Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

46 To study and find the conditions amenable to dedicating lanes for CAV users, the team con- ducted a modeling and simulation-based study of CAV driver behavior on DLs on a selected set of diverse case-study sites. This chapter details the process that was used to select the case study sites based on the project objectives. The team identified a set of evaluation criteria to assess the case study sites. Figure 4.1 presents the overall approach to identifying and selecting the case study sites used for modeling CACC DLs. As shown in Figure 4.1, a set of initial candidate case study sites was created based on the team members’ extensive experience with modeling managed lanes and CACC applications. Evalua- tion criteria pertaining to case study site characteristics, managed lane characteristics, and CAV modeling feasibility were developed to down-select these case study sites to two or three that could help define guidelines for agency use in determining whether their specific applications would merit lane dedication. Case study site characteristics include features that define their operational and geographic characteristics, demand, modes, ITS strategies, and the existence of managed lanes. Managed lane characteristics include the features of the existing (or proposed) managed lane facility. These characteristics include operational rules, priority conditions, allowable modes, and access features. The team used CAV modeling feasibility to rank the test sites. 4.1 Initial List of Candidate Sites The team analyzed nine case study sites that were available for use in the modeling effort. Each candidate site represented a simulation-based corridor model for which a managed lane facility existed or had been proposed. The map in Figure 4.2 shows the initial candidate case study sites. Because of map scaling, some candidate test beds overlap (i.e., the candidate sites in St. Paul and Minneapolis, Minnesota, and in Maryland and Northern Virginia). Table 4.1 shows the preliminary list of case study sites that were evaluated and assessed for their effectiveness in achieving the project goals. The next section presents a brief description of the geographic and modeling characteristics of these candidate case study sites. 4.1.1 I-66 Corridor, Northern Virginia The candidate site on the I-66 corridor in Fairfax, Virginia, starts from the outside of the Capital Beltway (I-495) and extends for 13 miles as a 4-lane freeway segment that includes an HOV lane on the left-most lane and stretches to the west all the way through the interchange with US-29 to SR-234 (see Figure 4.3). C H A P T E R 4 Case Study Site Selection

Case Study Site Selection 47 This suburban test site includes six interchanges and two dedicated on-and-off ramps for an HOV lane that is separated from the GPLs. The average distance between interchanges is approximately 1.2 miles, yielding 0.6 miles and 2 miles of minimum and maximum interchange spacing, respec- tively. The test site experiences recurring congestion caused by high directional daily demand every weekday for the eastbound lanes (i.e., toward Washington, D.C.) during the a.m. peak and the west- bound lanes (i.e., toward Fairfax, Virginia) during the p.m. peak. Between 2:00 p.m. and 8:00 p.m., traffic volumes of the test bed range from 900 vphpl to 2,100 vphpl and include approximately Initial List of Case Study Sites Feasibility of Modeling CAV Applications Characteristics of the Managed Lane Facility Characteristics of the Case Study Sites Selected Case Study Sites Figure 4.1. Selection process for the case study sites. Source: NCHRP 20‐102(08) project team; base map from www.HERE.com. Figure 4.2. Initial candidate case study site mapping.

48 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles 1,500 vphpl of peak HOV traffic volumes. This simulation model is currently available in the U.S.DOT’s Open Source Application Development Portal for academic/research use. Based on field observations, the existing simulation model includes a traffic stream with varying vehicle compositions (FHWA Class 4 and above). The existing freeway deploys several ITS strategies along the corridor—hard shoulder running, lane use control signals, VMS, and advanced ramp metering. US-29 is a parallel arterial and an alternate route to I-66 and is acces- sible via the six interchanges included in the existing model. Currently, the parallel roadway is not included in the simulation model. The I-66 managed lanes operate as far-left, single-lane, time-of-day HOV-2 lanes in both eastbound and westbound directions. User-type restrictions along the existing HOV-2 lanes allow only vehicle classes with two or more vehicle occupancy requirements. The existing managed HOV-2 lanes operate on a time-of-day basis with restrictions applying during a.m. and p.m. peak periods on weekdays. No physical barrier separates the managed HOV-2 lanes from the mixed-use lanes. Currently, only double solid white lane markings are used to No. Case Study Corridor Location Length of Corridor Freeway Average Annual Daily Traffic (AADT) Range 1 I- 66 Northern Virginia 13 150,000–160,000 2 US 101 San Mateo, California 8.5 200,000–250,000 3 I-15 San Diego, California 22 250,000–300,000 4 I-35 MnPASS Lanes St. Paul, Minnesota 15 39,000–125,000 5 I-94 St. Paul–Minneapolis, Minnesota 14 132,000–179,000 6 I-290 Managed Lanes Chicago, Illinois 14.5 159,000–211,000 7 I-75 HOV Lanes Detroit, Michigan 18.5 105,000–180,000 8 I-270 Corridor Maryland 26 175,000–270,000 9 I-95 Express Lanes Miami, Florida 20 94,000–260,000 Table 4.1. Initial candidate case study sites. Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Figure 4.3. I-66 case study site coverage.

