A Comparison of Non-Traditional Intersection Designs Using Microscopic Simulation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

A COMPARISON OF NON-TRADITIONAL INTERSECTION DESIGNS USING

MICROSCOPIC SIMULATION

Evangelos I. Kaisar, Assistant Professor, Civil Environmental and Geomatics Department,


Florida Atlantic University, Boca Raton, FL 33431, Tel: (561)297-4084, Fax 561-297 0493,
Email: ekaisar@fau.edu

Praveen Edara*, Assistant Professor, Department of Civil Engineering, University of Missouri-


Columbia, M0 65211, Tel: (573) 882-1900, Fax (573) 882-4784, Email: edarap@missouri.edu

Jarice D. Rodriguez-Seda, Instructor, Civil Environmental and Geomatics Department, Florida


Atlantic University, Boca Raton, FL 33431, Tel: Tel: 561-297 4084, Fax (561) 297-3936, Email:
jrodrigu@fau.edu

Steve Chery, Research Assistant, Civil Environmental and Geomatics Department, Florida
Atlantic University, Boca Raton, FL 33431, Email: schery3@fau.edu

Total number of words: 7410 (5410 text plus 8 figures and tables)

Revised paper submitted on November 15, 2010

* Corresponding author

TRB 2011 Annual Meeting Paper revised from original submittal.


ABSTRACT
In light of the growing traffic demand and the futility of the conventional solutions, many states
have been considering alternative intersection designs. Researchers have demonstrated the
benefits of several unconventional intersection designs and their implementation at different sites
throughout the United States and abroad have delivered significant improvement in traffic
performance compared to traditional designs. A signalized and un-signalized roundabout, a
Continuous Flow Intersection (CFI), and a Parallel Flow Intersection (PFI) have been evaluated
and compared in this research as viable alternatives to the traditional four-legged intersection.
Using traffic simulation platforms, AIMSUN and VISSIM, models of each intersection are
evaluated for low, medium, and high entrance volumes. For the balanced volume scenarios
studied in this paper, the analysis revealed that the Roundabout performs better at low entering
volumes while the CFI yields better results at high volumes. Statistical analysis of the results
indicated that it was not possible to statistically differentiate between the average delay time
estimates obtained from AIMSUN and VISSIM within 95% confidence level.

ii

TRB 2011 Annual Meeting Paper revised from original submittal.


1 INTRODUCTION
2 Recurring traffic congestion at roadway intersections has become a major concern in recent
3 years. Intersections, by their nature, present focal bottlenecks and vast conflict areas. Around the
4 world, various alternatives to conventional intersections are being considered to solve congestion
5 problems at intersections. The level of service (LOS) of intersections can be improved with the
6 application of unconventional intersection designs such as: Roundabout, Continuous Flow
7 Intersection (CFI) and Parallel Flow Intersection (PFI).
8 Roundabouts were developed in the United Kingdom in the 1960s to address issues of safety
9 and efficiency related to early uses of traffic circles. Roundabouts have been shown to be a safe
10 and effective alternative to designers and traffic engineers considering improvements to a
11 conventional intersection (1). The Federal Highway Administration (FHWA) recommends the
12 use of roundabouts as a way to improve intersection performance citing benefits including: 1)
13 less severe crashes than other intersection crashes; 2) increase in traffic capacity; and 3) increase
14 aesthetics values (2).
15 Other types of innovative intersection designs such as: CFI, PFI among others have also
16 been shown to improve both operational and safety performance issues associated with
17 conventional intersections. The benefits provided by these novel designs are attained mostly by
18 rerouting and prohibiting left turning movement at the main junction of the intersection. The
19 CFI eliminates left-turn conflicts with opposing through movement by displacing the left turn
20 lane to the left side of the road at a distance upstream of the main intersection. Many researchers
21 have studied and compared a variety of these designs. Although each individual intersection
22 offers particular strengths and weaknesses, the CFI has been shown to produce results that, not
23 only surpass other designs but also comparable to grade separation. In 2006 Parsons developed
24 a PFI design which reroutes left turns (3). Past studies have demonstrated that the PFI yields
25 results that considerably improve traffic from conventional intersections and at the same time,
26 are similar to that of the CFI (4, 5, 6). Although conventional intersections are an important
27 element of the surface transportation network, oftentimes they are unable to accommodate heavy
28 traffic flow and heavy left turn movement, which may result in high delays and accidents.
29 Hence, the uses of unconventional intersection designs have become an essential tool to reduce
30 congestion and improve safety.
31 The main objective of this research is to evaluate the operational performance of four
32 unconventional intersection schemes: 1) un-signalized roundabout, 2) signalized roundabout, 3)
33 CFI, and 4) PFI using two commercially available computer traffic simulation models.
34
35 LITERATURE REVIEW
36 Unconventional Intersection Designs
37 There are several published studies about the geometric characteristics and operational
38 benefits of a roundabout. In 2001, Thorson et al. (7) compared a roundabout operation to four-
39 way stop and signalized intersections. The microscopic simulation platform NETSIM was used
40 to model the three alternatives. Johnson and Hange in 2002 (8) examined the benefits of
41 roundabouts as an alternative to traffic signals under various conditions. The study aimed to
42 provide traffic engineers the understanding of how the differences in roundabout and traffic
43 signal operation can be used to achieve safe and efficient traffic operation, reduce congestion,
44 and lessen the negative impacts often associated with roadway and highway projects. As traffic
45 engineers and researchers seek out innovations that enhance traffic flow, ultimately leading to
46 alleviation of congestion, Bared and Kaisar (9) investigated the roundabout as an alternative to

