1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

Modeling Diverging Diamond Interchange Under Constraints of Ramp Metering, HOV Lane and BRT Transit

Wei Zhang, Ph.D. PE (Corresponding Author) Highway Research Engineer/Intersection Safety R&D Program Manager Federal Highway Administration Turner Fairbank Highway Research Center 6300 Georgetown Pike, HRDS-10, McLean, VA 22101, USA Tel: (202) 493-3317; Fax: (202) 493-3417; Email: [email protected]

Nopadon Kronprasert, Ph.D. Lecturer Department of Civil Engineering Chiang Mai University 239 Huay Kaew Road, Muang District, Chiang Mai 50200, Thailand Tel: : +66 53-944156 ext 109; Email: [email protected] George Merritt, PE Safety and Geometric Design Engineer Federal Highway Administration Resource Center Safety and Design Technical Service Team 61 Forsyth Street, SW Suite 17T26, Atlanta, GA 30303, USA Tel: 404-562-3911; Fax: (404) 562-3700; Email: [email protected]

Word Count: 4,645 words text + 10 tables/figures X 250 Words (each) = 7,145 words Submission Date: November 9, 2015

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

ABSTRACT The Diverging Diamond Interchange (DDI) design is gaining public acceptance due to its proven ability in delivering mobility and safety at substantially lower cost than conventional designs. However, many field constrains have also been identified that can severely limit DDI’s capacity utilization. This prompts designers to develop realistic plans for a 2-phase DDI to function under the field constrains. This study investigated the potential of using DDI to improve a congested compact diamond interchange in the United States. The interchange faces many constraints such as congested freeway, high occupancy vehicle (HOV) lane on the on-ramp, transit buses on cross-street, bus rapid transit (BRT) service along the freeway, and potentially high pedestrian and bicycle traffic, etc. A micro-simulation network was built that included a freeway segment, five signalized intersections, and six right-in and right-out driveways over a mile-long corridor. A 6-lane DDI configuration was used to simulate performances under the projected 25-year traffic demands. Signal coordination was optimized in SYNCHRO and the resulting timing parameters were coded into VISSIM. The simulation network also implemented local ramp metering policy, on-ramp HOV lane with separate metering rates, and a bike-priority signal timing plan at the ramp intersections. In addition to traditional performance measures such as queues and travel times by movement at each intersection, this study also counted the number of vehicles entering and leaving the on-ramp in 6-minute intervals to evaluate the potential of queue backing up to the surface street. The simulation study indicates that a 6-lane DDI design can serve the future traffic demand without need of widening the bridge.

INTRODUCTION Since the opening of the first DDI in the U.S., located at I-44 and MO 13 in Springfield, MO, in 2009, at least 44 more DDIs have been constructed and opened to traffic. As of September 2014, 46 states have had one or more DDI projects in planning, design and/or construction. The number of DDIs completed by year is shown below (1): Year 2009 2010 2011 2012 2013 2014 2015 (as of February 2015)

Number of DDIs completed 1 4 5 8 14 12 1

It is safe to say that DDI has now been accepted by practitioners on a broad scale and is becoming a mainstream design in the toolbox. To help accelerate DDI’s mainstreaming, the USDOT funded a DDI field evaluation study; including DDI designs in a nation-wide promotion during the second round of Every Day Count (EDC2) program. The EDC2 program includes numerous trainings, project reviews, and technical assistances to agencies interested in implementing DDI or other forms of alternative intersection/interchange design. The Federal highway Administration (FHWA) is also developing a Highway Capacity Manual procedure for DDI (2).

