1

Hu, Tian, Xu, and Andalibian

1
2
3 AN ADVANCED SIGNAL PHASING SCHEME FOR DIVERGING DIAMOND
4 INTERCHANGES
5
6
7 Peifeng Hu, Ph.D.
8 (Corresponding Author)
9 Department of Civil & Environmental Engineering
10 University of Nevada, Reno
11 Reno, NV, USA 89557
12 Email: hupeifeng1981@gmail.com
13 Tel: (775) 784-1232
14 Fax: (775) 784-1390
15
16 Zong Z. Tian, Ph.D., P.E.
17 Department of Civil & Environmental Engineering
18 University of Nevada, Reno
19 Reno, NV, USA 89557
20 Email: zongt@unr.edu
21 Tel: (775) 784-1232
22 Fax: (775) 784-1390
23
24 Hao Xu, Ph.D.
25 Department of Civil & Environmental Engineering
26 University of Nevada, Reno
27 Reno, NV, USA 89557
28 Email: haox@unr.edu
29 Tel: (775) 784-6909
30 Fax: (775) 784-1390
31
32 Rasool Andalibian, Ph.D. Candidate
33 Department of Civil & Environmental Engineering
34 University of Nevada, Reno
35 Reno, NV, USA 89557
36 Email: randalibian@gmail.com
37 Tel: (775) 784-6195
38 Fax: (775) 784-1390
39
40
41 Nov 2013
42

TRB 2014 Annual Meeting Paper revised from original submittal.

 2

Hu, Tian, Xu, and Andalibian

43 ABSTRACT
44 A diverging diamond interchange (DDI), also called a double crossover diamond interchange
45 (DCD), is a relatively new interchange type in the U.S. Due to its many advantages over the
46 conventional diamond interchanges such as higher capacity and lower delay, there has been a
47 growing interest in its analysis methodologies and operational strategies. DDIs have shown
48 particularly high efficiency under conditions with heavy ramp traffic to arterial roads. However,
49 as the DDI concept is relatively new in the U.S., the research on DDIs is still considered to be in
50 the preliminary stage. In particular, most studies found in the literature only provided basic
51 phasing schemes, which were not optimized for efficiency. Furthermore, there is a lack of
52 commonly acceptable methodologies for obtaining critical signal control parameters, such as the
53 cycle length, and phasing splits. Therefore, it is an area topic in need of much research in order
54 to develop signal phasing schemes for achieving the optimized DDI performance. This study
55 proposed a new methodology for signal phasing scheme design at DDIs. With minor adjustment,
56 it can be applied to a variety of DDIs with different traffic demands and geometric
57 configurations. The proposed phasing design approach was specifically designed for DDIs, so it
58 could be employed as a guide or a reference for signal timing design at DDIs. At the DDI of
59 Moana Ln and U.S. 395 in Reno, Nevada, the proposed scheme was evaluated and compared
60 with various signal phasing schemes by using the hardware-in-the-loop simulation technology.
61 The simulation results indicated that the proposed approach significantly outperformed the one
62 currently being used at this interchange.
63 Keywords: Diverging Diamond Interchange, Traffic Signal Control, Signal Phasing Scheme,
64 Webster Method, VISSIM.

65 INTRODUCTION
66 The DDI concept was first introduced to the United States by Gilbert Chlewicki in 2003 (1).
67 The first DDI was constructed in Springfield, Missouri in 2009 (2). Before that, the only known
68 DDIs existed in the communities of Versailles, Le Perreux-sur-Marne, and Seclin in France (3).
69 Field and simulation studies have demonstrated many advantages of DDIs over the conventional
70 diamond interchanges. As a result, seven states (Missouri, Utah, Georgia, Kentucky, Maryland,
71 New York, and Nevada) have constructed DDIs with demonstrated operational improvements
72 by the end of year 2012. More than 17 states are currently constructing DDIs in the U.S. (3).