3
TRB 2011 Annual Meeting Paper revised from original submittal.
1 diamond interchanges. They compared the delay caused by a diamond interchange with the
2 delays at interchanges containing double or single roundabouts. Data generated from computer-
3 simulated models for roundabout traffic operations and for signalized intersections were used as
4 the basis for the comparisons. Bared and Edara (10) scrutinized the simulated capacity of
5 roundabouts and the impact of roundabouts within a progressed signalized road. Traffic
6 simulation was used to study the performance of urban single-lane and double-lane roundabouts
7 in isolation. The impact of signalized intersection proximity to roundabouts was also studied by
8 inserting a roundabout within an arterial corridor.
9 Bernetti et al. (11) compared the operational performances of un-signalized and signalized
10 roundabouts under critical traffic conditions in Italy. Capacity and delays were the two
11 performance measures used as the basis of the comparison between the two alternative
12 intersection designs. Goldblatt et al. (12) compared traffic performance at a CFI to that of
13 conventional intersections. Reid and Hummer (13) compared travel time between seven
14 unconventional arterial intersection designs – the quadrant roadway intersection, median U-turn,
15 superstreet median, bowtie, jughandle, split intersection, and CFI designs. Jagannathan and
16 Bared (14) assessed the design and operational performance of three different configurations of
17 the CFI intersection design (also known as crossover displaced left-turn intersection).
18 The parallel flow or paraflow intersection (PFI) was introduced as a likely solution to satisfy
19 the aforementioned objectives at many intersection locations where improvements are needed
20 (4). Cheong et al. (5) compared the performance of a PFI, a CFI, and an upstream-signalized
21 crossover. Dhatrak et al. (6) conducted an extensive performance analysis of the PFI and CFI
22 under various traffic conditions for both full and partial designs. Using traffic simulation they
23 compared the operational performance of the designs based on the maximum through and left-
24 turn movement throughputs.
25
26 Computer Simulation Models
27 The two computer traffic simulation packages selected for this study were the stochastic and
28 microscopic simulation models VISSIM and AIMSUN. These models are commonly used for
29 traffic simulation studies and have the capability of modeling different intersection designs,
30 arterial and freeway operations, with a wide variety of highway geometric characteristics, traffic
31 control and signalization functions. AIMSUN was developed by Transport Simulation Systems
32 (TSS) in Spain. VISSIM was developed by PTV AG in Germany. It employs links and
33 connectors to form its network, rather than nodes and links (like AIMSUN).
34 Fang and Elefteriadou (16) studied and compared three widely used traffic simulation
35 packages: AIMSUN, CORSIM and VISSIM. The two intersection designs selected as case
36 studies were a diamond interchange and a Single Point Urban Interchange (SPUI) located in
37 Arizona. The models were calibrated to adjust parameters related to driver behavior and vehicle
38 performance.
39 Jones et al. (17) compared CORSIM, which according to the authors is “the most widely
40 used microsimulation program in the US,” SimTraffic, and AIMSUN. In 2005 Xiao et al. stated
41 that AIMSUN and VISSIM are “two of the best-regarded simulators” available on the market
42 (18). Both models have the capabilities of modeling complex networks as well as various types
43 of interchanges and are therefore used in this research to evaluate and compare the performance
44 of the selected intersection designs.
45