Case Study Site Selection 49 separate the lanes and to indicate no lane changing and no access. Access points between the dedicated HOV-2 lanes and the mixed-use lanes are permitted only along areas with dashed lane striping. The I-66 also has hard shoulder running lanes on the far-most right lanes in both directions. These lanes operate from 5:30 a.m. to 11:00 a.m. in the eastbound direction and from 2:00 p.m. to 8:00 p.m. in the westbound direction. Lane utilization is indicated via VMS, which show a green arrow for permitted use and a red cross for closed for use unless exiting. The simulation case study was developed and calibrated using the PTV Vissim micro- simulation software, which allows external API-based control of simulation components, including driver behavior, making it a good candidate for CAV modeling. Driver behavior was calibrated to replicate field-observed corridor travel time, speed, and traffic volume. Freeway speed and volume information for the case study site are available, in 5-minute inter- vals and classified by lanes, via the FHWA’s Saxton Transportation Operation Laboratory (U.S.DOT 2018). The existing ITS strategies along the corridor were included in the traffic simulation model. 4.1.2 US-101 Corridor, San Mateo, California The US-101 case study site is located within the County of San Mateo, California, and stretches from Redwood City to the City of Burlingame. The length of the modeled US-101 freeway facil- ity is approximately 8.5 miles, with a parallel arterial, El Camino Real (SR-82), of similar length. Drivers can divert to the parallel arterial via seven possible interchanges. The extent and coverage of the US-101 corridor model is illustrated in Figure 4.4. Source: NCHRP 20-102(08) project team; base map from www.openstreetmap.org. Figure 4.4. US-101 case study site coverage.

50 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles 4.1.3 I-15 Corridor, San Diego, California The I-15 case study site is made up of a 22-mile stretch of the I-15 corridor facility and associ- ated parallel arterials. It extends north-to-south from the interchange with SR-78, just below the City of Escondido, California, to the interchange with Balboa Avenue, approaching San Diego, California. This facility is shown in Figure 4.5. The corridor passes through a suburban area. A network of arterials runs concurrent with the I-15 freeway, and drivers on the Interstate are able to divert via 18 possible interchanges, including Pomerado Road and Ted Williams Parkway. The I-15 corridor has a ramp metering information system and traffic light synchroniza- tion, both used for an active traffic demand management system. Speed and volume detectors are located throughout the freeway. Existing ITS strategies, specific to active traffic demand management systems, were included in the existing model. Within the limits of the simulation model, congestion during peak periods has been recorded to be approximately 50% higher than Source: NCHRP 20-102(08) project team; base map © San Diego Geographic Information Source (SanGIS) 2015, accessed at San Diego Association of Governments (SANDAG) website (www.sandag.org). Legend ICM Network Figure 4.5. I-15 case study site coverage.