1

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

DDI design has many advantages over competing designs, most notable, high capacity, low cost, and short construction time. It is particularly attractive in retrofitting congested diamond interchanges. As DDIs are implemented in different traffic and roadway environments, issues and challenges associated with DDI operation and capacity utilization start to surface. The issues and challenges identified so far include, but are not limited to: 1. Congested signal intersection adjacent to the DDI crossover intersections that limit the capacity utilization of the DDI design. 2. BRT vehicles or large trucks that need to exit the freeway and re-enter the freeway at the same interchange. 3. Nearly saturated freeway segment at interchange location that requires ramp metering to keep some vehicles on the ramp. 4. High pedestrian traffic which requires a protected walk phase to cross the surface street at DDI crossover and adjacent intersections. 5. Frequent passage of emergency vehicles through the DDI that disrupt signal coordination. Rather than being rare events, the above issues (and other identified challenges associated with the DDI design as well) tend to occur at many if not most DDI sites. This is due to DDIs often being chosen at high traffic and high crash locations where such issues already exist, and existing interchange has experienced unacceptable level of service. Practitioners and researchers in the U.S. have started tackling the above issues from different directions. To solve the problem presented by issue #1, the National Highway Cooperative Research Program is funding a study to develop guidance on treating congested signal intersections adjacent to DDI’s crossover intersections (3); Zhang and Kronprasert proposed solutions to a real project in Alaska using relaxed bow-tie, superstreet, and quadrant road intersection designs to treat congested signal intersections adjacent to DDI crossovers (4, 5). Their analyses suggested that converting nearby intersections into at-grade alternative intersections can cut the signal cycle length of nearby intersections significantly by eliminating the protected left-turn phases and pedestrian crossing phase. Doing so makes the adjacent signal controlled intersections more compatible with DDI’s 2-phase operation, leading to sizable increase in system throughput. To address the challenge presented by issue #2, Chlewicki proposed the concepts of expanded DDI and diverging double roundabout (6). Although these concepts have not been implemented in the field, as ideas accumulate, field implementation of one or more such ideas is expected in the near future. In Utah, practitioners have installed continuous flow intersection and through-turn intersection designs adjacent to single point urban interchanges (SPUI). In July 2014, TRB hosted the First Alternative Intersection and Interchange Symposium in Salt lake City, Utah, providing a knowledge exchange platform for solving many of the above issues.

PROBLEM DESCRIPTION The interchange studied is I-805 and Palm Ave, about 16 miles southeast of downtown San Diego, CA, and about 3 miles north of the Mexico border. The location map and peak hour traffic counts are shown in Figure 1. This is an overpass diamond interchange built in the 1970s. The freeway underneath the bridge has 4 lanes in each direction, and becomes five lanes after the on-ramp joins the freeway. The fifth lane continues to the next interchange and turns into the exit lane. The bi-directional average daily 2

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13

14 15 16 17 18 19 20 21 22 23 24 25 26

traffic (ADT) on the freeway is 165,300 between freeway exit ramp and entrance ramp, and 226,130 downstream of the on-ramps. During the peak period, the freeway segment is nearly saturated, and the on-ramp traffic must be carried by the fifth lane. The overpass bridge sits on steep slopes. On Palm Ave, there are multiple bus lines and high pedestrian traffic. The center-to-center distance between the two ramp intersections is about 650 ft (198.25 m). In each direction of travel, the bridge has one 10-ft (3.05 m) raised side walk, one 6-ft (1.83 m) bike lane, and two through lanes; at the center of the bridge are back-to-back double left-turn lanes. During the peak hours, on ramp left-turn traffic often fill up the leftturn pocket, back up to the through lane, and block through traffic on the surface street. A large retail and industrial center is being planned at a location 2 miles south which will generate additional traffic to this interchange. The future plan for this site includes a park and ride lot at one quadrant, HOV lane and bike lane on the ramps, BRT service along the freeway, and ramp metering control to on-ramp traffic. Any improvement designs proposed must function under the above field constrains.