4
TRB 2011 Annual Meeting Paper revised from original submittal.
1 METHODOLOGY
2 This section describes the required input data for the simulation analysis including, the
3 geometric characteristics of each of the intersections, the traffic demand data, and the signal
4 timing optimization procedure for each intersection design. Traffic and geometric data was used
5 as input into the computer simulation software, signal timing was optimized and simulation
6 studies were conducted. Once performance measures of effectiveness were extracted from the
7 simulation studies, a statistical analysis was performed to establish whether the differences
8 between computer simulation packages were statistically significant. Figure 1 illustrates the
9 methodology used in this research.
10

11
12
13 FIGURE 1 Methodology flowchart
14
15 The measures of effectiveness (MOE) used in this research were those recommended in
16 literature. Also, the selected MOE needed to be similarly computed by the two traffic simulation
17 models. The primary MOE used in this research is average control delay. This parameter is the
18 criteria recommended by the HCM2000 methodology to determine the LOS of an intersection
19 (19). Fang and Elefteriadou (16) used delay to compare the simulation performance. In addition
20 to average delay, the other parameter used as MOE was the number of stops. The number of
21 stops is the number of times, on average, a driver has to stop while travelling through each type
22 of intersection to complete their movement.
23 Signalized and un-signalized roundabout, CFI and PFI intersection were modeled in
24 VISSIM and AIMSUN, with identical geometric dimensions in the length of the approaches and
25 turning lanes (right and left turns), and the number of through lanes. A four-lane roadway with
26 two lanes in each direction was studied. For the CFI and the PFI, the distance between the main
27 junction and the sub-intersections had to be determined. The capacity of the intersection
28 depends on that distance since it also affects the capacity of storing the left-turn traffic. Many

5
TRB 2011 Annual Meeting Paper revised from original submittal.
1 researchers including Jagannathan and Bared have suggested that the distance between the main
2 and the sub-intersections should be within 300-500ft long in order to be able to operate
3 efficiently (5). The distance between the main and sub-intersections is set at 325ft. In addition,
4 the right-turning pockets are set at 250ft and the deceleration lanes are set at 325ft. Figure 2
5 presents the geometric elements of the PFI. Figure 3a-3b present the screenshots of the CFI and
6 PFI models in AIMSUN and VISSIM. A conventional intersection with geometrics comparable
7 to these designs in terms of number of through, left turn, and right turn lanes, length of storage
8 for left and right turn bays was also simulated for comparisons.
9

10
11
12 FIGURE 2 Geometric elements of Parallel Flow Intersection (PFI).
13
14

6
TRB 2011 Annual Meeting Paper revised from original submittal.
1
2 (a) Parallel Flow Intersection (PFI)models

3
4 (b) Continuous Flow Intersection (CFI) models

5
6 (c) Roundabout models
7
8 FIGURE 3 Screenshots of the (a) PFI, (b) CFI and (c) roundabouts models in AIMSUN
9 (left) and VISSIM (right).
10

7
TRB 2011 Annual Meeting Paper revised from original submittal.
1 As for the dual-lane roundabout networks, the geometric features, as recommended by the
2 FHWA’s roundabout guide (20) were used in order to achieve safe and efficient operation. The
3 inscribed circle diameter was 180ft; the entry and exit radii were approximately 131ft; the
4 approach and departure width were set at approximately 24ft; and the circulatory road width was
5 31ft. Figure 3c presents the screenshots of the roundabout models in AIMSUN and VISSIM.
6 Each intersection was evaluated under different volume levels ranging from low to high in
7 order to simulate both peak and off-peak conditions. Although, both balanced and unbalanced
8 flow conditions were tried only the results of balanced flows are shown in this paper. Under a
9 balanced flow condition, the volumes on all four approaches were similar whereas an unbalanced
10 volume condition considers a major and minor road at the intersection.
11 To emulate peak and off-peak traffic data that can be observed in the field, volumes ranging
12 from 1,000 vph- 6,000 vph were tested for each intersection designs. These volumes were
13 chosen because typically in the field, 1,000 vph represent low entering flow at an intersection
14 during off-peak period whereas 6,000 vph represent high flow situation where conventional
15 intersections are performing close to saturation. A constant left-turning traffic of 15% (of the
16 through traffic) was assumed on all four approaches. These volumes were generated based on
17 field trends and works of previous researchers. Bared et al. (10) used similar volumes with low
18 (1,450 vph), medium (2,890 vph), high (5,410 vph), and peak (5,752 vph) entering volumes to
19 compare the operational performance of several unconventional intersections. Ten different
20 volume scenarios were generated at an increment of 500 vph (see Table 1).
21
22 TABLE 1 Traffic volume scenarios studied
23
Movement Flow Rate
Traffic
Eastbound Westbound Northbound Southbound
Volumes
(vph) (vph) (vph) (vph)
(vph)
L T R L T R L T R L T R
1,000 30 195 25 30 195 25 30 195 25 30 195 25
1,500 44 293 38 44 293 38 44 293 38 44 293 38
2,000 59 391 50 59 391 50 59 391 50 59 391 50
2,500 76 502 63 76 502 63 76 502 63 76 502
3,000 89 586 75 89 586 75 89 586 75 89 586 75
3,500 103 684 88 103 684 88 103 684 88 103 684 88
4,000 118 782 100 118 782 100 118 782 100 118 782 100
4,500 133 879 113 133 879 113 133 879 113 133 879 113
5,000 148 977 125 148 977 125 148 977 125 148 977 125
6,000 177 1173 150 177 1173 150 177 1173 150 177 1173 150
24
25
26
27