Case Study Site Selection 51 off-peak hours in the peak direction. The measured daily VMT varies from the average value of all days observed by no more than a 10% margin. The simulation model includes varying heavy vehicle percentages within the traffic stream by time of day. No transit vehicles were included in the model. The I-15 freeway also includes express lanes that are separated by a concrete median barrier. These lanes are located between the GPLs northbound and southbound. The median barriers are moveable to manage congestion during peak hours. The standard lane configuration is two northbound lanes and two southbound lanes. This configuration can be changed to three south- bound lanes and one northbound lane to mediate peak-hour traffic demand. Currently, these are the only two-lane configuration choices available. The managed lane facility operates as HOT lanes using distance-based dynamic pricing. Motorcycles and all vehicles with two or more occupants can access the express lanes with no charge. SOVs also are allowed to access the express lanes, but pay a fee. Heavy vehicles (Class 4 and above) are restricted from the express lane facility. Designated ramp entrances and exits to this facility exist to and from SR-163, the I-15 south GPLs, and the I-15 north GPLs. There are two entrances and two exit flyover access ramps near the center of the express lanes facility granting direct access to and from SR-56. There are six access points each in the northbound and southbound directions between the express lanes and the GPLs. The simulation case study site was developed and calibrated using Aimsun microsimulation software. This software allows external API-based control of simulation components, including driver behavior, making it a good candidate for CAV modeling. Driver behavior was calibrated to replicate field-observed corridor travel time, speed, and traffic volume. Roadway speed and volume data were available through the Caltrans Performance Measurement System (PeMS) by specific days with precision of 1-minute intervals (Caltrans 2018). The detector data also classi- fies the traffic volume by lanes. 4.1.4 I-35E MnPass Lanes, St. Paul, Minnesota The I-35 MnPASS lanes that pass through the dense urban area of St. Paul, Minnesota, also represent conditions for assessing feasibility of dedicating lanes to CAVs. On the northern half of the 15-mile study corridor, the managed lane freeway transitions to suburban and rural struc- tures. The freeway corridor also contains a system-to-system interchange with I-694 where the two freeways run concurrently for approximately 1 mile. This facility is shown in Figure 4.6. Existing calibrated models in both CORSIM and Vissim formats are owned by the Minnesota Department of Transportation (Minnesota DOT). Traffic count and speed data detection by lane is archived daily. For this study, traffic count and speed data along the ramps and mainline were obtained through Minnesota DOT’s Regional Traffic Management Center detector data (Minnesota DOT 2017). Turning-movement counts at the ramp terminals, which were available from previous signal retiming projects at most of the study area interchanges, also were used in the model. This corridor experiences typical a.m. and p.m. peak-hour commuter demand and mild to moderate congestion, with the southbound traffic experiencing heavier demand during the a.m. peak and the northbound traffic experiencing greater demand during the p.m. peak Outside of these commuter rush hours, demand drops off considerably, and traffic moves under free-flow conditions. Alternate arterial routes exist (i.e., US-61), and even the regional freeway network provides alternate routes; however, these regional routes were not included in the scope of this microsimulation project. Traffic on the corridor is largely commuter traffic and mostly consists

52 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles of passenger vehicles with a small proportion made up of commercial trucks. On this corridor, transit is not significant enough to affect operations greatly; therefore, transit was not explicitly modeled. Existing ITS strategies include VMS, ramp metering, and traffic speed/volume detec- tors feeding to the traffic management centers. The existing managed lane facility is present on the southern half of the study corridor (south of I-694) and consists of a single lane in each direction. Studies are being performed to assess the feasibility of expanding the current facility to the north. HOV, transit vehicles, and motorcycles can use the current facility at no charge, whereas SOVs pay to use the facility. Large commercial vehicles (with more than two axles and weighing more than 26,000 pounds) are restricted from the managed lane during peak hours but can use the managed lane during non-peak travel times. The facility is currently operated as a managed lane during commuter rush and as a GPL during off-peak hours. A solid double white line indicates that access to the managed lane is restricted. Frequent access areas, which also are major weaving areas, are indicated with striping. Buses are permitted to run along the shoulders of I-35E in the northbound and southbound direction for an approximate 2-mile stretch north of I-694 and an approximate 3-mile stretch south of I-694. Buses can use the outside shoulder along these stretches of I-35E when congestion slows travel speeds to 35 mph or slower. Buses using the shoulder may only exceed adjacent general traffic speeds by 15 mph. Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Legend Study Area Interchange Figure 4.6. I-35E case study site coverage.

