic volumes,
lane configurations, and optimal cycle lengths. It modeled these cases in
TRANSYT-7F and compared both total system delay and average
intersection delay.
For saturation flow rates, it uses 1800 vphgpl for the through traffic,
1700 vphgpl for the SPUI left turns, and 1600 vphgpl for CDI left turns and all
right turn movements. The study states that off-ramp lefts, when in 2 lanes
and opposed by left turns (like the SPUI), exhibit a reduction in saturation
flow rates by 5% to 15%. No support is given for this statement, but it is the
rationale for selecting the 1700 vphgpl saturation flow rate. This reduces the
saturation flow rate by 5.5%.
For the SPUI, it uses clearance intervals (yellow plus all-red time on
the signal) of 7.5 seconds for through movements and 8 seconds for the left-
turn movements. For the CDI, it uses 5.5 seconds for the through
movements and 6 seconds for the left-turn movements. These values are
based on the ITE Manual of Traffic Signal Design (1982) which applies the
dilemma zone using intersection width and approach speed to determine
adequate clearance times.
The signal phasing for the SPUI was a three-phase plan with no
overlap. The phasing for the CDI was a four-phase plan with overlap on two
phases.
6


The conclusion reached was that the CDI had less delay for four
locations. At the fifth location, the SPUI had less delay. This location had
heavy balanced left turns from the off ramps and very light through traffic.
The study also looked at one situation with frontage roads to
demonstrate the higher delay at the SPUI when a fourth signal phase is
added for the frontage roads.
1.2.2.2 Fowler Study
Fowler (1993) looked at twelve fabricated traffic volume scenarios to
compare the SPUI and the TUDI. These scenarios varied the directional split
of the cross street through movements, the volume of the cross street left
turns, and the balance of the off-ramp left turns. The paper gives traffic
volumes but no lane configurations or optimal cycle lengths.
The first part of the study developed a spreadsheet to calculate the
volume-to- capacity ratio based on traffic volumes, saturation flow rates,
cycle lengths, lost times, and the TUDI capacity overlap in the signal timing.
These scenarios were then modeled in TRANSYT-7F, and the volume-to-
capacity ratios were compared. Then the scenarios were modeled in
TRANSYT with 80% of the traffic volumes, and both v/c ratios and delay
were compared.
7


For saturation flow rates, the study used 1800 vphgpl for all
movements in both the SPUI and TUDI. For right turns at the SPUI, 2700
vphgpl was used. For clearance intervals, it used 6 seconds per phase for
the SPUI and 3 seconds per phase for the TUDI. The different assumptions
in the saturation flow rates and the clearance intervals could explain the
different results between this study and the Leisch et al. (1989) study.
In all cases, the SPUI had lower delay. The volume-to-capacity ratio
was lower for the SPUI in eleven out of the twelve cases.
1.2.2.3 Garber and Smith Study
Garber and Smith (1996) conducted a nationwide survey of the state
highway agencies regarding the prevalence of and attitudes toward the
SPUI. The state highway agencies' perceptions were that the SPUI has
higher capacity, requires less ROW, and accommodates large trucks better
than the TUDI. The study also examined the operational and safety
differences between the diamond and the SPUI and developed guidelines for
when to select a SPUI or TUDI.
The operational analysis consisted of three parts. First, it examined
six existing SPUIs and five existing diamonds for the delay at each
interchange. Second, it compared the each of the five diamond interchanges
with a theoretical SPUI using the same traffic characteristics. In this
8


analysis, the existing signal timing was used for the diamond, but an
optimized signal timing was used for the SPUIs.
The study acknowledged the inconsistency of this analysis and
developed a third analysis. In this analysis, the diamond and SPUI were
compared using 10 high volume and 10 low volume scenarios. The total
entering volume of the low volume scenarios ranged from 2700 to 2970
vehicles. The total entering volume of the high volume scenarios ranged
from 5700 to 5950 vehicles. In these scenarios, the through volumes and left
turn volumes were balanced and unbalanced in a number of ways to
determine which cases the SPUI or diamond worked better. The diamond
had a lower delay in the low volume scenarios and the SPUI had a lower
delay in the high volume scenarios. In the high volume scenarios, there were
two scenarios where the SPUI had a higher delaywhere higher through
volumes opposed the higher left-turn volumes.
This study report is sketchy in the details provided. No assumptions
about the NETSIM modeling are given and no statistical analysis for the third
analysis was given. Statistical tests were performed on the first two
comparisons and no significant differences were found between the delay at
the diamond and SPUI. The study gives some good guidelines about when to
use the SPUI and TUDI, but gives no clear conclusion about which one
operates better.
9


1.2.2.4 Pate and Stover Study
Pate and Stover (1992) examined guidelines for urban arterial
interchanges. The study specifically looks at the SPUI, TUDI, and the Left-
Hand Exit Single Signal. It does a thorough comparison of these three
interchanges based on cost, operations, and ROW. An analysis was done in
TRANSYT-7F for four different cases:
1) Low Volume, Low Turning Movements;
2) Low Volume, High Turning Movements;
3) High Volume and Low Turning Movements; and
4) High Volume, High Turning Movements.
The delay for all these cases was lowest for the SPUI. No statistical analysis
was done to determine if the differences were significant. The study states
that the performance of the SPUI decreases when the volumes of the
simultaneous left turns vary greatly. It uses the results of the Leisch et al.
(1989) study to support this claim. Based on this analysis, the report claims
the TUDI is the best option for the urban setting.
1.2.2.5 Gupta Study
Gupta (1993) compares the SPUI and TUDI using a microsimulation
software developed specifically in this study. The study claims that the
existing traffic modeling software is inadequate to analyze the traffic
10


operations at a SPUI. The study shows that the SPUI operates better than
the TUDI in all eight scenarios. The total entering volumes for those
scenarios range from 2800 to 7700 vehicles per hour. The study also
chooses to analyze delay in average delay per distance (sec/veh/mile). This
accommodates for the fact that the vehicles must travel a longer distance
through the TUDI.
1.2.2.6 Yousif Study
Yousif (1990) looks at optimizing one SPUI with an actuated signal.
The study looks at all the independent variables of an actuated signal that
contribute to delay and creates a linear regression for the delay model. The
study does a sensitivity analysis of all the independent variables. The
sensitivity analysis for saturation flow rate was surprising. There was very
little variation in the delay output from the sensitivity analysis for saturation
flow rate/ discharge headway. Given that the saturation flow rate varied from
1.7 to 2.6 seconds., a larger variation in the delay would be expected. The
delay varied from 14.50 to 19.64 seconds/vehicle. The sensitivity analysis
for change interval showed a wider range of delay. The delay varied from
8.34 to 22.80 seconds/vehicle for the clearance intervals ranging from 1 to 10
seconds.
11


This study also looked at comparing the SPUI to a CDI. The analysis
showed that the SPUI was better. However, the interchange geometries (like
left turn bay lengths) were not comparable.
1.2.2.7 Abbey Study
Abbey (1991) did a case study that compared two existing
interchanges: one SPUI and one TUDI, in Salt Lake City. The study used the
AASHTO Red Book method to calculate delay cost. The delay cost
calculation is based three components: 1) slowing down, stopping, and
starting up at traffic signals, 2) idling in queues, and 3) the time it takes to get
through an intersection. (Abbey, p. 34)
The SPUI had an overall lower delay cost than the TUDI. The
substandard left-turn storage at the TUDI and overall higher traffic volumes
at the TUDI may have contributed to these results.
1.2.2.8 Bonneson Study
A study done by Bonneson (1992) compares the SPUI to a high-type
at grade intersection (AGI). Since the SPUI can fit in the same space as an
AGI, it is viewed as an alternative to a highly congested AGI. In this study,
three SPUIs and two AGIs were studied. Four variables, discharge
headway, startup lost time, free flow speed and end use were quantified and
12


an ANOVA and linear regression were performed on these dependent
variables.
The paper also gives a general operational comparison of the two
intersection types in a qualitative form. The AGI has a higher phase capacity
but the SPUI has an lower overall delay and a lower demand-to-capacity
ratio. This is attributed to the fact that the major movement in the SPUI is
separated out from the intersection.
For the left-turn movement, the capacity decreases as the wheel path
increases. This is partly due to the longer clearance interval. However, the
larger radius encourages higher speeds and lower headways. In this study,
these two seem to negate each other.
1.2.3 Operational Parameters of the SPUI and TUDI
Poppe, Radwan, and Matthias (1991) look specifically at some of the
key operational parameters of the SPUIthe saturation flow rate of left and
through movements and the clearance lost time per phase. Three SPUIs in
the Phoenix area were studied, and the saturation flow rate, start up lost
time, and clearance lost time were measured. The off-ramp left turn, cross-
street left turn, and cross-street through movements have mean saturation
flow rates of 2069, 2045, and 1935 vehicles per hour green per lane (vphgpl),
respectively. The mean start up lost times for the three movements were
13