8
TRB 2011 Annual Meeting Paper revised from original submittal.
1 Although signal timing (cycle length and phase split) optimization was not the objective of
2 this study, it was necessary to determine an optimum signalization plan to evaluate the
3 effectiveness of intersection configuration. For a single intersection the Webster’s formula (21)
4 was used to minimize the average delay which has been the basis of the delay minimization
5 problem for isolated intersections. More complicated formulations have been developed but it’s
6 been proven that their results when compared to those of Webster’s are only marginally better.
7 In some cases, where only the entering flow data were considered, Webster’s formulation
8 yielded results comparable to observed delay in the field (22). Therefore, the Webster’s formula
9 was considered appropriate for the purpose of this study.
10 After obtaining all the geometrical and phase assignment inputs from the intersection, the
11 model was applied and the problem was solved using Excel solver. The following values were
12 used for the parameters in the solver – a maximum time of 100 seconds, 100 iterations, a
13 precision of 0.000001, a 5% tolerance, and a convergence value of 0.0001. The solver was also
14 set to assume non-negative numbers for all values.
15 In order to validate the results found from the Excel solver, the signal timing for the same
16 intersection used in the case study was optimized using PASSERII–02 and TRANSYT-7F.
17 PASSERII-02 minimizes delay for an intersection using the delay formulation from the
18 HCM2000. PASSER II uses a bandwidth optimization technique, which is a numerical search
19 technique known as Interference Minimization. The reader is referred to Messer et al. (23), for a
20 complete description of PASSERII optimization procedure. To search for the optimum solution,
21 TRANSYT-7F uses a genetic algorithm (24).
22 Once the required data were inputted in the programs, they searched for an optimal solution
23 and generated the optimum cycle length and the optimum splits. The results obtained from
24 TRANSYT-7F were discarded because the program generated solutions consisting of a
25 minimum of 6 phases for every attempt. This might be because the program was designed with
26 the purpose of mostly optimizing networks of intersection rather than isolated intersections. The
27 results obtained from PASSERII as well as those from Excel Solver were inputted into an
28 AIMSUN model of the case study intersection. The results showed that the signal settings from
29 PASSERII provided lower delays than the solver-based model. Moreover, the delay values
30 obtained from using the signal control plans generated by the model were comparable to that of
31 PASSERII. However, because the problem formulated was non-linear, Excel Solver failed to
32 provide a solution on several occasions. Therefore, PASSERII–02 was used to optimize the
33 signal timings for the conventional intersection design and the signalized roundabout.
34 The signal timing plan used by Jagannathan and Bared (14) was applied to the CFI
35 evaluated in this study (see Figure 4). As the PFI design also operates with two phases similar to
36 the CFI, the signal timing plan selected for the CFI was implemented for the PFI as well.
37