Case Study Site Selection 53 Driver behavior was calibrated in CORSIM per Minnesota DOT guidelines. Vissim models were calibrated as well to replicate existing travel speeds and congestion levels during the a.m. and p.m. peak-hour rush periods. 4.1.5 I-94 Managed Lanes (Proposed), Minneapolis, MN The proposed I-94 managed lane facility will be in a dense urban area between Minneapo- lis and St. Paul, Minnesota, and will include system interchanges with I-35W and I-35E (see Figure 4.7). The length of the facility included in the simulation model is approximately 14 miles, extending from I-394 on the west to US-61 on the east, with 32 interchanges modeled. Traffic on the existing corridor is largely commuter traffic and mostly is made up of passenger vehicles with a smaller proportion of commercial trucks. Transit is not a significant enough component of this corridor to impact operations greatly; therefore, transit was not explicitly modeled. The corridor experiences typical a.m. and p.m. peak-hour commuter demand and moderate to heavy congestion during these peak periods. Alternate arterial routes are available with lim- ited river crossings, but the alternate routes were not included in the microsimulation modeling of this project. The facility currently has VMS, ramp metering, and traffic speed/volume detec- tors. The proposed project is to construct an expansion of the managed lane (MnPASS) system to include this corridor. The managed lane would be a single lane in each direction. Buses would be permitted to run along the shoulders of I-94 between Highway 280 and Downtown St. Paul. As with other bus shoulder-running applications in the area, buses would be permitted to use the outside shoulder when congestion slows travel speeds to 35 mph or slower, with bus speeds limited to no more than 15 mph faster than the adjacent general-purpose traffic. The proposed facility would be operated as a HOT lane, allowing free access to high-occupancy passenger cars, transit vehicles, and motorcycles. SOVs would be able to access this facility with a fee. Heavy vehicles are restricted from access to the facility. The proposed operating rules involve time-of-day plans, operating each managed lane as a mixed-use lane during off-peak hours while operating it as a HOT lane during peak periods. Driver behavior was calibrated in CORSIM per Minnesota DOT guidelines. Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Figure 4.7. I-94 case study site coverage.

54 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles The existing calibrated models were in CORSIM format and owned by the Minnesota DOT, which presented significant challenges to modeling CAV behavior due to limitations in uti- lizing external API to code varying driver behaviors. Traffic count and speed data along the ramps and mainline were obtained through Minnesota DOT’s Regional Traffic Management Center detector data (Minnesota DOT 2017). Traffic count detection is by lane and is archived daily. Turning-movement counts at the ramp terminals were available from previous signal reti- ming projects at most of the study area interchanges. At interchanges where turning-movement counts were not available, new turning-movement counts were collected. 4.1.6 I-290 Managed Lanes, Chicago, Illinois The I-290 facility runs through a dense urban area in Chicago and metropolitan communi- ties to the west of downtown Chicago (see Figure 4.8). A managed lane facility, which includes a single lane in each direction, is proposed on this corridor. The length of the facility included in the simulation model is approximately 14.5 miles, extending from I-294 on the west to I-90 on the east, with 21 interchanges modeled. The corridor experiences typical a.m. and p.m. peak- hour commuter demand and heavy congestion during these 2- to 3-hour peak periods. Outside of these peaks, traffic demand reduces enough to allow for free-flow operations along I-290. There are no restrictions on transit or heavy vehicles on the existing facility and, due to left-side entrance/exit ramps along the corridor, commercial heavy vehicles can utilize all lanes. Alternate arterial routes are available (Roosevelt Road being the primary alternate route), but the alternate routes were not included in the microsimulation modeling of this project. Traffic on the I-290 case study corridor is largely commuter traffic, consisting mostly of pas- senger vehicles with a smaller proportion of commercial trucks. Transit is not a significant com- ponent of this modeled corridor. Commuter rail is present, running immediately adjacent to Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Figure 4.8. I-290 case study site coverage.