Figure 1. Project location and satellite view of the interchange The stakeholders of this project include FHWA CA Division, CALTRANS, City of San Diego, and San Diego Association of Governments (SANDAG). CALTRANS hired two consulting firms to perform traffic impact study. The study indicated that the existing interchange cannot handle the future traffic demand, and capacity upgrade must be made to this critical interchange. It should be noted that since the freeway cannot accommodate all the on-ramp demand, some on-ramp vehicles must be temporarily stored on the ramp. If the ramp queue backs up to the surface street, it will quickly deteriorate the interchange’s operation. Therefore merely trying to increase the interchange’s traffic capacity (such as constructing a brand new interchange) is not a viable solution. The capacity boost must be within a reasonable range and in tandem with the freeway segment’s traffic receiving capacity. To solve such a problem, the conventional approach is to widen the bridge to increase the directional throughput and add 3

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10

11 12 13 14 15 16 17 18 19 20 21 22 23

loop ramp(s) at one or two quadrants to store the on-ramp vehicles. This option was studied by CALTRANS. The FHWA team was tasked to investigate the feasibility of a DDI design at this location. SIZING the DDI The forecast traffic demand at this location is shown in Figure 2. The traffic patterns at the ramp intersections indicate heavy NB on-ramp traffic demand during morning peak, and heavy SB off-ramp left-turn demand during afternoon peak. To evaluate the feasibility of the DDI design, the first step is to determine the number of lanes needed in each direction.

Figure 2. Forecast future AM/PM peak hour traffic count A planning tool called Capacity Analysis for Planning of Junctions (CAP-X) developed by FHWA was used for this purpose. CAP-X is an easy to use tool for estimating traffic capacity of 15 different types of alternative intersection designs. The intersection capacity estimated by CAP-X is determined from critical lane volume method. It can also be used for sizing the number of lanes required for specific type of alternative intersection designed to carry the forecast traffic demand. The CAP-X analysis indicated that a 5-lane DDI (3-lane EB and 2-lane WB) or larger and double on-ramp and off-ramp left-turn lanes would be needed to serve the forecast traffic demand. With the 5-lane DDI design and assuming pedestrians crossing the freeway are handled by a protected center median as done at most overpass DDIs, the existing bridge would be wide enough for the proposed 5-lane DDI. Figure 3 shows the possible lane reconfiguration. One can see that the roadway width needed by DDI is less than existing 4

Zhang, Kronprasert, Merritt 1 2 3 4 5

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

interchange, which means if the bridge is structurally sound, no widening would be needed. This is a key advantage of the DDI design and also the main reason of its increasing popularity. In this case, the protected median for pedestrian/bicycle traffic can be made much wider than 12 ft (3.66 m) to fully utilize the bridge deck’s width and increase the angle of the crossover paths.

Figure 3. Suggested lane re-configuration from 6-lane diamond to 5-lane DDI One can see from Figure 3 that there is enough room to make the protected central median for Pedestrians/bicyclists wider or add one more lane to the DDI. For this project, although the NB on-ramp demand is high (Figure 2), the total traffic demand at each ramp intersection is not extraordinarily high. Since the freeway segment is congested and ramp metering must be enforced to limit the number of vehicles that can enter the freeway during the peak periods, sizing up the DDI design further won’t help reduce congestion at the interchange unless the freeway congestion is reduced or removed.

EVALUATING FEASIBILITY of 5-LANE DDI This project has many challenges and constraints that were not experienced at any of the 45 DDIs already constructed. Due to the complexity of constraints involved, the FHWA team adopted a step-bystep approach to solve this problem. First, a micro-simulation was performed to quantify the performance improvement that the 5-lane DDI may bring over the existing condition; the results were summarized and presented to the customers for discussion, to ensure the key assumptions used in the simulation were realistic and implementable in the field; if there is enough spare capacity, then 5