9
TRB 2011 Annual Meeting Paper revised from original submittal.
1
2 FIGURE 4 Signal timing control settings for the CFI (Jagannathan and Bared, 2004)
3
4 RESULTS
5 Using the two computer microsimulation models, AIMSUN and VISSIM, the
6 aforementioned ten balanced traffic volume scenarios were tested for each of the intersections.
7 Ten replications with different seed numbers were used, and the results were averaged, to
8 account for the randomness of simulations. Two primary MOEs: average delay time and number
9 of stops, were used to assess the operational performance of the intersections. In addition, a
10 sensitivity analysis for the average delay times obtained is conducted to determine which
11 unconventional intersection design is an appropriate alternative to the conventional intersection
12 at different volume levels. The simulation results for each intersection design were compared to
13 establish the similarities and differences between the two simulation models. Finally, a
14 statistical analysis was conducted to compare the results obtained from the two models.
15 Furthermore, a general comparison of the two models with regards to network-coding process,
16 simulation processing time, and data output was provided to emphasize some of the key
17 differences between models.
18 The average delay for the conventional intersection gradually increased by approximately
19 one second for every 500 vph increment for a total entering volume of 1,000 vph to 3,500 vph.
20 As the total entering volume further increased, the average delay per vehicle drastically
21 increased. At the highest tested volume (6,000 vph), an average delay of approximately
22 53seconds was experienced per vehicle. This was consistent with the hypothesis that the
23 conventional intersection does not perform well at volume levels close to saturation conditions.
24 According to the criteria established by the HCM2000, from 1,000 vph to 3,500 vph, the
25 conventional intersection performed at LOS C. While a LOS D was achieved from 4,000 vph to
26 4,500 vph, the unconventional intersection performed at LOS F for flows higher than 5,000 vph.

10
TRB 2011 Annual Meeting Paper revised from original submittal.
1 Vehicles in the un-signalized roundabout experienced the lowest delay compared to the
2 other intersections evaluated in this research at the low volume levels with average delay time of
3 less than 10 seconds/vehicle up to 2,000 vph. The results showed that introducing signals at the
4 roundabout is beneficial and effective especially at higher volume level when the roundabout
5 capacity is exceeded. Up to 2,500 vph, both the un-signalized roundabout and the signalized
6 roundabout performed at a LOS A, although the un-signalized roundabout produced slightly
7 lower delays at these volume levels. However, beyond 2,500 vph, introducing the signals
8 significantly reduced the average delay time at the roundabout.
9 Compared to the conventional intersection, the PFI performed better in terms of lower
10 delays for most of the volume scenarios. For an entering volume of 1,000 vph, the PFI resulted
11 in a LOS B while the conventional intersection had a LOS C. For entering volume between
12 1,000 vph and 4,500 vph, the average delay times per vehicle produced by the PFI were similar
13 to the conventional intersection. For volumes greater than 4,500 vph the PFI outperformed the
14 conventional intersection.
15 The CFI, on the other hand, had the best overall performance compared to the other
16 intersection designs evaluated in this research. At the low entering flow levels (1,000 vph to
17 2,500 vph), the CFI produced average delay times that were similar to the conventional
18 intersection. However, at the high flow levels, the CFI produced lower average delays than all
19 the other alternatives; with the conventional intersection and the roundabout performing at a
20 LOS F, the PFI at LOS E while the CFI had a LOS C. Figures 5a-5b graphically illustrate the
21 average delay time trends of each intersection designs for 15% left-turn volume.
22 For the low traffic volume levels (1,000 vph to 2,000 vph) the un-signalized roundabout
23 showed smaller average delays than the other alternatives. However, as the volume increased,
24 the average delay for the un-signalized roundabout rapidly increased. The signalized roundabout
25 performed better than the un-signalized roundabout at higher volume levels but the average
26 delays for the signalized roundabout were considerably higher than the other intersections. At
27 traffic volumes higher than 2,000 vph, the CFI showed lower average delay than the
28 conventional intersection as well as the other unconventional intersection designs. The PFI
29 performance was similar to the CFI until the entering volume reached 5,000 vph, beyond which
30 the average delays found in the CFI were lower than the PFI. For example at 5,000 vph, the
31 average delay in CFI was 28 seconds compared to 37 seconds in the PFI.
32

11
TRB 2011 Annual Meeting Paper revised from original submittal.
Conventional

Roundabout
400
Ave. Delay Time (s/veh) 400.0000 Signalized
350 Roundabout
350.0000 CFI
300
300.0000 PFI
250.0000
250

200.0000
200
150.0000
150
100.0000
100
50.0000
50
0.0000
1000
1,000 2000
2,000 3000
3,000 4000
4,000 5000
5,000 6000
6,000
TotalEntering
Total Enetring Volume
Volume(vph)
(vph)
1
2 (a)VISSIM model
3
Conventional

Roundabout
400.0000
400 Signalized
Roundabout
350.0000
350 CFI
Ave. Delay Time (s/veh)