1.84, 1.43, and 1.55 seconds, respectively, and the mean clearance lost
times were 6.67, 3.97, and 6.41 seconds, respectively.
Hook and Upchurch did a follow up to this study in 1992, where they
looked at these same parameters for the TUDI and performed a statistical
analysis on the comparison. The off-ramp left turn, cross-street left turn, and
cross-street through movements have mean saturation flow rates of 1915,
1935, and 1895 vphgpl, respectively. The mean start up lost times for the
three movements were 1.49, 1.39, and 1.58 seconds, respectively, and the
mean clearance lost times were 3.52, 3.87, 3.80 seconds, respectively. The
comparison found that the mean saturation flow rate for the off-ramp left
turns was significantly different between the SPUI and TUDI. The mean
clearance lost times for both the off-ramp left turns and the cross-street
throughs were significantly different. All the other parameters were not
significantly different.
Poppe et al. (1991) reveals the basis for the Leisch et al. (1989)
statement that dual left- turns have a lower saturation flow rate per lane. The
basis is a 1961 study by Capelle and Pinnell where they identified that
drivers tend to stagger as they go through a dual left turn. This rationale is
the basis of the Highway Capacity Manual lane utilization factor at dual-lane
left turns. In the Poppe et al. (1991) study the Hook and Upchurch (1992)
study, and the Sheffer and Janson (1997) study, the saturation flow rates for
14


dual left-turns were in the range of 1900-2100 vphgpl. These studies
indicate that todays drivers may be more comfortable with dual left-turns
than drivers were in 1961.
Bonneson (1992) investigated the discharge headways and lost times
at three SPUIs and two at-grade intersections (AGI). The study found that the
off-ramp left turn, cross-street left turn, and cross-street through movements
have discharge headways of 1.86, 1.88, and 2.07 seconds, respectively.
The start up lost times for the three movements were 2.52, 2.50, and 2.09
seconds, respectively. The SPUI left turn discharge headways were lower
than for the cross-street through movements. This finding is consistent with
the Poppe et al. study.
Bonneson and Messer (1998) investigated the effect that traffic
pressure has on saturation flow rates. Traffic pressure is defined as the
lowering of the headways for each lane when there are more vehicles
queued in that lane. Traffic pressure is expressed in terms of volume per
cycle per lane. Twelve interchanges with closely spaced ramp terminals
were studied and a model for saturation flow rates and start up lost time on
the cross-street through movements was developed. Discharge headways
for the through movements at the SPUI and TUDI were 1.86 and 1.67
seconds per vehicle, respectively. The 1.67 seconds/vehicle for the TUDI
represents only one interchange which had a large down grade. Start up lost
15


times were 2.52 and 2.88 seconds, respectively. The study recommends
using a cross-street through movement saturation flow rate of 2000 vphgpl.
Dixon (1998) calibrated and validated a NETSIM model of a SPUI
based on an existing SPUI interchange at 1-215 and California Avenue in Salt
Lake City. The actual flow rates ranged from 165 vph on the westbound
through movement to 666 vph on the westbound left turn. With these
volumes, this interchange would be considered a low volume interchange.
For the discharge headway, Dixon used 1.9 seconds per vehicle for all the
cross street movements and 2.0 seconds per vehicle for the off-ramp
movements. This finding contradicts the findings of Poppe et al. (1991) and
Bonneson (1992). The start lost times were 1.7 seconds for the cross street
movements and 1.9 second for the off-ramp movements. This finding agrees
with the Poppe et al. (1991) results.
1.3 SPUI/TUDI Interchange Selection Criteria
Many factors contribute to which interchange type is appropriate for a
given location. Pate and Stover (1992) provides a table that compares the
TUDI and the SPUI. This information can be found in Table 1.1.
16


Pate and Stover (1992) also states a TUDI advantage: When an
accident occurs on the mainline between the ramp gores, traffic can be
diverted off the mainline and through the ramp terminals using the off and on-
ramps (p. 59).
Characteristic TUDI SPUI
Right-of-way Requirements Moderate Moderate
Construction Costs Moderate High
Sight Distance Requirements Low Moderate
Length of Vertical Curves Low Moderate
Driver Expectancy Meets Slightly Violates
Accommodation of Pedestrians Good Poor
Accommodation of Heavy Vehicles Poor Good
Table 1.1 Comparison Between SPUI and TUDI (Pate & Stover, 1992, p. 61)
17


Leisch et al. (1989) recognized that any efficiency with the SPUI is lost
when frontage roads are present. The additional signal phase and lost time
add up.
The SPUI does not work well when the skew angles are greater than
30 degrees. Bridge structures get too long and sight distance can be a
problem.
Abbey (1991) states that the TUDI works better than the SPUI when
there is a large through volume and relatively little turning volume. He also
questions the occurrence of this volume scenario given that interchanges
facilitate traffic getting off and on the mainline. If there is little turning volume,
there would be no need for a interchange and an overpass would be
adequate (p. 39).
Garber and Smith (1996) states the SPUI is more efficient when
handling left-turn movements and the Dl is more efficient for the arterial
through movement (p.51). This finding was based on questionnaire results
not on any findings from the operational analysis. This statement is a
perception that the state highway agencies have about the SPUI and TUDI.
Bonneson (1992) found that for the left-turn movement, the capacity
decreases as the wheel path increases. This is partly due to the longer
clearance interval. However, the larger radius encourages higher speeds
and lower headways. These two factors appear to negate each other.
18


1.4 Traffic Modeling Software
This section looks at traffic modeling software that are appropriate to
evaluate the intersections at ramp terminals and issues associated with
these software. A number of studies have been done to examine the
capabilities and limitations of various software.
Transportation Research Circular No. 430, entitled Interchange
Operations on the Local Street Side: State of the Art, (Chen et al, 1994)
contains a number of articles on interchange characteristics and software
modeling issues associated with interchange capacity analysis. A
questionnaire was sent to members of the Transportation Research Board
Freeway Operations Committee asking which modeling software they had
used to evaluate interchange operations. TRANSYT-7F was the most
common response with fifteen responses. The Highway Capacity Software
and the PASSER II and III were next most frequently used. TRAF-NETSIM
was fifth with six responses. Another survey was sent to state highway
agencies asking which software they used most often in interchange
evaluation. The Highway Capacity Software was the number one response
(Chen, et al., 1994).
The study identified that TRANSYT-7F is useful for interchange type
selection, intersection spacing, signal timing optimization, queue analysis
and spillback analysis. TRAF-NETSIM is useful for interchange type
19


selection, intersection spacing, queue analysis, spillback analysis, and
weaving on cross roads. (Chen, et al, 1994).
The National Cooperative Highway Research Program Web
Document 12 entitled Capacity Analysis of Interchange Ramp Terminals,
(Messer and Bonneson, 1998) discusses traffic modeling software and its
capabilities and limitations. The purpose of this research project was to
evaluate the capabilities and limitations of existing software, and to develop
new software that encompasses the problems encountered at ramp
terminals. After evaluation of the software, it ranked the software in the
following order: NETSIM, TEXAS, and TRANSYT-7F. NETSIM was
selected because it can simulate almost all aspects of interchange/arterial
traffic operations desired. Its capability to view the simulation process gives
the analyst an added sense of the fidelity of the simulation in progress
(Messer and Bonneson, 1998, p.2-13).
This research also takes into account the effect of adjacent
intersections on traffic flow through ramp terminals at arterial streets and the
weaving between ramp terminals and adjacent intersections (Messer and
Bonneson, 1998).
McGhee and Arnold, Jr., (1997) recommend using NETSIM for
isolated intersections where congested, oversaturated conditions exist and
for nonisolated intersections where queuing and spillback occur. These
20


recommendations were made in a study establishing capacity analysis
procedures for the Virginia Department of Transportation.
Gupta (1993) states that NETSIM is not appropriate for analyzing the
SPUI because it does not take into account the size of the intersection or the
large turning radii that the off-ramp left turns can achieve. The model
estimates the total distance a vehicle has to travel, including intersection
length, and uses this distance to calculate the free flow travel time. Because
the model does not recognize the large intersection length, the total distance,
and hence, the free flow travel time is underestimated, and this means that
the total delay is overestimated (p.15).
Gupta (1993) also states that NETSIM assumes a constant speed of
22 feet/second for all left turn vehicles. But the left turning vehicles at SPUI
travel at speeds comparable to speeds of through vehicles. If the average
speed is 20 mph, it corresponds to 44 feet/second which is twice as much as
the speed assumed by the NETSIM model. This may result in
overestimation of delay values for left turns (p.15).
In a study that calibrated NETSIM for a SPUI, Dixon (1998) notes two
drawbacks of NETSIM: 1) the lack of movement specific input variables, and
2) the limitation of not be able to extend left and right turn bays past an
upstream node. Dixon commented that the program could be more accurate
21


if one could designate different headways and start up lost times for each
movement on an approach.
Jim Powell, Session Moderator of the Transportation Research
Circular No. 430 (1994), notes that the default parameters and statistical
distributions in NETSIM were based on the model calibration done in 1973.
These default values are based on drivers in Washington, D.C. He states
that these values are typically good for peak hour conditions. He also states
that few users of NETSIM have the ability to modify embedded stochastic
distributions due to the cost and amount of work involved.
Chen, et al.(1994) states that in TRANSYT-7F, the queues are vertical
and the program does not model spillback from left turn bays that could
impact the saturation flow rate of the through lanes.
1.5 Operational Issues at Ramp Terminals
There are many issues that contribute to the efficiency of the
intersections of a ramp terminal. The geometry of the intersections should
be adequate to handle the volume of traffic. Left turn bays should be long
enough to prevent left turn vehicles from blocking through lanes. Right turn
lanes on the off-ramps also help separate traffic and keep the ramps from
backing up. In the cases of the SPUI and TUDI, the geometry of the ramp
22


terminal intersections is less than desired because the surrounding street
network is already existing.
The impact of the ramp terminals on the arterial street is also
important. When adjacent intersections are too close to the ramp terminal
intersections, spillback may occur at any of the intersections. The weaving
on the arterial will be difficult as vehicles need to cross a number of lanes to
make a left or right turn. The spacing of intersections may also make signal
coordination and progression difficult. The optimal signal timing for the ramp
terminal intersections may not be the optimal signal timing for the rest of the
arterial.
1.6 Operations of a Single Point Urban Interchange
The single point urban interchange, shown in Figure 1.1, is a recent
development in interchange design. The four ramp terminals come together
in a single intersection under or over the mainline highway and the free right
turns do not cross the intersection. Due to the geometry, the intersection is
often 150-300 ft of open pavement. The angle that the off-ramp left turns
approach the intersection produces a large left-turning radius. Because of
the larger intersection, a higher clearance interval is needed. The advantage
is that there is only one intersection to negotiate.
23