73 Studies have been conducted to evaluate the performance of DDIs (1, 4-6). Most of
74 these studies compared the performance of DDIs to conventional diamond interchanges and
75 single point urban interchanges. Researchers sought signal timing plans either manually or by
76 using Synchro to optimize signal timing for each traffic scenario. Since Synchro does not
77 provide the specific function of signal timing optimization at DDIs, some of the research
78 interpreted the two crossover intersections of DDIs as two separate intersections operated by
79 two controllers (2). The microscopic simulation software of VISSM was then employed to
80 evaluate the performance. The existing research results have indicated that DDI works better
81 than the conventional interchange designs in terms of delay, number of stops, queue length, and
82 capacity (2, 6).

TRB 2014 Annual Meeting Paper revised from original submittal.

 3

Hu, Tian, Xu, and Andalibian

83 FHWA’s research suggested that the DDI signal timing could be two-phase control, with
84 each phase dedicated for the alternative opposing movements (7). Signals at a DDI can be fixed
85 time and may be fully actuated to minimize delay. Detectors can be used on all approaches of
86 both crossovers, and durations of signal phases may vary on a cycle-by-cycle basis. A DDI can
87 be operated by one controller or two. FHWA also indicated a three-phase signal operation by
88 adding a dummy phase as the spacing between the crossover junctions is very close. Until now,
89 there is no commonly accepted signal timing methodology specifically for DDIs. Therefore,
90 professionals rely on personal experience or conventional approaches for other types of
91 interchanges.

92 This study proposed a new methodology for the signal phasing scheme design at DDIs.
93 With minor adjustment, it can be applied to a variety of DDIs with different traffic demands and
94 geometric configurations. The proposed phasing design approach was specifically designed for
95 DDIs, so it could be employed as a guide or a reference for signal timing design at DDIs. This
96 research paper consists of four major sections. Section 1 provides a comprehensive literature
97 review pertinent to DDIs. Section 2 briefly introduces the current operation scheme and presents
98 a mathematical model for traffic signal timing design of DDIs based on Webster’s method.
99 Section 3 compares performance of the current traffic signal operation scheme and signal timing
100 by the proposed methodology with simulation results at the DDI of Moana Ln and U.S. 395
101 interchange in Reno, NV. The results indicated that the performance of the proposed scheme is
102 better than the current operation scheme at the studied interchange under the traffic demand
103 conditions. The last section offers conclusions of this study.

104 METHODOLOGY

105 Proposed Operation

106 The proposed phasing scheme is for one controller operation and can be used with all control
107 types: pre-timed, semi-actuated, and fully actuated. It is a general phasing scheme design
108 methodology for DDIs. The proposed phasing scheme is shown in Figure 1 and Figure 2. The
109 proposed operation operates very efficiently by applying overlapping phases extensively.

110 Phasing Scheme

111 As shown in Figure 1, phases 1 and 5 have the same split that is approximately equal to the
112 travel time (
,, sec) between locations of “11” and “7” (or between “4” and “8”) as depicted
113 in Figure 3. Phases 1 and 5 start and terminate at the same time, so they have the same “Max 1”
114 or “phase splits” in the controller settings for the same maximum recall. The two phases allow
115 the southbound (SB) off-ramp to be released earlier by
,, sec. Due to the early release,
116 there are no stops at the next signal, which means improved capacity. Phase 7 serves as the
117 green extension for phase 2, which allows the vehicles passing through location “5” in phase 2
118 to get to location “8”. These vehicles can then be served with green signal when phase 3 comes
119 on, which reduces the control delay of the northbound (NB) off-ramp traffic. In addition, phase
120 7 is set to “Max Recall” for safety and its duration must be less than the minimum green of
121 phase 4, which is set to be “Min Recall” in the controller. Another reason for setting “Max
122 Recall” on phase 7 is to avoid the queue between nodes “3” and “4” exceeding the segment

TRB 2014 Annual Meeting Paper revised from original submittal.