300.0000
300 PFI
250.0000
250

200.0000
200

150.0000
150

100.0000
100

50.0000
50

0.0000
0
1000
1,000 2000
2,000 3000
3,000 4000
4,000 5000
5,000 6000
6,000
TotalTotal Enetring
Entering Volume
Volume (vph)
(vph)
4
5 (b)AIMSUN model
6 FIGURE 5 Average delay time for different designs: (a) VISSIM5.10 and (b) AIMSUN6.0
7
8 To determine the appropriate alternative for each of the volume level considered: low (1,000
9 vph to 2,500 vph), medium (3,000 vph to 4,000 vph), and high (4,500 vph to 6,000 vph), a
10 sensitivity analysis was conducted. There are two major classes of sensitivity functions: analytic
11 and empirical. Analytic sensitivity functions are used for well-defined systems usually using
12 partial derivatives whereas empirical sensitivity functions observe system changes due to
13 changes in certain parameters. The latter are used for un-modeled systems (25).
14 Bloomberg and Dale (26) performed a sensitivity analysis to compare the relative travel
15 times for seven different routes on a congested network as predicted by CORSIM and VISSIM.
16 After calibrating and validating the models and ensuring that the models produce results that are
17 comparable to field observations, results for two sets of runs were obtained where one set used
18 traffic demands 10% lower than the base case and the other used 10% higher than the base case.

12
TRB 2011 Annual Meeting Paper revised from original submittal.
1 This sensitivity analysis established whether the relative comparisons between the scenarios
2 were consistent even with higher or lower traffic demands. Then, a system was developed to
3 compare the travel times for the seven alternatives where the system travel time values were
4 scaled to 100. For instance, if the least travel time for a scenario is 30 minutes for an alternative,
5 that alternative is given a score of 100. If another alternative produced a travel time of 40
6 minutes for the same scenario, it would be given a score of 100x30/40=75. This procedure is
7 repeated for all alternatives and all scenarios. The alternative with the maximum score at the end
8 is the best alternative.
9 A similar scoring method was employed in this research for each aforementioned volume
10 scenarios. For each of the entering volumes, the alternative with the lowest average delay time is
11 awarded a score of 100, and then each of the other alternatives is awarded a relative score as
12 described above. A total score is determined for each of the flow level separately so that the best
13 alternative for each level can be determined. Then, these values were added in order to
14 determine the overall best alternative among the intersection designs. Table 2 presents the results
15 of the sensitivity analysis for the AIMSUN and VISSIM models.
16 The sensitivity analysis results were consistent in suggesting that for the low volume level,
17 the roundabout was the best alternative while the CFI was the best alternative for the medium
18 and high volume levels. Moreover, the CFI was considered the best overall alternative compared
19 to the other intersections.
20 To establish whether the differences between the results from the two microsimulation
21 platforms were significant, statistical analyses were conducted. A paired t-test was applied to
22 compare the results from the two simulation platforms. This statistical test uses independent
23 samples to make inferences about population means. For the purpose of this research, values
24 obtained from the two models for one of the intersections are selected for the statistical analysis.
25 The paired t-test was designed in this case to test whether the two microscopic platforms report
26 significantly different average delay time values for the same intersection. A 95% confidence
27 level (α=0.05) was chosen and the average delay time values obtained from the two platforms for
28 the CFI were selected for statistical analysis. The results of statistical analysis are shown in Table
29 3.
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

13
TRB 2011 Annual Meeting Paper revised from original submittal.
1 TABLE 2 Sensitivity Analysis of Designs in AIMSUN and VISSIM
2
3 AIMSUN