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

additional field constraints were gradually added to the simulation model, and interim results were discussed with customers again to examine if the traffic control policies (such as ramping metering) and other constraints were modeled properly. After each step of iteration, the DDI’s throughput capacity was maintained or reduced slightly, but the simulation model itself would be one or more steps closer to field condition. If after all the identified constraints have been implemented, the DDI’s capacity is still within a reasonable range relative to the expected traffic demand, then the 5-lane DDI design can be considered viable from traffic perspective. Otherwise, either additional lane(s) must be added to the DDI design or other forms of interchange be considered before continuing the simulation analysis. This study was performed over a period of five months, with one face-to-face meeting in San Diego, and five Webinars. Simulation Network and Signal Timing Coordination The authors built a VISSIM model from Google Maps and referenced project documents when specifying model details. The simulation network included five signalized intersections as shown in Figure 2 and 6 un-signalized right-in-right-out (RIRO) intersections in between. The peak hour traffic counts at the 6 RIRO intersections were provided by customers but not shown in Figure 2 (to reduce visual clutter). Based on the peak hour traffic counts at each intersection, the origin-destination matrix was developed using the maximum likelihood method. The O-D pairs in the matrix form the static routes of all vehicles in the simulation network, and enable the tracing of any vehicles by type and route for performance measures such as travel time, speed/acceleration profiles, and so on. It is the authors’ standard practice to use static O-D routes when simulating local traffic network. O-D routes can be determined using other methods, or dynamically assigned by the micro-simulation software. Due to page limit constraint, the O-D matrix will not be presented in this paper. The key simulation setups are: Vehicle speed: Freeway: 65 mph (105 km/h, I-805) Arterial: 35 mph (56 km/h, Palm Ave), 45 mph (73 km/h, Dennery Rd), 25 mph (40 km/h, Driveways) Vehicle composition Freeway – 95% Car and 5% Heavy vehicle Arterial – 98% Car and 2% Heavy vehicle Interchange Lane Configurations 6-lane diamond – Existing (2 TH lanes in each direction & back-to-back double left-turn lanes) 5-lane DDI (3-lane EB, 2-lane WB, double on-ramp/off-ramp LT lanes, single RT on-ramp) Signal Timing Diamond – 3-phase signal DDI – 2-phase signal The next step is to optimize the signal coordination plan. For any DDI project, this is the most critical step in simulation study. The signal timing parameters used shall be realistic, conforming to local standards/policies, and implementable in the type of signal controller(s) in the field. Synchro was used for this purpose. The dominate route used for optimization was from freeway SB exit to EB through 6

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

route. Signal coordination parameters were set such that no congestion will form on this route; or if congestion is unavoidable, the signal timing plan will minimize congestion on this route. The signal timing parameters used at each signalized intersection are shown in Figure 4. For the 6 RIRO intersections, some are protected by flexible bollards (which separate the through traffic on major road from RIRO traffic), traffics at such intersections were modeled as free-in and free-out; some are controlled by STOP sign, at these intersections, right-in traffic was modeled as free right-in, and rightout traffic was modeled as must finding a gap of 6 to 8 seconds to enter the major road.

Figure 4. Signal timing setup Simulation Results and Incremental Refinements Tables 1 and 2 show the measures of efficiency (MOEs) for the 6-lane diamond and 5-lane DDI during the AM peak. Comparing the MOEs shown in these two tables, one can see that the 5-lane DDI reduces intersection delay at 4 signal intersections, and increases intersection delay by 3.6 seconds at the east most signal intersection, which is furthest from the DDI crossover. The level of service improves from C to B at two intersections. Note the significant reductions in approach delay at many intersections. Similar simulation analysis was performed for PM peak traffic. It needs to point out that when adding in other site constraints such as ramp metering rate, and ramp vehicle storage need, the interchange’s performance under AM peak traffic becomes critical due to the capacity constraint imposed by the congested freeway. 7

Zhang, Kronprasert, Merritt 1

Table 1. MOE of 6-lane diamond interchange under future AM Peak traffic (No build)