 4

Hu, Tian, Xu, and Andalibian

123 between these two nodes. Although Phase 8 is not assigned to any of the movements, the phase
124 with a “Min Recall” ensures phases 3 and 7 not to come on at the same time as they are actually
125 conflicting phases. Without phase 8, phase 7 will extend to run simultaneously with phase 3. If
126 phase 8 is set up with “Max Recall, the signal operation the DDI is not efficiently as phase 3
127 cannot gap out when its pertinent traffic is low. Figure 2 shows the locations of phases in the
128 proposed DDI phasing scheme. The SB on-ramp right-turn traffic and the NB on-ramp right-
129 turn traffic are controlled by “yield” type in this study. They can also be controlled by overlap
130 phases such as phase 4 for SB on-ramp right-turn traffic and overlapping phases 1 and 3 for NB
131 on-ramp right-turn traffic. Therefore, the proposed phasing scheme can be easily adjusted for
132 controlling other DDIs with different configurations.

133
134 Figure 1 Phases, Rings, and Barrier for Proposed Operation

135
136 Figure 2 Phase Location Diagram for Proposed Operation

TRB 2014 Annual Meeting Paper revised from original submittal.

 5

Hu, Tian, Xu, and Andalibian

137
138 Figure 3 Critical Nodes of a DDI

139 Signal Timing

140 1. Assumptions and Basic Strategies

141 Similarly to Webster’s method, the proposed operation assumes that the v/c ratios for critical
142 lane groups are equal, and that the green times allocated to critical lane groups are assumed to
143 be proportional to their saturation flow rates. Unlike the traditional Webster’s method, the
144 principle of deriving a real cycle length and the real splits (the cycle length and the splits shown
145 in a controller) of a DDI by the proposed operation is adding a certain of additional times
146 (coming from overlapping phases) to the real cycle length for calculating its effective splits in
147 proportion, and then deducting the additional times from the effective cycle length and splits to
148 obtain the real cycle length and the real splits, which will be input into a controller.
149 The proposed method introduces phases 2 and 6 into the signal operation, which lead to
150 two different timing schemes: phase 6 is one of the critical phases or phase 2 is the one of the
151 critical phases. The criterion for selecting the critical phases between phases 2 and 6 is
152 determined by the traffic demands and configurations of a DDI. The performance of these two
153 schemes is different as their operation efficiencies are different. The flow chart in Figure 4
154 indicates the basic steps for deciding which scheme to use.

TRB 2014 Annual Meeting Paper revised from original submittal.

 6

Hu, Tian, Xu, and Andalibian

155
156 Figure 4 Overview of Proposed Operation
157 2. Signal Timing Scheme 1 of Proposed Operation

158 Signal timing scheme 1 is for the situation that phases 6, 4, and 3 are critical phases. Phase 1 is
159 predetermined by the travel time of
,, sec.

160 1) Cycle Length

161 For the traffic signal operation shown in Figure 1, the DDI traffic signal timing parameters must
162 satisfy the relationship in the following equation:
163     
  
,, 
,,      2 ∗
,,   (1)
164 where
165  : effective green time of phase 3 (s);
166  : effective green time of phase 4 (s);
167 
: effective green time of phase 6 (s);
168  : sum of splits of phases 3, 4, and 6 shown in Figure 1 (s);

TRB 2014 Annual Meeting Paper revised from original submittal.