4
5
6 VISSIM

7
8
9

14
TRB 2011 Annual Meeting Paper revised from original submittal.
1 TABLE 3 Paired t-test Results
2
Computer Simulation Software 3
Statistics
AIMSUN (sample 1) VISSIM (sample 2)4
Sample size 10 10 5
Sample mean (delay, s) 22.199 22.228 6
7
Sample standard deviation 3.2388 1.9449
8
Pooled Variance(Sp) 2.67
9
|t| 0.025 10
t α /2 2.262 11
12
13 Since tpaired was equal to 0.025085 and less than t α / 2 (2.262), it was concluded that the population
14 means were not significantly different for a 95% confidence level. The result of the t-test
15 indicated that, for this CFI design, it was not possible to statistically distinguish between the
16 average delay time values obtained from AIMSUN and those from VISSIM in 95 out of 100
17 trials.
18
19 CONCLUSIONS
20 This research compared the operational performance of four unconventional intersections:
21 roundabout, signalized roundabout, CFI, and PFI. The intersection designs were compared to a
22 conventional intersection to determine which intersection can be used as an alternative when
23 considering improvements to a conventional intersection.
24 For the balanced flow scenarios studied in this paper, the simulation results showed that the
25 roundabout yielded the lowest average delay time among the intersection designs for low
26 entering volumes. The roundabout produced average delay times less than 10 seconds/vehicle
27 for entering volumes of less than 2,500 vph, LOS A according to the HCM2000. At higher
28 volumes however, the roundabout quickly reached its capacity and the average delay times
29 rapidly increased with increase in entering volumes. Introducing signals at the roundabout
30 resulted in lower delays for flows higher than 2,500 vph. Although the signalized roundabout
31 had lower average delays than the un-signalized roundabout at higher flows, the average delays
32 per vehicle were higher than the CFI and PFI designs. The PFI showed average delays that were
33 lower than the conventional intersections for most of the cases and similar or lower than the CFI
34 at low volumes. For the low and medium entering volumes, the PFI had a LOS C while
35 performing at a LOS E for the highest tested volume (6,000 vph). The CFI produced the lowest
36 delays at the medium and high entering volumes.
37 The result of statistical analysis indicated that it was not possible to statistically differentiate
38 between the average delay time values obtained from AIMSUN and those from VISSIM within
39 95% confidence level. This means that the performance of designs is independent of the chosen
40 simulation platform.
41 A sensitivity analysis was performed to determine the appropriate alternative for each of the
42 volume level considered: low (1,000 vph-2,500 vph), medium (3,000 vph-4,000 vph), and high
43 (4,500-6,000 vph). From the sensitivity analysis, it can be concluded that the roundabout would
44 be the best alternative, to conventional design, for locations where traffic volume is less than
45 3,000 vph. Implementing signals at the roundabout can provide the benefits of reduced delay at
46 medium volume and still perform better than the conventional intersection. For locations where

15
TRB 2011 Annual Meeting Paper revised from original submittal.
1 the traffic volume is high, the CFI turned out to be the best alternative. Overall, the CFI was
2 considered the best alternative compared to the other intersections evaluated in this research with
3 at least a LOS C performance for all the volume scenarios tested. It is noted that these findings
4 are applicable to the balanced flow scenarios studied in this paper. Additional traffic scenarios
5 considering variations of type of turning movements should be researched to be able to
6 generalize the findings of this study.
7 The focus of this paper was to compare traffic performance of different intersection designs.
8 In reality, the selection of a design entails several other important considerations such as right-
9 of-way requirements, construction and maintenance costs, safety performance, environmental
10 and social impacts, accommodation of pedestrians and bicycles, adjacent access requirements.
11
12 REFERENCES
13
14 1. National Cooperative Research Highway Program. Report 572. Project 3-65: “Applying
15 Roundabouts in the United States”, Transportation Research Board, TRB, National
16 Research Council, Washington, D.C., 2007.
17 2. US Department of Transportation. “Roundabouts: A Safer Choice.” FHWA-SA-08-006.
18 http://safety.fhwa.dot.gov/intersection/roundabouts/fhwasa08006/fhwasa08006.pdf
19 3. Quadrant Engineering. “The Paraflow: An Innovative New Traffic Intersection.
20 Introduction to the parallel flow intersection.” July 2009. http://www.paraflow.info/
21 4. Parsons, G. F. "The Parallel Flow Intersection: A New Two-Phase Signal Alternative,".
22 Institute of Transportation Engineers. ITE Journal.
23 FindArticles.com.http://findarticles.com/p/articles/mi_qa3734/is_200710/ai_n21100072
24 Accessed September 26, 2009.
25 5. Cheong, S., Rahwanji, S., and Chang, G. “Comparison of Three Unconventional Arterial
26 Intersection Designs: Continuous Flow Intersection, Parallel Flow Intersection, and
27 Upstream Signalized Crossover.” 11th International IEEE Conference on Intelligent
28 Transportation Systems, Beijing, China, 2008.
29 6. Dhatrak, A., Edara, P., and Bared, J. Performance Analysis of Parallel-Flow Intersection
30 and Displaced Left-Turn Intersection Designs in: Transportation Research Record:
31 Journal of the Transportation Research Board, 2171, pp. 33-43, 2010.
32 7. Thorson , S.L., Campbell, D.D., Gross, P.D. Comparison of Roundabout Operations to
33 Four Way Stop and Signal Controlled Intersections Using Netsim Simulations. Nevada
34 DOT Research Report RDT 01-008. 2001.
35 8. Johnson, M.T., and Hange, W.A. “Modern Roundabout Intersections: When to Use
36 Them? A Comparison with Signalized Intersections.” ITE Journal, 2002.
37 www.ksu.edu/roundabouts/news/ITEPaper.pdf
38 9. Bared, G. J., and Kaisar, E. I. “Does Your Interchange Design Have You Going Around
39 in Circles?” Public Roads, November/December 2002, Vol. 66, No. 3
40 10. Bared, J., Edara, P., and Jagannathan, R. Design and operational performance of Double
41 Crossover Intersection and Diverging Diamond Interchange. Transportation Research
42 Record, TRR No. 1912. Transportation Research Board, National Research Council,
43 Washington, D.C., 2005.
44 11. Bernetti, G., Dall’Acqua, M., and Longo, G. “Unsignalized vs. Signalized Roundabouts
45 under Critical Traffic Conditions: a Quantitative Comparison.” Association for European
46 Transport, 2003