2 3 4

Table 2. MOE of 5-lane DDI under future AM Peak traffic

5 6 8

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Pedestrian and Bicycle Traffic The next issue to address is how well the 5-DDI design can handle the pedestrian and bicycle traffic on the surface street. This is reflected in two aspects: ped/bike traffic at the DDI crossover intersections, and ped/bike traffic at other signal intersections. Since pedestrian and bicycle traffic demands were not given in the future traffic forecast, they can only be assumed in the analysis. Under the no-build condition, due to the 3-phase signal operation and the constant traffic conflicts at the ramp intersections, pedestrian crossing in any direction must be protected. Therefore under no-build condition, protected walk phase will take away the green time available for vehicle traffic, and reduce the ramp intersections’ capacity. For DDI design, since only two phases exist at the crossover intersections, pedestrians can move simultaneously with vehicles during the green phase when getting into and out of the protected central median. In this aspect, pedestrian traffic doesn’t reduce the intersection’s vehicular traffic capacity. The simulation analysis showed that even if 200 pedestrians per hour were present at the crossover intersections, it won’t noticeably degrade the interchange’s level of service. Discussion with the customer suggested that due to heavy right-turn on-ramp traffic at the east ramp, for safety reason, a pedestrian activated signal is needed on the right-turn lane leading to freeway on-ramp to resolve the conflict between pedestrians and right-turn on-ramp traffic. This push-button pedestrian signal only potentially reduces capacity to the right-turn on-ramp traffic (when heavy pedestrian traffic coincides with vehicular peak traffic). To a large extent, it doesn’t truly affect the DDI’s ability of moving through and left-turn on-ramp traffic. It should be noted that pedestrian crossing the major road at nearby intersections next to the interchange ramps may have negative impacts to the interchange’s capacity. Such impacts will be similar whether the interchange is a 6-lane diamond or 5-lane DDI. For bicycle traffic, two types of bikers were considered - the casual bikers who act like pedestrians and will use the pedestrian path at the interchange; the experienced bikers who act like drivers and will use the traffic lane rather than the protected center median. No special treatment is needed for causal bikers. For experienced bikers, if their demand is low, say less than 50 bikes/hr (for 126 seconds cycle length, this translates into average 1.75 bikes per signal cycle) , they can use the green time assigned to vehicle in their direction of travel, and will not affect the LOS of the interchange. If their demand is high, say 6 to 10 bikes per signal cycle, experienced bikers may be expected in any signal cycle, some special treatment may be necessary. For this scenario, the authors implemented an intersection treatment and signal timing plan originated from Europe as shown in Figure 5 – a 12-ft (3.66 m) or so space upstream of the intersection stop bar is reserved for bikers, the stop lines for motorized vehicles begin at the end of the bicycle storage area. During the yellow/red phases, bikers arriving at the intersection move forward to fill up the bicycle storage area; when the signal turns green, a 6-second or so green time is dedicated to the bike traffic; and followed by green time for vehicles. The above idea was acquired by the lead author while attending a bicycle safety Webinar, and implemented into the simulation by Dr. Kronprasert. No attempt was made to identify a literature source for this treatment. Such intersection design and signal operation reduce the DDI’s capacity slightly, but not much. Bicycle detectors may be installed at the intersection, and the protected bike phase can be skipped if no bicycles are detected during the yellow and red phases. The simulation analysis up to this stage indicated that the 5-lane DDI design can still perform at acceptable LOS. Ramp Metering and HOV Lane on the Ramp

9

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13

14 15 16 17 18 19 20 21 22 23 24 25 26 27

The next constraints to be considered are ramp metering rates and HOV lanes on the ramp. According to the local policy, the ramp metering rates for the two ramps are set as follow: AM Peak, 6:00 am to 9: 00 am:

NB Ramp: 165 vehicles every 6 minutes SB Ramp: 48 vehicles every 6 minutes

PM Peak, 4:00 pm to 7:00 pm:

NB Ramp: 165 vehicles every 6 minutes SB Ramp: 65 vehicles every 6 minutes

HOV lane on the ramp is another requirement of this project. In the simulation study, HOV lane is only implemented in the NB ramp as shown in Figure 3, because this is the critical ramp that makes the interchange fail or remain. The SB ramp has low on-ramp demand and plenty of ramp storage space.