 7

Hu, Tian, Xu, and Andalibian

169
,,: split of phase 1, fixed and determined by the travel time between the two signals
170 “11” and “7” through nodes “11,” “12,” and “7” shown in Figure 3; and
171 : total lost time per cycle (s);
172 For this kind of DDI, the effective cycle length ( ) for the three critical lane groups
173 (SBL, eastbound (EB), and westbound (WB)) satisfy the following relationship with  :
174     2 ∗
,, (2)
175 Replacing  ,  , 
and  in Equations (1) and (2) yields:
 
176 ∑
 ∗      (3)
 

177 Then, based on the assumption,      for  ≠ " ,the effective cycle length is:
# # #
178   %  ∑'
 * (4)
∑'
 &  ) $
$ $ (
(
(

179 where
180 + : flow ratio for critical lane group + ; and
181 ,: sum of all critical lane groups.
182 For this traffic signal timing plan, the total lost time is:
183   -  -  -
(5)
184 -  -,  -.,  -,  ,  /0  1 (6)
185 where
186 - : lost time of critical lane group  (  4 345 6) (s);
187 -, : start-up lost time of critical lane group  (  4 345 6) (s);
188 -., : clearance lost time of critical lane group  (  4 345 6) (s);
189 , : yellow interval of critical lane group  (  4 345 6) (s);
190 /0 : all-red interval of critical lane group  (  4 345 6) (s); and
191 1 : extension of effective green time of critical lane group  (  4 345 6) (s).
192 -  -,  2 (7)
193 The lost time of phase 3 is 2 sec, since it is overlapped by phase 1. Phase 1 is overlapped
194 by phase 2. Therefore, the westbound through traffic has start-up lost time of 2 sec as shown in
195 Equation (7).
196 The real cycle length is computed by:
#
197      
,,  *   
,, (8)
$
(

198 2) Effective Green Times and Phase Splits

TRB 2014 Annual Meeting Paper revised from original submittal.

 8

Hu, Tian, Xu, and Andalibian

199 Effective green times can be calculated by:

8
200 7  ∑'8 ∗ (  ) (9)
 

201 where
202 7 : effective green time for phase  (  3, 4, 6).
203 The actual green time shown in Figure 1 can be obtained by:
204   ∅  ,  /0 (10)
205   7  ∅ (11)
206    7 (12)
207   
 
7  ∅ (13)
208 where
209 ∅ : is equal to
,, (s).
210 The phase splits 3, 4, and 6 are:
211 ∅    ,  /0 (14)
212 The + in Equation (4) can be replaced by:
 
213 +  =3{ ?,@A,B , ?,C,D } (15)
?,@A,B ?,C,D
 
214 +  =3{ A,@,F , A,@G,H } (16)
A,@,F A,@G,H

215 where
216 I, , : traffic volumes of lane group between nodes “14” and “7” (veh/h);
217 J, ,: saturation flow rate of lane group between nodes “14” and “7” (veh/h);
218 I,K,
: traffic volumes of lane group between nodes “5” and “6” (veh/h);
219 J,K,
: saturation flow rate of lane group between nodes “5” and “6” (veh/h);
220 I ,,L: traffic volumes of lane group between nodes “1” and “8” (veh/h);
221 J ,,L: saturation flow rate of lane group between nodes “1” and “8” (veh/h);
222 I , ,M: traffic volumes of lane group between nodes “10” and “9” (veh/h); and
223 J , ,M: saturation flow rate of lane group between nodes “10” and “9” (veh/h).

224 3. Signal Timing Scheme 2 of Proposed Operation

225 Signal timing scheme 2 is for the situation that phases 2, 3, and 4 are critical phases. In this
226 scheme, phase 1 is also predetermined by the travel time of
,, sec. Phase 1 serves SBL off-
227 ramp traffic with phase 6. Phase 1 also serves the WBT through traffic. Phase 7 is neglected in
228 the scheme 2 for studying the signal timing as phase 2 is the critical phase. It is not efficient to

TRB 2014 Annual Meeting Paper revised from original submittal.