Figure 1.1 si"9'e
point Urban Interchange
24


The signal phasing for the single point urban interchange, shown in
Figure 1.2, is a three-phase signal plan. The cross-street through and right
movements go in one phase, the off-ramp left turns go in the next phase, and
the cross-street left turns and off-ramp right turns go in the third phase.
Figure 1.2 Signal Timing: Single Point Urban Interchange
25


1.7 Operations of a Tight Urban
Diamond Interchange
The tight urban diamond interchange is defined as a diamond
interchange with 250-400 feet between the two ramp terminal intersections.
The TUDI is shown in Figure 1.3. The intersections are a conventional size,
and the clearance intervals are lower than those at the SPUI are. Because
the spacing between the two intersections is short, there is less storage for
the left turning and through movements.
The close proximity of the two intersections means that the traffic
signals need to work together to optimize the traffic flow and to prevent
spillback between the two intersections. Figure 1.4 shows the phasing
between the two intersections. The signal timing has four phases with two
overlapping phases.
26


Figure
1.3 TigW Urban
Diamond interchange
27


Figure 1.4 Signal Timing: Tight Urban Diamond Interchange
28


2.0 Methodology
The Leisch et al. (1989) and Fowler (1993) papers set up the classic
conflict regarding the operational analysis of the SPUl compared to the TUDI.
Both of these studies made assumptions that need to be reexamined in light
of new data. Leisch et al. (1989) heavily penalized the dual left turn flow rates
of the SPUl while Fowler (1989) underestimated the clearance interval for the
SPUl. These two assumptions could have contributed to the conclusions
reached by each study. The other studies covered in the earlier literature
review tend to favor the SPUl. However, not all the studies define their input
parameters. This study uses field data from Poppe et al. (1991) and Hook
and Upchurch (1993) to reanalyze the same traffic volume scenarios that
Leisch et al. (1989) and Fowler (1993) examined.
2.1 Study Approach
The approach of this study is to model 17 traffic volume scenarios in
both TRANSYT-7F and NETSIM and to determine whether the SPUl or the
TUDI operates more efficiently. TRANSYT-7F is used to obtain the optimal
signal timing for each scenario. These pre-timed signal timings are then
used in NETSIM. The average overall delay and the average delay for
29


individual turn movements for both the SPUI and TUDI in both software
packages will be analyzed.
2.2 Traffic Volume Scenarios
This study examines the same traffic volume scenarios as presented
in the Leisch et al. (1989) and Fowler (1993) studies, but uses more recent
field-verified data to make assumptions on clearance intervals, saturation
flow rates, and start up lost times. Five traffic scenarios from Leisch et al.
(1989) and twelve traffic scenarios from Fowler (1993) can be found in Table
2.1. The volume scenarios for Leisch et al. (1989) are labeled with an L
and the Fowler (1993) volume scenarios are labeled with an F. These
volume scenarios were chosen to gain a direct comparison with the results
from the Leisch et al. (1989) and Fowler (1993) studies. Total volumes range
from 5525 to 9900 vehicles per hour. Left turn volumes range from 190 to
1300 vph per approach. Right turn volumes range from 100 to 580 vph per
approach. Through volumes range from 900 to 3000 vph per approach.
2.3 Lane Geometry of the SPUI and TUDI
Since this study models high volume cases, the lane geometry was
optimized. Both interchange types have three through lanes in each
direction on the cross street with 400-foot right turn lanes and 400-foot dual
30


LOCATION DESIGN HOUR VOLUME
L1 360 I I I 720 V ,oo 900 bio 4 j rr 730 100 X 84oll f120
L2 120 | | | 190 V 100 ^ l W .3 i860 210 > j r r 690 ^ n r
L3 210 I I I 220 v 100 -J |J~- S 1400 1800 T 1 N. 590 I I |270
L4 130 I I I 210 V 580 ^ L- -* 1910 ^ ^ .il
L5 610 | I 690 V 1M l^ 1410 100 V 720 1 1 ( 590
F1 "J IT 5. 300 -> v r r 800 ==? xoilF
F2 300 J M 70 V joo ; 3000 i s \ 300 1 1 1 400
F3 300 IN 1300 V 200 * 1200 is si r?=m N 900 1 1 1 400
F4 300 J I 1 1300 V 200 L'*' -7 3D00 300 > j 800 S£s vir N 900 1 1 1 400
Table 2.1 Traffic Volume Scenarios
31


LOCATION DESIGN HOUR VOLUME
F5 300 III700 V 200 1200 700 4 j r p 300 x Moil ro
F6 300JM 700 V 200 -** 3000 700 300 s? , sir > 300 I I I 400
F7 300 ||, 1300 V 2Q0 I**- * 1200 700 4 j rr 300 x T1 ro
F8 30 J 1 I 1300 V 200 l^ 3000 700 4 j 4 p- 300 r ^ si r N 900 I I I 400
F9 300, | | 700 y 3Qo L'- 1900 400 'i Sir \ 300 1 1 1 400
F1 0 300 1 1 I 700 V.. .. 300 3noo 400 4 j 4 r 400 ,m v oooil r
F1 1 300 1 1 I 1300 V J L 3 Lv*- 1200 400 4 A 4 P 400 Sir X 900 I I I 400
F1 2 30 J I I 1300 > 300 L'*- 3000 400 r p 400 *£ = Sir > 900 I I I 400
Table 2.1 Traffic Volume Scenarios (cont)
32


left turn lanes. The off-ramps have 400-foot right turn lanes and 400-foot
dual left turn lanes. This lane geometry is constant for all 17 traffic
scenarios. This lane geometry is represented in Figures 1.1 and 1.3 for the
SPUI and TUDI found in Chapter 1.
2.4 TRANSYT-7F Modeling
TRANSYT-7F (Version 7.2) is a macroscopic signal network
optimization program. It was chosen because it could provide the optimal
signal timing for both the TUDI and the SPUI and because both Leisch et al.
(1989) and Fowler (1993) used previous versions.
The intersections were input with the basic coding required for
TRANSYT-7F with minimum green times for each signal phase. The
program optimized the signal timing to achieve the best operations for each
traffic scenario. It analyzed cycle lengths from 60 to 180 seconds at 15-
second intervals.
The following default settings were adjusted for both the SPUI and
TUDI. The average vehicle length for queue was set at 20 feet. The network
platoon dispersion factor was set at 50. The extension of effective green
time was set at 3 seconds.
Off-ramp right turns in the SPUI have green time when the cross
street left turns go. In order to minimize the time allocated to the off-ramp
33


right turns in the SPUI, these turns were coded as permitted turns during the
other two phases.
A sample printout of these runs for both the SPUI and TUDI can be
found in the appendix.
2.5 NETSIM Modeling
NETSIM is a microscopic stochastic simulation program. It was
chosen based on its ability to simulate heavy traffic situations and the
recommendations of the literature.
The network layout for the SPUI can be found in Figure 2.1 and for the
TDUI in Figure 2.2. Because of the stochastic nature of the program, three
runs were made for each interchange type/volume scenario and the average
delay of those runs was calculated. Each run was simulated for 3600
seconds. In each run, the random number generators were changed.
In order to control the input volumes, it was decided to model the
interchange ramp terminals in NETSIM rather than using both NETSIM and
FRESIM.
It was also decided to model just the intersections at the ramp
terminals themselves rather than add intersections on either side. This
34


Figure 2.1 NETSIM Network Layout of SPUI
35


sooj
8002
Figure 2.2 NETSIM Network Layout of TUDI
*800*
36


keeps the analysis focused on the operations of the intersections themselves
rather than introducing effects on traffic flow from adjacent intersections.
A sample printout for both the SPUI and TUDI can be found in the
appendix.
2.6 Assumptions
The saturation flow rates, clearance intervals, and start up lost times
were based on the field results from Poppe et al. (1991) and Hook and
Upchurch (1993).
2.6.1 Saturation Flow Rates
The saturation flow rates used in TRANSYT for the SPUI were 2070
vphgpl for the off-ramp left turns, 2000 vphgpl for the cross street left turn,
and 1900 vphgpl for the cross street through movement. The saturation flow
rates used in TRANSYT for the TUDI were 1900 vphgpl for the off-ramp left
turns, 2000 vphgpl for the cross street left turn, and 1900 vphgpl for the cross
street through movement.
NETSIM does not have a saturation flow rate variable but it does have
a discharge headway variable. This discharge headway was manipulated to
achieve the saturation flow rates in the range of 1900-2000 vphgpl. The
discharge headway of 1.4 seconds was used on the SPUI off-ramp and 1.5
37


seconds was used for all other movements on the SPUI and TUDI. The
original values for the discharge headway were 1.8 for the SPUI off-ramps
and 1.9 seconds for the remaining movements. This resulted in saturation
flow rates in the range of 1200-1400 vphgpl. Dixon (1998) noted a drawback
to NETSIM is that the individual movements on a link cannot have their own
discharge headway.
2.6.2 Clearance Intervals
The values for the clearance intervals are based on the dilemma zone
calculation to clear the intersection. The clearance interval for both
TRANSYT and NETSIM for the SPUI was a 5-second yellow with a 3-second
all-red. These values are based on a 200-foot intersection with an approach
speed of 30 mph. The clearance interval for the TUDI was a 3-second yellow
with a 2-second all-red. These values are based on a 75-foot intersection
with a 30-mph approach speed.
2.6.3 Start Up Lost Time
The start up lost time for both the SPUI and TUDI was 2 seconds for
all movements. Since the literature review had start up lost times ranging
from 1.5 seconds to 2.88 seconds, the default 2-second start up lost time
was used.
38