Figure 5. Bicycle oriented intersection design and signal timing for DDI design (also shown on the NB ramp is the HOV lane, located in the middle of LT on-ramp and RT on-ramp lanes) The NB ramp metering rate of 165 vehicles every 6 minutes translates into one vehicle released into the freeway every 2.18 seconds. For each passenger car lane, it becomes one vehicle released into the freeway every 4.36 seconds. The ramp metering works like this: If there is no vehicle on the HOV lane, the signal head on HOV lane remains RED and the green time rotates through the two passenger car lanes; if there are vehicles on the HOV lane, the green time rotates through all 3 lanes on the ramp. In any case, the vehicle release rate is kept between 2 or 3 seconds per vehicle. In the simulation, it was assumed that 10% of the on-ramp traffic is HOV traffic; and vehicle detectors were set up at proper locations to count the number of vehicles entering and leaving the NB ramp in 6-minute intervals. This result was used to assess whether the NB ramp has enough storage space under future traffic when the prescribed ramp metering rate is enacted. 10

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10

11 12 13 14 15 16 17 18 19

Figure 6 shows the dynamic queuing in number of vehicles under given traffic arrival rates during the AM and PM peak periods. One can see that the maximum number of vehicles that may stay on the NB ramp is 42 vehicles during AM peak, and 37 vehicles during PM peak. Assuming vehicles queue up evenly on the regular lanes, the number of vehicles expected on each regular lane would be 42*0.9/2 =18.9. Therefore it is safe to say that the maximum queue on any regular lane would not exceed 20 vehicles, or (20 veh * 25 ft/veh =) 500 ft (152.5 m). Measuring from the Google Maps, the existing NB ramp is approximately 1,500 ft (457.5 m). Even if the ramp meters are located in the middle of the ramp, there is sufficient space to store the NB on-ramp vehicles.

Figure 6. NB on-ramp vehicle storage need analysis. Table 3 shows estimated travel times and link speeds on critical O-D routes between the no-build and 5lane DDI designs. One can see that DDI’s 2-phase signal operation delivers significant reduction in intersection control delay and keeps traffic moving at reasonable speeds. Figure 7 shows the traffic scene of the 5-lane DDI under the constraints of ramp metering and HOV lane.

11

Zhang, Kronprasert, Merritt 1

Table 3. Comparison of travel time and link speed on select routes (1 ft = 0.305 m, 1 mph = 1.61 km/h)

2 3

4 5 6 7 8 9 10 11

Figure 7. Traffic scene of 5-DDI design under ramp metering and HOV lane constraints BRT vehicles BRT buses travel along the freeway. Their typical operation is to exit the freeway at a designated interchange, unload the passengers, and then re-enter the freeway from the same interchange. This constraint may be handled in any of the following ways: 12

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

1. If the bridge is wide enough for U-turn movement, channelize the BRT bus to turn left at the exit ramp onto the surface street; upon arriving at the adjacent DDI crossover intersection, activate a bus priority signal to enable the BRT bus to make a U-Turn in clock-wise direction, and get onto the left-turn on-ramp lane in the opposite direction. 2. Created a dedicated BRT bus U-Turn location on raised median of the surface street, and provide priority through signal operation to facilitate such BRT bus movement. The DDI design is inherently unfriendly to this type of off-ramp and then on-ramp movement in the same direction of travel. The decision makers must weigh the benefit of the DDI design (operation, safety and cost) against the inconvenience that this design may bring to certain traffic movement, and decide whether DDI design is a suitable choice. Please note that for most interchanges, the traffic demand for off-ramp followed by on-ramp in the same direction is negligible.