 9

Hu, Tian, Xu, and Andalibian

229 calculate the cycle length and splits by adding phase 7 into the effective green time as phase 2
230 itself can serve NBL off-ramp traffic well, or in other words, phase 7 is not effective for the
231 entire intersection although it can still add an additional green time to phase 2. All the symbols
232 of the signal timing scheme 2 are same as the signal timing scheme 1.
233 1) Cycle Length

234 For the traffic signal operation shown in Figure 2, the DDI traffic signal timing parameters
235 should satisfy the relationship in Equation 17:
236        
,,     
,,   (17)
237 Under this condition, the effective cycle length for the three critical lane groups (NBL,
238 EB, and WB) satisfy the following relationship with  :
239    
,, (18)
240 Replacing  ,  ,  and  in Equations (17) and (18) yields:
 
241 ∑
 ∗      (19)
 

242 Then, based on the assumption,      for  ≠ " , the effective cycle length is:
# # #
243   %  ∑'
 * (20)
∑'
 &  ) $
$ $ (
(
(

244 The real cycle length is calculated as:

#
245       *  (21)
$
(

246 For this traffic signal timing plan, the total lost time is:
247   -  -  - (22)
248 -  -,  -.,  -,  ,  /0  1 (23)
249 -  -,  2 (24)
250 The lost time of phase 3 is only 2 sec, since it is overlapped by phase 1. Phase 1 is
251 overlapped by phase 2. Therefore, the westbound through traffic has start-up lost time of 2 sec
252 as shown in Equation (24).
253 2) Effective Green Times and Phase Splits

254 Effective green times can be calculated by:

8
255 7  ∑'8 ∗ (  ) (25)
 

256 The actual green time illustrated in Figure 1 can be obtained by:
257   ∅  ,  /0 (26)

TRB 2014 Annual Meeting Paper revised from original submittal.

 10

Hu, Tian, Xu, and Andalibian

258 
   7 (27)
259   7  ∅ (28)
260    7 (29)

261 CASE STUDY

262 The proposed schemes have proved to work well with one controller operation at DDIs (8). In
263 this study, the interchange performance with the proposed scheme was compared with the
264 performance of the existing traffic operation.

265 Site Description

266 The interchange of Moana Lane and U.S. 395 is a diverging diamond interchange in Reno, NV,
267 as shown in Figure 5. The interchange has high right-turn traffic demands on the off-ramps. No-
268 turn-on-red is used on the southbound off-ramp due to the dual-lane geometry. Other peak hour
269 volumes of the interchange are shown in Figure 5.

E Moana Ln

Year 2015
AM (PM)
Peak Hour Volumes
AM Peak: 7:00-8:00
PM Peak: 17:00-18:00
270
271 Figure 5 Bird’s-Eye View and Peak Hour Volumes of the Moana Lane/U.S. 395
272 Interchange (8, 9)

273 Existing Traffic Operation and Performance

274 The current actuated-coordinated signal timing plans running at the studied interchange were
275 developed by the City of Reno in 2012, with consideration of the heavy right-turn movements
276 on the off-ramps and southbound no-turn-on-red. The current signal phasing scheme has been
277 evaluated by Hu et al. in September, 2012 (8). The existing timing plan is displayed in Figure 6
278 and Figure 7.
279

TRB 2014 Annual Meeting Paper revised from original submittal.

 11

Hu, Tian, Xu, and Andalibian

280
281 Figure 6 Phase, Ring, and Barrier Diagram by City of Reno
282

283
284 Figure 7 Phase Designation Diagram by City of Reno
285 Phases 4 and 1 are not conflicting phases. There is no accurate way to determine the phase 1’s
286 split. Therefore, there is no accurate quantitative method to derive the cycle length for the
287 current method as well. The cycle lengths of the actuated-coordinated signal timing plans at the
288 DDI are provided as the existing actuated-coordinated plans for the adjacent signals: 110-sec
289 cycle for AM and 130-sec cycle for PM. In this study, the phase 1 split is assumed to be 10% of
290 the cycle length. Phases 2, 3, and 4 share the remaining time in proportion to their critical
291 saturation flow ratios. Table 1 shows each phase split in year 2015 AM and PM peak hours.
292 Table 2 provides the yellow and all-red intervals of the two actuated-coordinated plans for
293 current operation used in VISSIM simulation models.
294
295
296

TRB 2014 Annual Meeting Paper revised from original submittal.