2.6.4 Network Speed/Free Flow Speed
The network speed in TRANSYT is 30 mph and the free flow speed
for all movements in NETSIM is 30 mph.
2.6.5 Default Parameters in NETSIM
There are a number of default parameters in NETSIM which were
used in this analysis for both the SPUI and TUDI. These default values can
be found in Table 2.2.
39


DEFAULT LINK GEOMETRIC DATA
WIDTH OF LANES 12 FEET
LONGITUDINAL DISTANCE FROM THE STOP 4 LINE TO THE NEAR CURB FEET
FORWARD SIGHT DISTANCE AT STOP LINE 1000 FEET
LANE CHANGE DATA
PARAMETERS VALUE UNITS
ENGLISH
DURATION OF LANE CHANGE MANEUVER 3 SECONDS
MEAN DRIVER REACTION TIME 1 SECOND
TIME REQUIRED FOR SUCCESSIVE LANE CHANGES 2 SECONDS
DECELERATION AT BEGINNING OF LANE CHANGE MANEUVER 5 FEET/SECOND2
DIFFERENCE IN VEHICLE'S DECELERATION OVER THE DISTANCE BETWEEN ITS POSITION WHEN IT BEGINS TO RESPOND TO AN OBSTRUCTION AND THE POSITION OF THE OBSTRUCTION -
FOR MANDATORY LANE CHANGE: 10
FEET/SECOND**2
FOR DISCRETIONARY LANE CHANGE: 5
FEET/SECOND**2
PANIC DECELERATION RATE OF LEAD VEHICLE FOR COMPUTATION OF CAR-FOLLOWING LAW FEET/SECOND**2 12
PANIC DECELERATION RATE OF FOLLOWER VEHICLE FOR COMPUTATION OF CAR-FOLLOWING LAW FEET/SECOND2 12
DRIVER TYPE FACTOR FOR DRIVER AGGRESSIVENESS 25
URGENCY THRESHOLD 10*SECONDS**2/FEET 2
SAFETY FACTOR FOR COMPUTATION OF PERCEIVED RISK OF LANE CHANGE 8 FACTOR*10
PERCENT OF DRIVERS WHO COOPERATE WITH A LANE CHANGER 50 %
HEADWAY BELOW WHICH ALL DRIVERS WILL ATTEMPT TO CHANGE LANES 2 SECONDS
HEADWAY ABOVE WHICH NO DRIVERS WILL ATTEMPT TO CHANGE LANES 5 SECONDS
FORWARD DISTANCE SCANNED BY DRIVER FOR A TURN MOVEMENT /
BUS STATION IN ORDER TO ASSESS NEED FOR A LANE CHANGE 300 FEET
Table 2.2 Default NETSIM Values
40


3.0 Results
This chapter presents the optimal signal timing for each case for both
the SPUI and TUDI, a discussion of the average delay calculations, the
results and discussion of the TRANSYT model, the results and discussion of
the NETSIM model, and the limitations in TRANSYT and NETSIM..
3.1 Optimal Signal Timing
Tables 3.1 and 3.2 show the optimal signal timing for each volume
scenario for the SPUI and TUDI, respectively. The clearance interval for all
phases of the SPUI was 5 seconds yellow with 3 seconds all-red. The
clearance interval for all phases of the TUDI was 3 seconds yellow with 2
seconds all-red.
3.2 Calculation of Delay
Delay to vehicles is one of the most important measures of
effectiveness in a traffic study. According to the TRANSYT Users Guide
(FHWA, 1991), delay represents an indirect cost to motorists in the time they
waste at an intersection and direct cost in the amount of fuel used when
waiting at an intersection.
41


Table 3.1 Optimal Signal Timing for SPUI from TRANSYT-7F
Case Overall Cycle Length sec Phase A Green Time sec Phase B Green Time sec Phase C Green Time sec
L1 75 14 19 18
L2 90 32 17 17
L3 75 22 12 17
L4 90 35 12 19
L5 75 23 16 12
F1 105 42 18 21
F2 180 93 29 34
F3 180 67 54 35
F4 180 81 46 29
F5 105 43 19 19
F6 180 95 30 31
F7 180 69 56 31
F8 180 83 47 26
F9 90 38 16 12
F10 135 74 23 14
F11 135 54 43 14
F12 180 88 52 16
Yellow time equals 5 seconds for each phase.
All-red time equals 3 seconds for each phase.


Table 3.2 Optimal Signal Timing forTUDI from TRANSYT-7F
Case Overall Cycle Length sec Intersection 4 * Intersection 5 Offset sec
Phase A Green Time sec Phase B Green Time sec Phase C Green Time sec Phase A Green Time sec Phase B Green Time sec Phase C Green Time sec
L1 60 13 19 13 14 13 18 11
L2 135 26 81 13 29 13 78 6
L3 90 23 39 13 16 20 39 26
L4 105 52 25 13 20 13 57 46
L5 90 13 31 31 29 13 33 9
F1 90 18 40 17 21 13 41 8
F2 180 37 92 36 33 13 119 116
F3 180 33 76 56 44 17 104 179
F4 180 34 79 52 38 13 114 5
F5 105 13 57 20 24 21 45 97
F6 180 15 113 37 33 30 102 98
F7 180 13 91 61 46 39 80 58
F8 180 14 97 54 39 29 97 72
F9 90 13 44 18 21 13 41 83
F10 135 14 78 28 26 13 81 77
F11 135 13 62 45 35 16 69 93
F12 180 19 92 54 41 17 107 121
Yellow time equals 3 seconds for each phase.
All-red time equals 2 seconds for each phase.
Intersection 4 is the master controller.


Both TRANSYT and NETSIM present approach delay. Approach
delay includes decelerating to stop at the back of a queue, stopping at the
intersection, and accelerating to get back to free flow speed. Stopped delay
measures only the time a vehicle is actually stopped at the intersection.
The comparison of delay between the SPUI and TUDI is not
straightforward. The SPUI has one intersection while the TUDI has two
intersections. The left turns and through movements at the TUDI experience
delay at both intersections. The average delay at the TUDI was calculated
by calculating the total delay for each movement and then dividing by the
total number of entering vehicles. According to the TRANSYT Users Guide
(FHWA, 1991), the average delay per vehicle is the total delay on the link
divided by the total flow on the link (p.7-25). The system average delay
given in TRANSYT is calculated by dividing the total delay by the total
number of vehicles on each link in the system. This method effectively
double counts a vehicle that goes through more than one intersection and
underestimates the actual system delay.
To calculate the delay for each individual movement on the TUDI, the
movement was examined to determine how many intersections the
movement travels through. The average delay for each movement per
intersection was added together. As shown in Figure 3.1, all the left turns
and through movements must travel through the two intersections. Only the
44


Figure 3.
1 TUDI Movements
that T ravel
Through Both
intersections


right turns travel through one intersection. For example, the northbound left
turn off the ramp goes through two intersections. It is the northbound left turn
on the westerly intersection and a westbound through movement on the
easterly intersection. The delay for the northbound left turn in Case L1 in
NETSIM is 26.1 sec/veh. This includes 24.3 sec/veh for the northbound left
turn and 1.8 sec/veh for the westbound through movement.
3.3 TRANSYT-7F Results
This section discusses the comparison of the overall intersection delay
and the delay by individual movement between the SPUI and TUDI in
TRANSYT-7F. Cases with spillback and a comparison with Leisch (1989)
and Fowler (1993) studies are also included.
3.3.1 Overall Intersection Delay
The overall intersection delay and volume-to-capacity ratio for ten of
the seventeen cases can be found in Table 3.3. The other seven cases were
discarded since the results were not reliable. The overall delay is lower for
the SPUI in all cases. Figure 3.2 shows a bar graph comparison between
the SPUI and TUDI.
46


Case Total Entering Volume vph SPUI Cycle Length v/c Delay sec % sec/veh TUDI Cycle Length v/c Delay sec % sec/veh
L1 5780 75 80 25.0 60 90 31.8
L2 6140 90 89 31.5 135 86 36.6
L3 6040 75 92 27.9 90 90 35.2
L4 5525 90 84 23.6 105 85 30.5
L5 6410 75 80 23.9 90 92 38.4
F1 6900 105 95 36.8 90 91 42.7
F2*
F3*
F4*
F5 6400 105 92 34.0 105 90 44.7
F6*
F7*
TV 00 *
F9 6600 90 90 25.6 90 89 35.9
F10 8400 135 95 38.1 135 98 57.7
F11 7800 135 96 48.0 135 96 75.1
F12*
* Cases discarded due to spillback
Table 3.3 Overall Intersection Statistics Comparison from TRANSYT-7F
The delay values for these cases are not significantly different
according to a two-tailed t test at a 95% confidence interval. A sample
calculation of the t test can be found in the appendix.
3.3.2 Delay by Individual Movement
The delay for each individual movement for both the SPUI and TUDI
can be found in Table 3.4. According to a two-tail t test at a 95% confidence
interval, the off-ramp left turns, the cross street left turns, and the through
47