16
TRB 2011 Annual Meeting Paper revised from original submittal.
1 12. Goldblatt, R., Mier, F. and Friedman, J. "Continuous Flow Intersections." Institute of
2 Transportation Engineers. ITE Journal, July 1994, pp. 35-42.
3 13. Reid, J.D., and Hummer, J.E. “Travel Time Comparisons between Seven Unconventional
4 Arterial Intersection Designs.” TRR 1751: Transportation Research Board. National
5 Research Council, Washington, D.C., 2001.
6 14. Jagannathan, R., and J.G. Bared. Design and Operational Performance of Crossover
7 Displaced Left-Turn Intersections. In Transportation Research Record: Journal of the
8 Transportation Research Board, No. 1881, Transportation Research Board of the
9 National Academies, Washington, D.C., 2004, pp. 1–10.
10 15. El Esawey, M. and Sayed, T. (2007). Comparison of Two Unconventional Intersection
11 Schemes: Crossover Displaced Left Turn and Upstream Signalized Crossover. In
12 Transportation Research Record: Journal of the Transportation Research Board, No.
13 2023, Transportation Research Board of the National Academies, Washington, D.C., pp.
14 10-19.
15 16. Fang, C. F., and Elefteriadou, L. “Some Guidelines for Selecting Microsimulation
16 Models for Interchange Traffic Operational Analysis.” ASCE Journal of Transportation
17 Engineering, Vol. 131, No. 7, pp. 535-543, 2005.
18 17. Jones, L. S., Sullivan, A. J., Cheekoti, N., Anderson, D. M., Malave, D. “Traffic
19 Simulation Software Comparison Study.” University Transportation Center for Alabama,
20 Report 02217. June 2004.
21 18. Xiao, H., Ambadipudi, R., Hourdakis, J., Michalopoulos, P. “Methodology for Selecting
22 Microscopic Simulators: Comparative Evaluation of AIMSUN and VISSIM.” Intelligent
23 Transportation Systems Institute, University of Minnesota. May 2005
24 19. Special Report 209: Highway Capacity Manual, 3rd ed. TRB, National Research Council,
25 Washington, D.C., 2000.
26 20. Robinson, B.W., L. Rodegerdts, W. Scarbrough, W. Kittelson, R. Troutbeck,
27 W. Brilon, L. Bondzio, K. Courage, M. Kyte, J. Mason, A. Flannery, E. Myers, J.
28 “Roundabouts: An Informational Guide,” U.S. Department of Transportation, Federal
29 Highway Administration, Publication No. FHWA-RD-00-067, June 2000.
30 21. Webster F. V. “Traffic signal settings.” Technical Report 39, Road Research Laboratory,
31 1958.
32 22. Dion, F., Rakha, H., and Kang , Y. “Comparison of Delay Estimates at Under-saturated
33 and Over saturated Pre-timed signalized Intersections”. Transportation Research Part B,
34 Volume 38, Issue 2, pp. 99-122, 2004.
35 23. Messer, C. J., R. H. Whitson, C. L. Dudek, C. L., and Romano, E.J. “A Variable
36 Sequence Multiphase Progression Optimization Program.” In Highway Research Record
37 445, Highway Research Board, National Research Council, Washington, DC, 1973, pp.
38 24-33.
39 24. Transportation Research Center, TRANSYT-7F User’s Guide, March, 1998.
40 Gainesville, Florida.
41 25. Bahill, A. T., Smith, D. E., Szidarovszky, F., and Karnavas, J. W. “Sensitivity Analysis, a
42 Powerful System Validation Technique.” The Open Cybernetics and Systemics Journal,
43 2008, 2, 39-56.
44 26. Bloomberg, L., and Dale, J. “A Comparison of the VISSIM and CORSIM Traffic
45 Simulation Models On A Congested Network” Transportation Research Record, 1727,
46 pp. 52-60, 2000.

17
TRB 2011 Annual Meeting Paper revised from original submittal.

You might also like