 12

Hu, Tian, Xu, and Andalibian

297 Table 1 Cycle Lengths and Phase Splits of Current and Proposed Operations
Current Operation Proposed Operation
2015 AM 2015 PM 2015 AM 2015 PM
Cycle (sec) 110 130 110 130
Phase 1 (sec) 11 13 10 10
Phase 2 (sec) 23 24 30 29
Phase 3 (sec) 23 34 17 32
Phase 4 (sec) 53 59 53 59
Phase 5 (sec) N.A. N.A. 10 10
Phase 6 (sec) 23 24 30 29
Phase 7 (sec) N.A. N.A. 12 12
Phase 8 (sec) N.A. N.A. 58 79
298
299 Table 2 Clearance Intervals Used in VISSIM Simulation Models of Current Operation
Operation Phase 1 2 3 4 5 6 7 8
Critical Movement 10->9 3->8 3->8 12->7 None 12->7 None None
Current
Yellow (sec) 3.5 3.5 3.5 3.5 3.5
Operation
Red (sec) 1.5 3.5 3.5 2.5 2.5
Operation Phase 1 2 3 4 5 6 7 8
Critical Movement 3->8 3->8 None 12->7 None 10->11 5->4 None
Proposed
Yellow (sec) 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3
Operation
Red (sec) 3.5 3.5 2.5 2.5 1.5 1.5 1.5 0

300 Proposed Operation

301 Based on the traffic volumes provided by NDOT, the proposed signal operation is from the
302 schemes presented in this study shown in Figures 1, 2, and 4. Its cycle lengths and phase splits
303 are summarized in Table 1. In fact, the cycle lengths based on the proposed operations are close
304 to the current operation’s cycle lengths in both AM and PM peak periods. In order to eliminate
305 the cycle length’s effects on the current and proposed operations, the proposed cycle lengths are
306 assumed as the current method’s cycle lengths both in AM and PM peak periods. These splits
307 were acquired by the methodologies for proposed operations introduced in this study. The study
308 uses the clearance intervals in VISSIM simulation models as shown in Table 2.
309 Table 2 These intervals are defined based on proposed operations and the specific geometry
310 conditions at the Moana DDI. The proposed signal operation is an actuated-coordinated control
311 plan, which has been successfully tested in a Naztec 2070 traffic controller (10).

TRB 2014 Annual Meeting Paper revised from original submittal.

 13

Hu, Tian, Xu, and Andalibian

312 Simulation Results

313 Each VISSIM model ran five simulations with random seeds. The warm-up time in the
314 simulation models was 300 sec, followed by 3600 sec of run time with data compiled and
315 collected for each traffic movement at the Moana DDI. The average delays of five runs of the
316 current and proposed operations as well as the compared results of their average delays are
317 presented in Table 3. The average maximum queues of the five runs of the current and proposed
318 operations as well as their average maximum queues are provided in Table 4. The percent
319 values in Table 3 are obtained by taking the difference of the average delays from the proposed
320 operation and the current operation and then dividing by the average delay found in the current
321 operation. For example, the value “11%” shown in the table is obtained by (17.9-
322 16.1)/16.1*100%. The value “16.1” comes from the current operation average delay of traffic
323 movement “10->9” during the AM peak hour. Similarly, the number “17.9” is the proposed
324 operations average delay of the same traffic movement in the period. Based on simulation
325 results, proposed operations brought about less average delay for most movements but greater
326 average delay of movement “10->9” compared to the operation developed by the City of Reno
327 for 2015 AM and PM peak hour periods. The reason for less delay of movement “10->9” by the
328 current operation is that it adds phase 1 to this traffic movement compared to the proposed
329 operation. The average delay for all vehicles of the proposed operation dropped by 17% and
330 28% in AM and PM peak hours, respectively, compared to the current operation.
331 Similar to the data in Table 3, Table 4 summarizes the results of average maximum queues. The
332 average maximum queue of movement “10->11” under the proposed operation decreased by
333 15% and 44% in the AM and PM peak hours, respectively, compared to the current operation.
334 The average maximum queue of movement “5->4” in the proposed operations reduced 21% in
335 both the AM and PM peak hour over the current operation, specifically from 222.1 to 174.8 feet
336 in the AM peak and from 308.9 to 245.5 feet in the PM peak. Of the other traffic movements
337 including “1->8,” “1->2,” “12->7,” and “14->7,” the proposed operation performed better than
338 the current operation during the peak periods. However, proposed operations increased the
339 maximum queue by 0% and 24% of movement “10->9,” and increased 11% and 14% of the
340 maximum queue of movement “3->8” ” in the AM and PM peak hour periods to 69.0 and 84.3
341 feet, respectively. The same reason for greater average delay of traffic movement “10->9”
342 brought about the larger average maximum queue of this movement. The reason for longer
343 maximum queues of movement “3->8” experienced with the proposed operation is that this
344 operation allows the release of northbound left (NBL) traffic earlier, which stops in front of
345 node “8.” The maximum queues of this movement were less than 84.3 feet. This maximum
346 queue is acceptable for reducing the delay in front of node “4.” Of the other movements, both
347 operations performed well with no noticeable differences for their low traffic volume, even
348 though they seemed much different.
349