Figure 3.2 Overall Intersection Delay in TRANSYT-7F
Traffic Volume Scenario


Table 3.4 TRANSYT-7F Average Delay by Movement
WB LT WB THRU WB RT NB LT NB RT
SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI
Delay Delay Delay Delay Delay Delay Delay Delay Delay Delay
Case sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh
L1 27.1 39.0 31.5 26.9 7.1 14.1 27.1 33.3 2.2 23.8
L2 43.8 77.4 31.2 23.9 6.0 11.5 45.2 59.4 8.5 53.4
L3 36.3 41.0 25.2 27.5 6.7 14.0 36.4 44.0 4.2 42.9
L4 38.7 22.0 26.3 21.6 9.1 15.2 36.8 37.0 1.2 55.7
L5 28.4 40.9 23.8 38.0 4.7 17.7 29.7 35.2 16.8 54.9
F1 69.5 65.9 23.0 32.4 6.8 13.7 38.0 39.5 22.8 48.0
F5 36.6 57.8 21.6 28.9 6.1 18.0 37.1 38.5 23.5 58.5
F9 39.0 50.5 17.8 24.1 4.5 14.7 31.3 33.9 21.6 48.0
F10 83.0 144.0 36.0 44.4 3.4 11.8 48.6 47.3 25.8 228.0
F11 83.0 165.0 30.0 38.1 3.4 17.9 39.5 73.5 87.5 52.4
Bold values represent the lower delay.


Table 3.4 TRANSYT-7F Average Delay by Movement
EBLT EB THRU EB RT SB LT SB RT
SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI
Delay Delay Delay Delay Delay Delay Delay Delay Delay Delay
Case sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh
L1 29.8 44.9 31.5 26.6 7.1 14.1 24.6 28.3 3.5 32.8
L2 29.6 83.3 30.5 16.2 6.0 10.5 30.3 57.2 6.4 61.9
L3 28.4 47.2 32.5 30.8 6.7 14.0 26.2 35.9 3.1 42.4
L4 28.9 58.7 18.9 42.1 6.4 30.2 34.0 41.9 4.7 43.0
L5 29.3 46.2 23.5 40.7 4.7 19.8 28.7 30.8 19.0 45.5
F1 34.8 77.9 39.5 38.3 8.4 18.3 66.3 56.3 2.2 43.8
F5 60.2 84.5 35.1 33.4 5.7 11.0 54.3 67.1 2.2 49.7
F9 39.0 79.7 27.2 31.2 4.8 13.9 49.1 51.5 2.1 40.5
F10 83.0 130.3 22.0 29.2 3.7 14.3 84.9 66.1 34.4 54.2
F11 83.0 143.0 49.4 70.2 3.7 24.5 67.3 115.1 7.6 34.3
Bold values represent the lower delay.
o
O


movements are not statistically different. All four of the right turns are
statistically different. However, the average delay on the right turns is so
small in relation to the other movements, it has no effect on the overall delay.
3.3.3 Cases with Spillback/Oversaturation
The TRANSYT Users Guide (FHWA, 1991) states that the program
disregards the effect of spillback in its optimization process and vertically
stacks vehicles in a queue. It states that if spillover does occur, the program
will not give realistic results. The guide also states that if the volume to
capacity ratio is greater than 95% and the signals are pre-timed, the estimate
of delay will not be reliable.
Based on these two statements, the seventeen traffic volume cases
were examined and seven cases were discarded because both spillback
occurred on a number of movements and the volume to capacity ratio was
greater than 99%. Figure 3.3 shows the volume to capacity ratio for each
case. To determine if spillback occurred, the maximum back of queue for
each movement was examined and compared against the storage length of
that movement. Cases that had 6 or more movements with spillback were
considered to be unreliable. A summary of these seven cases can be found
in Table 3.5. Of these seven cases, the TUDI had spillback on more
movements than the SPUI.
51


140
3 Q
Q. 3
co i-
S


NV\\\\y\vv\\\\\vv\vwvVvVvv\vvv
\\\\\\\\\\\\\\\\\\\\S\X\\\\|
N\\\\\\\\\\\\\\N\\\\\\\\\\\\
[VWWVVvVvWvWWVVvVVVWN

pv\\>^!^\\\VCCv\\\\\\\\\\\\\Vs\
N\\\\\\\\\\\\\\\\\\\\\\\V
IS\NS\\\\\\\\\\^
^\\\\\>>\\^
X\\\\\\\\\\\\\\\\\\\\\\Y1
ss
o
CM
(%) oiiey o/A
Figure 3.3 Volume to Capacity Ratio Comparison in TRANSYT-7F
52
Traffic Volume Scenario


Case Total Entering Volume vph Cycle Length sec SPUI v/c % Delay sec/veh Cycle Length sec TUDI v/c % Delay sec/veh
F2 8700 180 103 75.4 180 103 86.7
F3 8100 180 103 90.6 180 102 139.2
F4 9900 180 120 291.2 180 113 234.1
F6 8200 180 99 61.1 180 99 96.7
F7 7600 180 99 73.9 180 96 95.3
F8 9400 180 118 251.9 180 110 239.9
F12 9600 180 107 135.6 180 107 178.1
Table 3.5 Summary of Spillback Cases from TRANSYT-7F
In Cases F4 and F8, the off-ramp left-turn volumes and the through
volumes were very high, and one of the cross-street left-turn volumes was
high. In Case F4, the delay for the SPUI was 291.2 seconds/vehicle and the
delay for the TUDI was 234.1 seconds/vehicle. In Case F8, the SPUI delay
was 251.9 seconds/vehicle and the TUDI delay was 239.9 seconds/vehicle.
While a direct comparison cannot be made between the two interchange
types, the overall delay in both these cases is at Level of Service F. This
indicates that the geometry of the intersections is not adequate. Perhaps a
fourth through lane needs to be added or longer left-turn storage lengths are
needed.
53


3.3.4 Comparison with Leisch and Fowler
This study showed that for all five traffic volume scenarios from the
Leisch et al. (1989) study, the SPUI had the lower overall intersection delay.
Leisch et al. (1989) reported that the TUDI had the lower delay in four out of
the five cases. The overall intersection delay and optimum cycle lengths for
the SPUI and TUDI for both studies can be found in Table 3.6.
The differences in the results could be due to a number of factors.
This study used much higher saturation flow rates for both interchange types.
Table 3.7 shows a comparison of the saturation flow rates between the two
studies. The most significant difference is in the off-ramp left-turn saturation
flow rate. This study used 2070 vphgpl and Leisch et al. (1989) used 1700
vphgpl. In the Leisch et al. (1989) study, the TUDI had the shorter cycle
lengths, but the SPUI had shorter cycle lengths in this study. Both studies
used the same clearance intervals for both interchange types.
Another possible difference could be in the delay calculation. If Leisch
et al. (1989) used the system delay value from the TRANSYT output for the
TUDI, the delay valves could be underestimated.
This study confirmed the results of the Fowler (1993) study even
though the modeling was done differently and different assumptions were
made. Fowler modeled the TUDI as one node in TRANSYT where this study
modeled the TUDI as two nodes. The seven cases that had volume to
54


Table 3.6 and Table 3.7
U1
cn
Case SPUI TUDI
Leisch This study Leisch This study
Cycle Length sec Delay sec/veh Cycle Length sec Delay sec/veh Cycle Length sec Delay sec/veh Cycle Length sec Delay sec/veh
L1 95 37.9 75 25.0 85 41.8 60 31.8
L2 120 55.7 90 31.5 90 39.9 135 36.6
L3 110 41.3 75 27.9 80 38.3 90 35.2
L4 120 46.5 90 23.6 85 26.1 105 30.5
L5 100 40.5 75 23.9 75 34.4 90 38.4
Table 3.6 Comparison to Leisch Results Cycle Length and Overall Delay
SPUI TUDI
Movement Leisch This Study Leisch This Study
Off ramp Lefts 1700 2070 1600 1900
Cross street Lefts 1700 2000 1600 2000
Cross street Throughs 1800 1900 1800 1900
Table 3.7 Comparison to Leisch Results Saturation Flow Rates in vphgpl


capacity ratios over 96% in this study had volume to capacity ratios over 96%
in the Fowler (1993) study.
Fowler (1993) modeled all twelve cases at 80% of their volumes to
avoid the high volume to capacity ratios. In this cases, the TUDI had higher
volume to capacity ratios than the SPUI in eleven of the twelve cases. The
TUDI had a higher delay in all twelve cases with the delay ranging from 30 to
40 seconds.
Of the ten cases that this study analyzed, the TUDI had the higher
volume to capacity ratio in 4 cases. The SPUI had the higher volume to
capacity ratio in the other six cases. The TUDI had the higher delay in all ten
cases with the delay ranging from 30 to 75 seconds. The higher delay
values in this study result from using the Fowler (1993) traffic volume
scenarios at 100% rather than 80%.
3.4 NETSIM Results
This section discusses the comparison of the overall intersection delay
and the delay by individual movement between the SPUI and TUDI in
NETSIM. Cases with spillback and observations from TRAFVU are also
discussed.
56