DISCUSSIONS In this paper, the authors described the incremental process that the FHWA team used in improving the simulation model and presented key results for assessing the suitability of a 5-lane DDI design at the project location. The major challenge here is the congested freeway, which limits the number of vehicles that can enter the freeway during the peak period. This constraint is particularly acute at the NB ramp. The existing diamond interchange cannot function well under current or future traffic demand primarily because of this constraint. A 5-lane DDI can provide enough boosts in interchange capacity, and at the same time, not overwhelm the NB freeway on-ramp. The constraints imposed by pedestrian/bicycle traffic on the surface street, ramp metering and HOV lane on the ramp, etc. will all reduce the DDI’s throughput at the crossover junctions, but none will be significant enough to result in failure on any critical route. The good side of these constraints is that although they increase vehicle travel time getting into the NB ramp, they also reduce vehicle’s wait time on-ramp, and the overall travel times for vehicles going the NB freeway route remain unchanged. Although the DDI design is inherently unfriendly to off-ramp and on-ramp movement like the BRT bus, the DDI design was conceived based on the fact that for the vast majority of interchanges, this type of traffic demand is very low to the point that they can be neglected.

CONCLUSIONS 1. When solving seemingly complicated traffic analysis problems, it is advisable to use a divideand-conquer approach, start from simple scenarios, and gradually add additional constraints. 2. The analyses show that a 5-lane DDI design can satisfactorily serve the future traffic demand at this location. 3. DDI is feasible at locations where the freeway is congested; under this circumstance, the DDI’s capacity should not be over designed, and surface street signal coordination should be set to more uniformly distribute travel delays on different links of the travel route. 4. Field constraints such as heavily pedestrian/bicycle traffic on surface street, and ramp metering, etc. generally reduce the interchange’s traffic handling capacity, but can be managed. Under the 13

Zhang, Kronprasert, Merritt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

circumstance described in this paper, some constraints counteract each other (i.e. pedestrian protection on surface street and ramp metering), and don’t result in cumulative deterioration of the traffic operation. 5. The 5-lane DDI design cuts the travel time on critical routes significantly; the speeds on those links, although increased, remain safe at 26 mph or less. 6. DDI design is inherently unfriendly to BRT movement along the freeway. It is up to the decision makers to evaluate the operational and cost benefits of the DDI design, and the inconvenience it brings to the off-ramp and then on-ramp movement of BRT bus. ACKNOWLEDGEMENT The authors wish to thank Jeff Shaw, FHWA Office of Safety, and Mark Doctor, FHWA Resource Center, for their contributions to this study, to Jeff Holm and Manuel Sanchez of the FHWA California Division for connecting the FHWA team with CALTRANS team, and staff members from the CALTRANS team for their comments, suggestions, and timely delivery of relevant project and policy documents to keep the study on pace. REFERENCES 1. Diverging diamond interchange Website: http://www.divergingdiamond.com/ 2. FHWA, Highway Capacity Manual (HCM) Methodology for Alternative Intersections / Interchanges.Website: http://www.fhwa.dot.gov/publications/research/operations/datamodelsims/13083/index.cfm 3. NCHRP 03-113, Guidance for Traffic Signals at Diverging Diamond Interchanges and Adjacent Intersections. Website: http://apps.trb.org/cmsfeed/TRBNetProjectDisplay.asp?ProjectID=3631 4. Zhang W. and Kronprasert N., Unlock Diverging Diamond Interchange’s Capacity Potential Using Relaxed Bowtie Design at Adjacent Signalized Intersection, Proceedings, 1st TRB Alternative Intersection & Interchange Symposium, Salt Lake City, Utah, July 20-24, 2014. 5. Zhang W. and Kronprasert N., Unlock DDI’s capacity by re-routing left-turns at nearby intersections. Proceedings, 2nd International Conference on Access Management, Shanghai, China, Sept. 22-24, 2104. 6. Gilbert Chlewicki, Variations of The Diverging Diamond Interchange, Presented at ITE 2010 Annual Meeting, Vancouver, Canada, August 8-11, 2010. 7. 1st TRB Alternative Intersection and Interchange Symposium, Salt Lake City, UT, July 20-23, 2014

14