TRB 2014 Annual Meeting Paper revised from original submittal.

 14

Hu, Tian, Xu, and Andalibian

350 Table 3 Average Delays from VISSIM Simulation Models (sec/veh)

Peak Hours AM PM

Movement Current Proposed Change (%) Current Proposed Change (%)

10->9 16.1 17.9 11% 22.1 26.9 22%

10->11 45.8 33.5 -27% 85 45.3 -47%
1->8 21.7 16.3 -25% 35.9 30.3 -16%
1->2 0.8 0.8 5% 2.7 2.1 -21%
3->8 7.7 8.7 14% 7.5 6.3 -15%
3->2 0.3 0.3 -7% 0.8 0.3 -67%
14->13 0.6 0.5 -4% 2.6 1 -59%
14->7 51.7 47.6 -8% 68.3 51.1 -25%
5->6 0.8 0.7 -8% 1.3 1.2 -12%
5->4 48.7 32.5 -33% 66.1 41.1 -38%
12->13 1.4 1.3 -8% 3.7 3 -18%
12->7 5.9 1.5 -75% 14.6 2 -86%
All 18.3 15.1 -17% 27 19.5 -28%
351
352
353 Table 4 Average of the Maximum Queue from Simulation Models (ft)
354
Peak Hours AM PM

Movement Current Proposed Change (%) Current Proposed Change (%)

10->9 288.2 287.2 0% 345.1 427.1 24%

10->11 206.4 175.3 -15% 479.4 267.8 -44%
1->8 250.7 229.9 -8% 634.5 594.8 -6%
1->2 41.7 30.8 -26% 365 325.4 -11%
3->8 62 69 11% 74 84.3 14%
3->2 13.5 37.7 180% 50.9 67 32%
14->13 0 0 N.A. 0 0 N.A.
14->7 274.8 279.7 2% 632.1 526.3 -17%
5->6 0 0 N.A. 0 0 N.A.
5->4 222.1 174.8 -21% 308.9 245.5 -21%
12->13 0 7.8 N.A. 100.3 59.2 -41%
12->7 101.8 95.6 -6% 279.6 238.9 -15%

TRB 2014 Annual Meeting Paper revised from original submittal.