The delay values shown in both the overall intersection and individual
movement delay is obtained by taking the average of the three runs for each
volume scenario.
3.4.1 Overall Intersection Delay
The overall intersection delay for all cases can be found in Table 3.8,
except for the 6 cases which had to be discarded due to spillback. The
overall delay is lower for the SPUI in all eleven cases. Figure 3.4 shows the
comparison between the SPUI and TUDI graphically. The delay values are
not significantly different according to a two-tailed t test at a 95% confidence
interval.
3.4.2 Delay by Individual Movement
The delay for each individual movement for each case for both the
SPUI and TUDI can be found in Table 3.9. According to a two-tail t test at a
95% confidence interval, the average delays on the major individual
movements (lefts and throughs) are not statistically different. The average
delay on the eastbound right turns is significantly different but is too small to
impact the overall intersection delay.
For the westbound left-turn movement, the SPUI had a lower delay
than the TUDI in all cases except for Case F1 where the TUDI delay was
57


Case Total Entering Volume vph NETSIM
SPUI TUDI
Delay sec/veh Delay sec/veh
L1 5780 25.3 29.8
L2 6140 27.6 32.8
L3 6040 24.5 30.4
L4 5525 22.8 25.8
L5 6410 23.3 31.2
F1 6900 28.6 30.7
F2*
F3*
F4*
F5 6400 29.4 36.6
F6 8200 46.6 49.1
F7*
F8*
F9 6600 23.3 30.9
F10 8400 30.9 36.3
F11 7800 48.4 53.4
F12*
*Cases discarded due to spillback
Table 3.8 Overall Intersection Delay Comparison from NETSIM
58


Figure 3.4 Overall Intersection Delay in NETSIM
Traffic Volume Scenarios


Table 3.9 NETSIM -Average Delay by Movement
WB LT WB THRU WB RT NB LT NB RT
SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI
Delay Delay Delay Delay Delay Delay Delay Delay Delay Delay
Case sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh
L1 25.7 45.3 28.2 29.5 5.1 6.2 32.9 26.1 7.0 9.2
L2 36.2 67.7 26.4 23.4 4.4 5.7 38.9 56.3 13.1 24.2
L3 28.1 42.5 23.4 26.5 5.2 6.0 31.7 40.2 10.4 10.7
L4 34.8 22.5 23.8 21.9 5.5 7.1 39.0 39.9 8.8 8.2
L5 28.4 37.6 22.3 36.2 4.8 5.7 39.2 39.0 11.6 11.7
F1 43.0 42.0 23.1 36.7 4.8 5.1 45.4 46.8 18.1 18.2
F5 37.0 59.9 21.8 30.2 5.6 6.1 42.2 43.6 18.3 29.0
F6 72.0 146.1 39.9 35.4 14.6 19.9 77.1 70.7 36.5 26.8
F9 37.4 50.8 17.8 25.1 4.8 5.2 37.6 39.0 16.9 23.7
F10 61.9 108.9 26.2 27.6 7.0 16.4 58.0 52.5 26.6 18.9
F11 59.6 113.6 28.8 37.9 4.9 5.5 136.8 67.2 31.5 44.1
Bold values represent the lower delay.


Table 3.9 NETSIM -Average Delay by Movement
EBLT EB THRU EB RT SB LT SB RT
SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI SPUI TUDI
Delay Delay Delay Delay Delay Delay Delay Delay Delay Delay
Case sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh sec/veh
L1 30.7 49.4 28.7 31.0 6.1 6.1 25.1 25.9 7.3 9.1
L2 33.1 83.2 25.8 17.3 6.2 6.1 29.7 61.9 11.9 16.3
L3 31.8 40.8 24.1 27.2 6.6 6.5 26.8 39.9 9.6 10.3
L4 31.4 55.4 18.9 39.3 6.0 6.4 36.8 46.7 11.8 16.1
L5 31.1 40.0 22.4 35.3 4.7 5.7 27.8 33.8 12.1 13.3
F1 39.1 63.3 28.7 26.1 8.4 10.1 44.0 40.5 10.9 8.4
F5 49.4 75.7 28.0 28.6 4.7 6.7 41.4 51.2 10.5 9.0
F6 90.1 107.2 31.9 31.7 5.6 8.3 78.9 82.3 36.7 41.4
F9 41.1 81.7 22.8 27.9 5.7 7.2 36.9 44.8 10.0 8.9
F10 73.6 78.2 20.7 28.3 5.8 8.2 54.8 60.1 31.2 36.8
F11 78.1 43.7 36.5 49.7 6.7 13.6 52.6 62.4 12.7 20.8
Bold values represent the lower delay.
o
O


only 1 second lower. For the eastbound left-turn movement, the SPUI had a
lower delay than the TUDI in all cases except for Case F11. In Case F11,
the TUDI delay was 43.7 seconds/vehicle and SPUI delay was 78.1
seconds/vehicle. A graphical comparison between the SPUI and TUDI for
these cross-street left-turn movements can be found in Figures 3.5 and 3.6.
For the northbound off-ramp left turns, the SPUI had a lower delay in
six out of the eleven cases. Out of the five cases where the TUDI had the
lower delay, only Case F11 showed a large difference in delay. The TUDI
had a delay of 67.2 seconds/vehicle and the SPUI had a delay of 136.8
seconds/vehicle. For the southbound left-turn movement, the SPUI had a
lower delay than the TUDI in all cases except for Case F1 where the TUDI
delay was only 3.5 seconds lower. A graphical comparison between the
SPUI and TUDI for these off-ramp left-turn movements can be found in
Figures 3.7 and 3.8.
For the westbound through movement, the SPUI had a lower delay in
eight of the eleven cases. The three cases where the TUDI had the lower
delay were Cases L2, L4, and F6. The SPUI also had a lower delay in eight
of the eleven cases for the eastbound through movement. Cases L2, F1,
and F6 had lower delays for the TUDI eastbound through movements. A
graphical comparison between the SPUI and TUDI for these cross-street
through movements can be found in Figures 3.9 and 3.10.
62


3 D
Q- 3
W I
G IS


^^gggjg

o





ES
i^ggj
o
*t
o
o
(M9A/08S) Aeiaa aBejaAv
Figure 3.5 Average Delay Comparison of Westbound Left Turns
63
Traffic Volume Scenario


40.0
3 Q
Q. 3
W H
S



^\\\\\\\^
^\\N\\\\\\S

^>\\\\\^N>\\>^
io
CM
O
IT)
O
O
O
lO
o
o
(qsA/oes) Ae|3Q dBejsAV
Figure 3.6 Average Delay Comparison of Eastbound Left Turns
64
Traffic Volume Scenarios


160.0
D Q
CL D
CO H
0 S
o
Tf
o
eg
o
o


k\\\\V\\SN
aWH




J\V
1 hhi
r
O
o
00
o
o
^j-
o
eg
o
o
(qaA/oas) Ae|eg a6ejaAv
Figure 3.7 Average Delay Comparison of Northbound Left Turns
65
Traffic Volume Scenario


120.0

D n
Q_ D
C/) f
IS



ss
k\\\\\\\\\\\\\\v
O)
LL
r


(i|0A/o9s) Ae|sa sBbjsav
Figure 3.8 Average Delay Comparison of Southbound Left Turns
66
Traffic Volume Scenario


009

, > n
Q_ 3
CO H
0 B

O
o
in
^^|-'< *
XSSSSSSSt0CWNNNSS>
I . . "
(IJ3A/33S) ABiaa 36BJ3AV
Figure 3.9 Average Delay Comparison of Westbound Through Movement
67
Traffic Volume Scenario


006
W h-
s
^h^HK^RSRRRSSRI

o o O o O o o o o
o d o d o o o d d
00 h- CO UD CO CN V
(qaA/oas) Aepa aBejaAV
Figure 3.10 Average Delay Comparison of Eastbound Through Movement
68
Traffic Volume Scenario


3.4.3 Cases with Spillback
Spillback usually means that the vehicle queue is so long that it blocks
the operations of the adjacent intersection. Since there were no adjacent
intersections in the network for the SPUI, spillback occurred at the entry
nodes. Spillback did not typically occur between the two intersections of the
TUDI. The spillback at the TUDI occurred at the entry nodes. The spillback
affected the number of vehicles that could enter the system.
There were eight cases where spillback occurred at the entry nodes.
In these cases, the actual number of vehicles through the network was lower
than the specified volumes. A summary of these cases and the actual
volumes can be found in Table 3.10.
Case Total Entering Volume vph NETSIM
SPUI TUDI
Volume veh Delay sec/veh Volume veh Delay sec/veh
F2 8700 8330 50.6 8700 52.3
F3 8100 7790 67.6 7660 81.4
F4 9900 8610 72.6 8320 94.9
F6 8200 8200 46.6 8200 49.1
F7 7600 7390 69.4 6990 82.0
F8 9400 8420 82.5 8700 77.1
F11 7800 7800 48.4 7800 53.4
F12 9600 8860 79.1 5940 130.9
Table 3.10 Summary of Spillback Cases from NETSIM
69


In two of the cases, F6 and F11, the spillback was minimal and
occurred on only one approach. The specified volumes were handled in both
the SPUI and TUDI by the network. These two cases were included in the
overall analysis.
In the other six cases, the spillback was extensive and occurred on
three or four of the approaches. These six cases were not included in the
overall analysis because of the variation in the volumes between the actual
and specified and the variation of volumes between the SPUI and TUDI.
Figures 3.11 and 3.12 show the movements where spillback occurred most
often. These movements are the eastbound and westbound through
movements and the northbound and soundbound left turn movements.
The seven cases that were discarded in the TRANSYT analysis
included the same six cases discarded in NETSIM plus Case 6 which
experienced some spillback in NETSIM. This shows that the programs have
similar limitations with regard to saturated flow conditions.
3.4.4 TRAFVU Observations
TRAFVU is the visualization program of the CORSIM modeling
software. It allows a users to watch the vehicles move through the network.
Observing the cases in TRAFVU gave some additional insight to how the
network is handling traffic. In the eleven cases with no spillback, the
70