 15

Hu, Tian, Xu, and Andalibian

355 CONCLUSION
356 Since there is no widely accepted signal timing methodology for DDIs, in the literature, traffic
357 engineers and researchers designed DDI signal timing plans based on their personal experience
358 and approaches for conventional interchanges, which have been proven to be inefficient. Some
359 of the signal phasing schemes in existing research did not consider the capability of signal
360 controllers, so they are not readily implementable in the field.
361 Because DDIs are a relatively new interchange design, there is a limited amount of related
362 literature. Most DDI related reports do not include detailed descriptions of feasible signal
363 phasing schemes. Where signal phasing was actually discussed, most reports did not follow the
364 NEMA convention, thus the phasing schemes were not readily implementable in the field. This
365 paper presented an innovative phasing scheme and a methodology for determining cycle length,
366 splits, and phase intervals for DDIs with consideration of the capacity of signal controllers. The
367 proposed phasing scheme employed the conception of overlap phases, which was inspired by
368 the TTI-4 signal strategy for conventional diamond interchanges. The phasing scheme and the
369 timing methodology had been fully tested by Hu et al. (8) for its validity under pre-time, fully
370 actuated, and actuated-coordinated control modes. The proposed phasing scheme and timing
371 methodology can also be easily modified to meet the specific DDI requirements under a variety
372 of geometric characteristics and traffic demand conditions. A case study was conducted using
373 the first DDI site in Nevada. The simulation results by VISSIM showed that the proposed
374 operation outperformed what was initially implemented by reducing vehicle delays.
375 Further research identified through this study includes the following. First, the
376 relationship between cycle length and interchange level delay needs to be further addressed.
377 Second, the relationship between traffic demands, phasing scheme, and signal timing
378 parameters, needs additional investigation. Finally, the effect of signal spacing on DDI’s
379 performance needs to also be thoroughly researched.
380

TRB 2014 Annual Meeting Paper revised from original submittal.

 REFERENCES

1. Chlewicki, G. New Interchange and Intersection Designs : The Synchronized Split-

Phasing Intersection and the Diverging Diamond Interchange. 2nd Urban Street
Symposium, Anaheim, CA, 2003, pp. 3–16.

2. Missouri Department of Transportation. Missouri’s Experience with a Diverging Diamond

Interchange. Report No.: OR 10 - 021, Jefferson, MO, 2010, pp. 4–22.

3. Wikipedia. http://en.wikipedia.org/wiki/Diverging_diamond_interchange. accessed Oct.

29, 2013.

4. Bared, J., P. Edara, and R. Jagannathan. Design and Operational Performance of Double
Crossover Intersection and Diverging Diamond Interchange. Transportation Research
Record, Vol. 1912, No. 1, Jan. 2005, pp. 31–38.

5. Bared, J. G., T. Granda, and A. Zineddin. Drivers’ Evaluation of the Diverging Diamond
Interchange. Publication No.: FHWA-HRT-07-048, McLean, VA, 2009, p. 11.

6. Stanek, D. Innovative Diamond Interchange Designs : How to Increase Capacity and

Minimize Cost. ITE 2007 Annual Meeting and Exhibit, Pittsburgh, PA, 2007, pp. 1–17.

7. Hughes, W., R. Jagannathan, D. Sengupta, and J. H. Hummer. Alternative

Intersections/Interchanges: Informational Report (AIIR). Publication No.: FHWA-HRT-
09-060, Washington, DC, 2009, pp. 217-259.

8. Hu, P., and Z. Tian. Proposed Signal Control at the Moana Lane / U . S . 395 Diverging
Diamond Interchange. NDOT Sponsored Research (to be published), Reno, NV, 2012.

9. Diverging Diamond Interchanges, Nevada Department of Transportation.

https://www.nevadadot.com/Traveler_Info/Safety/Diverging_Diamond_Interchange.aspx.
accessed Oct. 29, 2013.

10. Hu, P., and Z. Tian. Advanced Signal Control Strategies and Analysis Methodologies for
Diverging Diamond Interchanges. University of Nevada, Reno. Reno, NV, 2013.
http://www.youtube.com/watch?v=VVOvzOTHUDA. Accessed Oct. 29, 2013.

TRB 2014 Annual Meeting Paper revised from original submittal.