71


72


operations in TRAFVU ran smoothly. For the TUDI, there was a slight delay
on the second intersection for the cross-street through movements. The
delay was large enough to queue 7-10 vehicles per lane.
Cases F6 and F11 had slight spillback but NETSIM could
accommodate the designated volumes. In observing the SPUI in Case F6,
the westbound through movement occasionally blocked the westbound left-
turn lane. The westbound through movement of 3000 vehicles per hour was
the approach where the spillback was occurring. In the TUDI in Case F6,
there was spillback on both cross-street approaches.
In observing the SPUI in Case F11, spillback was detected on both
the northbound and southbound off-ramps. The left turning vehicles were
blocking the right turn lane. In the TUDI in Case F11, spillback was detected
on the eastbound approach.
Cases F4 and F12, the two highest volume cases, were also observed
in TRAFVU. In observing the SPUI in Case F4, the westbound through
movement blocked the westbound left turn lanes; the westbound left turn
lanes spilled back into the through lanes and both ramps were experiencing
spillback. The left turns from the ramps never dissipated.
In the TUDI in Case F4, there was spillback between the ramp
terminals for the westbound through movement. The signal timing between
the two intersections caused this spillback. The east intersection had a
73


green for through movement, but the west intersection had a red for the
through movement and the a long queue developed. This restricted the
northbound left turn movement and caused the northbound ramp approach to
spillback.
In observing the SPUI in Case F12, all the approaches were
experiencing spillback. The spillback on the off ramps was so large that the
queue of left turns never dissipated. The spillback blocked the off ramp right
turn lane. The eastbound left turns were spilling back into the center through
lane. The TUDI in Case F12 experienced these same problems.
3.5 Volume Limitations in NETSIM and TRANSYT
In this study, both TRANSYT and NETSIM had problems dealing with
high traffic volumes that would represent saturated flow conditions. NETSIM
worked well for a majority of cases where the traffic volumes were less than
8500 total entering vehicles. Eleven of the seventeen cases were modeled
with no problems and the delay was reasonable. In the other 6 cases,
NETSIM couldnt handle the volumes and spillback occurred. Figure 3.13
shows the average delay versus the total entering vehicles. As the traffic
volumes increase, the delays become very high and unreliable.
TRANSYT worked well for a majority of cases where the volume-to-
capacity ratio was less than 96%. Ten of the seventeen cases were modeled
74


120.0
3 D
Q. 3
h-
(ijeA/oes) Abisq aBejaAV
O
O
O
O
O
O
O
O)
O
O
O
00
O
O
O
h-
o
o
o
CO
o
o
o
o
o
m
Figure 3.13 Average Intersection Delay versus Total Entering Volume from
NETSIM
75
Entering Volume (vph)


with no problems. In the other seven cases, spillback occurred and the
model broke down. Figure 3.14 shows the average delay versus the volume-
to-capacity ratio. As the volume-to-capacity ratio increases, the average
delays become very high and unreliable.
Neither of the programs can handle volumes that neared saturation of
the system. It would seem appropriate that the model could handle saturated
flow conditions up to a certain volume before breaking down. Messer and
Bonneson (1997) commented that it was unknown whether NETSIM and
TRANSYT could provide accurate results at oversaturated and near
saturated conditions.
Many of our street networks function in saturated flow conditions. It is
disappointing that these traffic models can not handle congested
intersections and give meaningful results. These highly congested
intersections are often the ones that would benefit most from traffic modeling.
76


350.0


4
-4
3 O
0- 3
W h-
4
4
4
4
o
o
Csl
44 4
\
4

o
o
o
CO
I
o o O O o o
o o o o d o
o in o ID o LD
CO CM CM T~
(qeA/oas) Abiqq eBejeAV
o
o
CM
I-
o
d
o
Figure 3.14 Average Delay versus Volume-to-Capacity Ratio from
TRANSYT -7F
77
V/C Ratio (%)


4.0 Conclusion and Recommendations
This chapter presents a summary of this research, the conclusions
and observations from this study, and provides recommendations for further
research.
The purpose of this study was to determine whether a SPUI or TUDI
operates more efficiently at high volumes. A literature review was conducted.
There were seven previous studies that examined the operational efficiency
of the SPUI and TUDI. The two most important are the Leisch (1989) and
Fowler (1993) studies. These two studies had different results. Leisch
(1989) concluded that the TUDI works better and Fowler (1993) concluded
that the SPUI works better. Both studies made assumptions about saturation
flow rates and clearance intervals that could have skewed the results. This
study used field data obtained from other studies for saturation flow rate,
clearance intervals, and start up lost time. The Poppe (1990) study
examined these parameters for the SPUI. The Hook and Upchurch (1992)
study examined these parameters for the TUDI and compared them to the
Poppe (1990) study.
78


Seventeen traffic volume scenarios, taken from the Leisch (1989) and
Fowler (1993) studies, were modeled in both TRANSYT and NETSIM using
the field- verified parameters. The SPUI had a lower overall intersection
delay than the TUDI in both TRANSYT and NETSIM for the cases that could
be analyzed.
Seven cases in TRANSYT and six cases in NETSIM were discarded
because spillback occurred and the results were not reliable. A major
drawback to both models is that the results become unreliable in saturated
traffic flow conditions. In NETSIM, the model failed to process the correct
number of vehicle when spillback occurred. In TRANSYT, the queues are
stacked vertically, and the queue lengths are much longer than the network
would actually allow.
The literature review showed that higher saturation flow rates may be
used than previously thought. Field studies are showing saturation flow
rates of 1900 to 2000 vphgpl for through movements and saturation flow
rates of 1900 to 2040 vphgpl for dual left turns. These flow rates should be
used to modify the assumptions used in all the modeling software and
including the Flighway Capacity Manual.
This study questioned the methodology used to calculate average
system delay in TRANSYT. Average system delay calculations in TRANSYT
are based on the total delay divided by the number of vehicles per approach
79


for each intersection. This method double or triple counts vehicles that travel
through more than one intersection. The average system delay calculations
should be based on the total number of vehicles entering the system. This
calculation was used in this study to calculate the average system delay for
the TUDI which has two intersections.
Even though this study and many of the other research studies
covered in the literature review support the SPUI, there is still some
hesitation to use the SPUI. The support of the SPUI is counter to the
prevailing thoughts among some traffic engineers. Some traffic engineers
would not consider the SPUI a viable alternative for areas with high through
volumes. More positive papers and articles supporting the SPUI would help
improve the SPUIs reputation. A comprehensive study of the saturation flow
rates and the operational efficiency of existing SPUIs, especially ones with
high through volumes, would provide positive information to dispute the
previous claims that SPUIs dont work with high through volumes.
Other recommendations for further research center around the
modeling software. Modifications to both NETSIM and TRANSYT are
needed to handle saturated conditions in a realistic, reliable manner. These
modifications need to be field-verified and calibrated to handle spillback in
saturated and oversaturated flow conditions. This is very important because
saturated and oversaturated conditions are often found in the urban areas.
80


In these areas, the option of adding more lanes is limited, and the results
from these models need to be reliable to make the best decision
operationally. Another modification to NETSIM is needed to allow movement
specific discharge headways. This would increase the accuracy of the
model.
81


Appendix A: Sample Output File for SPUI TRANSYT-7F
82


******************************************************************************
Release 7.2
(TRANSYT-7F) February 1994
TRAFFIC SIGNAL SYSTEM OPTIMIZATION
PROGRAM
Sponsored by:
U.S. Department of Transportation
Federal Highway Administration
Developed by:
University of Florida
Transportation Research Center
************
Software Maintenance and User Support Furnished by:
Center for Microcomputers in Transportation (McTrans)
Transportation Research Center, University of Florida
512 Weil Hall, Gainesville, FL 32611-6585 USA
(904) 392-0378
TRANSYT/7 (C) British Crown Copyright.
TRANSYT-7F Copyright 1980-1994, University of Florida.
All Rights Reserved.
k************************************************************
Date of Run: 9/21/99 Start Time of Run: 19:40: 0 Data File: SPU LI.TIN
INPUT DATA REPORT FOR RUN 1
FIELDS:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
SPU_L1
1 60 180 15 3 20 2 3 20 1 1 60 0 0 0 0
--- 5----NOTE -
+ THE SEC/STEPS FACTOR IN FIELD 6 IS
INTERPRETED AS 2.000.
>>> 106 +++ WARNING +
2 4 0 0
1TRANSYT-7F:
SPU_L1
FIELDS:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
3 12 11 10 4 5 6 3 2 1 9 8 7 0 0 0
10 4 5 3 1800 30 50 0 20 0 0 0 0 0 0 0
INTERSECTION 4
13 4 0 1 12 5 3 12 5 3 12 5 3 0 0 0
21 4 1 1 2 3 20 401 402 407 408 -404 -410 0 0 0
22 4 4 4 5 6 20 401 406 407 412 -404 -410 0 0 0
THE SEC/STEPS FACTOR IN FIELD 6 IS TOO SMALL FOR CYCLE
LENGTHS ABOVE 120 SECONDS. IT WILL BE INCREASED TO
ALLOW A MAXIMUM OF 60 STEPS/CYCLE.
000000000000
PAGE 2
83


23 4 7 7 8 9 20 403 404 409 410 0 0 0 0 0
28 401 400 1900 100 0 0 0 0 0 0 0 0 0 0 0
28 402 400 5700 900 0 0 0 0 0 0 0 0 0 0 0
28 403 400 4000 810 0 0 0 0 0 0 0 0 0 0 0
28 404 400 1900 360 0 0 0 0 0 0 0 0 0 0 0
29 404 0 0 0 0 0 0 0 408 34 412 50 0 0 0
28 406 400 4140 720 0 0 0 0 0 0 0 0 0 0 0
28 407 400 1900 100 0 0 0 0 0 0 0 0 0 0 0
28 408 400 5700 900 0 0 0 0 0 0 0 0 0 0 0
28 409 400 4000 730 0 0 0 0 0 0 0 0 0 0 0
28 410 400 1900 320 0 0 0 0 0 0 0 0 0 0 0
29 410 0 0 0 0 0 0 0 402 34 406 50 0 0 0
28 412 400 4140 840 0 0 0 0 0 0 0 0 0 0 0
PLOT AND OPTION CARDS
52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
---72----NOTE -
+ A CARD TYPE 52 CAUSES RUN TO BE OPTIMIZED USING THE
DEFAULT NORMAL OPTIMIZATION STEP SIZES.
IF CARD TYPE 4 WAS CODED, IT IS IGNORED.
THE ABOVE WILL BE PROCESSED AFTER THE "BEST" CYCLE
LENGTH HAS BEEN SELECTED.
NO ERRORS DETECTED. TRANSYT-7F PERFORMS FINAL PROCESSING.
IF ANY ERRORS ARE DETECTED, FURTHER PROCESSING IS SUSPENDED.
PAGE 3
SPU LI
---70-----NOTE -
+
1TRANSYT-7F:
FIELDS:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
--- 74---NOTE -
+ THERE ARE A TOTAL OF 1 NODES AND 10 LINKS,
INCLUDING BOTTLENECKS, IF ANY, IN THIS RUN.
--- 77 --- NOTE -
+ THERE WERE A TOTAL OF 1 WARNING MESSAGES ISSUED
IN THE ABOVE REPORT.
1TRANSYT-7F: PAGE 4
SPU LI
CYCLE EVALUATION SUMMARY PERFORMANCE
CYCLE LENGTH (sec) STEP SIZE (steps) AVERAGE DELAY (sec/veh) PERCENT STOPS (%) FUEL CONSUMPTION (gal/hr) DISUTILITY INDEX NUMBER SATURATED LINKS PERFORMANCE INDEX
60 20 27.06 80 75.1 69.1 0 69.0681
75 25 24.60 78 71.7 64.6 0 64.5500
84


90 30 27.60 79 75.5 69.7 0 69.7066
105 35 30.57 79 78.9 74.3 0 74.3237
120 40 34.54 79 83.7 80.8 0 80.8410
135 45 36.71 78 85.7 83.8 0 83.8372
150 50 40.80 78 90.8 90.7 0 90.6640
165 55 44.59 79 95.3 96.8 0 96.7938
180 60 48.10 78 99.4 102.4 0 102.4272
BEST CYCLE LENGTH = 75 SEC. CYCLE SENSITIVITY = 16.2 %
---80----NOTE -
+ TRANSYT-7F OPTIMIZES THE SYSTEM USING THE BEST
CYCLE LENGTH AND HILL-CLIMB STEP SIZES AS
INDICATED BY CARD TYPE 52.
1TRANSYT-7F: Page 5
SPU_L1
CYCLE: 75 Seconds, 38 Steps

MOVEMENT/ NODE NOS. v/c (%) TOTAL TRAVEL (v-mi) TRAVEL TIME TOTAL AVG. (v-hr)(sec/v) TOTAL DELAY (v-hr) AVG. DELAY (sec/v) UNIFORM STOPS NO. (%) MAX OF EST BACK 0UEUE .CAP. FUEL CONS. (gal)
NB LEFT 76 63.68 8.46 36.2 6.32 27.1 714 ( 85) 16 32 11.16
RGHT 35 24.26 1.01 11.3 . 19 2.2 102 ( 32) 2 16 1.70
SB LEFT 40 27.29 1 27 12.7 . 35 3.5 159 ( 44) 3 16 2.25
RGHT 65 54.58 6.74 33.7 4.91 24.6 588 ( 82) 13 32 9.06
EB THRU 79 68.23 10.17 40.7 7.88 31.5 805 ( 89) 17 48 12 99
LEFT 80 61.40 8.76 38.9 6.70 29.8 703 ( 87) 15 32 11.28
RGHT 9 7.58 .45 16.2 .20 7.1 40 ( 40) 1 16 . 67
WB THRU 79 68.23 10.17 40.7 7.88 31.5 805 ( 89) 17 48 12.99
LEFT 72 55.34 7.34 36.2 5.49 27.1 621 ( 85) 14 32 9.70
RGHT 9 7.58 .45 16.2 .20 7.1 40 ( 40) 1 16 . 67
NODE 4 80 438.17 54 81 40.12 25.0 4 575 ( 79) 72.47
All MOEs are in units per hour.
1TRANSYT-7F: Page 6
S PU_L1
CYCLE: 75 Seconds, 38 Steps
85


SYSTEM-WIDE PERFORMANCE: ALL NODES
SYSTEM
PERFORMANCE MEASURES UNITS TOTALS
Total Travel
Total Travel Time
Total Uniform Delay
Total Random Delay
Total Delay
Average Delay
Passenger Delay
Stops: Total
Percentage
System Speed
Fuel Consumption
Operating Cost
Performance Index
veh-mi/hr 438
veh-hr/hr 55
veh-hr/hr 36
veh-hr/hr 4
veh-hr/hr 40
sec/veh 25.0
pax-hr/hr 48
veh/hr 4575
% 79
mph 8.0
gal/hr 72
$/hr 454
DI 65.5
Performance Index (PI): Disutility Index (DI):
Disutility Index Delay + Stops
No. of Simulations = 60, Links = 490 Elapsed Time =
1TRANSYT-7F:
SPU_L1
CYCLE: 75 Seconds, 38 Steps
TRANSYT-7F TRAFFIC SIGNAL TIMING TABLES
NETWORK-WIDE SIGNAL TIMING PARAMETERS
SYSTEM CYCLE LENGTH = 75 SECONDS
MASTER OFFSET REFERENCE LOCATION = INTERSECTION NO.
Key to Interval Types:
F : Fixed green.
V : Variable green.
Y : Yellow.
R : All-red.
An 'M' by an interval length means this is the minimum t
INTERSECTION CONTROLLER SETTINGS
.8 sec.
Page 7
4 START OF INTERVAL 1.
ime available.
86


INTERSECTION
4
PRETIMED SPLITS OPTIMIZED
Interval Number :
Intvl Length(sec):
Intvl Length (%):
Pin Settings (%):
Phase Start (No.):
Interval Type
Splits (sec):
Splits (%):
Links Moving
Offset = 0 sec
1 2 3 4
14 5 3 19
19 7 4 24
100/0 19 26 30
1 2
V Y R V
22 27
30 35
401 401
402 406
407 407
408 412
-404 -404
-410 -410
0 %.
5 6 7 8 9
5 3 18 5 3
7 4 24 7 4
54 61 65 89 96
3
Y R V Y R
26
35
403
404
409
410
This is the master controller.
>>> 193 + + + WARNING +
+
1TRANSYT-7F:
THE OFFSET FALLS WITHIN 1% OF AN INTERVAL
CHANGE POINT AT THE START OF INTERVAL NO.
1
SPU_L1
CYCLE: 75 Seconds, 38 Steps
TERMINATION CARD
90 000000000000
-92-NOTE -
+ END OF JOB!
1
Page 8
0 0 0
87


Appendix B: Sample Output File for TUDI TRANSYT-7F
88


******************************************************************************
* Release 7.2
(TRANSYT-7F)
TRAFFIC SIGNAL SYSTEM OPTIMIZATION
February 1994
PROGRAM
* Sponsored by:
*
* U.S. Department of Transportation
* Federal Highway Administration
Developed by:
University of Florida
Transportation Research Center
* Software Maintenance and User Support Furnished by: 4
* Center for Microcomputers in Transportation (McTrans) 4
* Transportation Research Center, University of Florida *
* 512 Weil Hall, Gainesville, FL 32611-6585 USA 4
* (904) 392-0378 *
* *
* TRANSYT/7 (C) British Crown Copyright. 4
* TRANSYT-7F Copyright 1980-1994, University of Florida. 4
* All Rights Reserved. 4
* <
******************************************************************************
Date of Run: 9/11/99 Start Time of Run: 10:31:36 Data File: TUD_L1.TIN
INPUT DATA REPORT FOR RUN 1
FIELDS:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
TUD_L1
1 60 180 15 3 20 2 3 20 1 1 60 0 0 0 0
-- 5--NOTE -
+ THE SEC/STEPS FACTOR IN FIELD 6 IS
INTERPRETED AS 2.000.
>>> 106 +++ WARNING +
+ THE SEC/STEPS FACTOR IN FIELD 6 IS TOO SMALL FOR CYCLE
LENGTHS ABOVE 120 SECONDS. IT WILL BE INCREASED TO
ALLOW A MAXIMUM OF 60 STEPS/CYCLE.
24 50000000000000
1TRANSYT-7F: PAGE 2
TUD_L1
FIELDS:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
3 12 11 10 4 5 6 3 2 1 9 8 7 0 0 0
10 4 3 2 1800 30 50 0 20 0 0 0 0 0 0 0
INTERSECTION 4
13 4 0 1 16 3 2 16 3 2 16 3 2 0 0 0
21 4 1 1 2 3 18 408 409 0 0 0 0 0 0 0
22 4 4 4 5 6 18 408 401 402 0 0 0 0 0 0
89