Case Study Site Selection 55 I-290 and within the I-290 median for the eastern half of the study corridor; however, due to limited interaction with the freeway, transit—including commuter rail—was not included in the microsimulation modeling. VMS are present indicating travel time along the corridor. Ramp metering, closed-circuit television (CCTV), and traffic speed/volume detectors also are present on the corridor. The operating rules proposed for this managed lane facility are HOV and HOT time-of-day restrictions by which the facility operates as a managed lane during peak hours and as a GPL during off-peak hours. Access to and from the managed lane is proposed to be indicated using dashed white line pavement striping only, and restricted (no) access is to be indicated with a solid double white line. Existing calibrated models in the Vissim model format were owned by the Illinois Department of Transportation (Illinois DOT). Traffic count and speed data along the ramps and mainline were used for calibration of the models and obtained from the Illinois DOT’s detector database. Traffic count detection is by lane and is archived daily. Turning-movement counts at the ramp terminals were collected at most of the study area interchanges as part of the project. 4.1.7 I-75 HOV Lanes, Detroit, Michigan The I-75 freeway corridor is based in a dense urban area immediately north of the Detroit city limits. This area includes a major system interchange with I-696. The length of the facility included in the simulation model is approximately 6 miles, extending from M-102 on the south end to 12 Mile Road on the north end, with five interchanges included (see Figure 4.9). The microsimula- tion model was prepared for detailed analysis of a subarea of a larger (18.5-mile) corridor being studied for the addition of a managed lane (from M-102 on the south end to M-59 on the north end). The corridor experiences typical a.m. (southbound) and p.m. (northbound) peak-hour commuter demand and moderate congestion during these peak periods. Outside of these peaks, traffic demand reduces enough to allow for free-flow operations along I-75 during most of the day. A managed lane facility with a single lane in each direction was proposed for immediate construction. The construction would require widening along the corridor for the addition of this lane in each direction. The proposed managed lane facility would extend 12.5 miles, from SR-59 to approximately 12 Mile Road. This would be the first managed lane facility along the freeway system in Michigan. Operational rules for this managed lane facility would be based on a time-of-day HOV restriction, by which the facility would operate as a managed lane during peak hours and as a GPL during off-peak hours. Alternate arterial routes are available, with Woodward Avenue being the primary alternate route; however, the alternate routes were not included in the microsimulation modeling of this project. Traffic on this corridor is largely commuter traffic, consisting mostly of passenger vehicles with a smaller proportion of commercial trucks. Transit is not a significant component of this modeled corridor. Existing ITS strategies and technologies along the corridor include VMS, CCTV, and traffic-count stations. Areas providing access to and from the managed lane are proposed to be indicated by dashed white line striping only. Restricted areas (with no access to or from the managed lane) are pro- posed to be indicated by a solid double white line. The entire corridor (18.5 miles) was given a macroscopic Highway Capacity Manual analysis of the basic freeway segments, merge/diverge areas, and weave areas. A more detailed micro- simulation analysis was conducted for a 7-mile section containing the system interchange with I-696 and proposed ramp-braiding alternatives. The microsimulation was conducted in Vissim (Version 6), and traffic count data was obtained from the Michigan Department of

56 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles Transportation (Michigan DOT) traffic count web portal for ramps and mainline counts along the corridor (Michigan DOT 2018). The model was set up as a ramp and mainline model only. Full interchange operations were not modeled. Speed and congestion data used for calibration were obtained from the Regional Integrated Transportation Information System (RITIS) maintained by the University of Maryland (CATT Lab 2018). 4.1.8 I-270 Corridor, Maryland I-270 is a 34.7-mile auxiliary Interstate Highway that travels between I-495 (the Capital Belt- way) just north of Bethesda, in Montgomery County, Maryland, and I-70 in the city of Frederick in Frederick County, Maryland. The corridor consists of a 32.60-mile main line plus a 2.10-mile spur that provides access to and from southbound I-495 (see Figure 4.10). Most of the southern part of the route in Montgomery County passes through suburban areas around Rockville and Gaithersburg. This portion of I-270 is up to 12 lanes wide and consists of a local-express lane configuration as well as HOV lanes that are in operation during peak travel times. North of the Gaithersburg area, the road continues through the northern part of Montgomery County as a 6- to 8-lane highway with an HOV lane in the northbound direction only. Farther north, I-270 continues through rural areas into Frederick County and toward the city of Frederick as a 4-lane freeway. The modeled length is approximately 26 miles. Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Figure 4.9. I-75 case study site coverage.

Case Study Site Selection 57 This corridor experiences very little diversity in demand conditions and traffic patterns, and currently operates at a high level of congestion throughout most typical days. The only parallel alternate route is SR-355, an arterial corridor with signalized intersections and significant busi- ness activity. Modes of transportation included in the model were passenger cars, buses, and heavy vehicles. The parallel alternative routes were not included in the model. The model included imme- diate facilities, such as highway interchanges and immediate signalized intersections at the interchanges. The current ITS infrastructure consists of speed detectors, video monitoring infrastructure, and VMS. The Maryland DOT is in the process of procuring an innovative congestion-management upgrade project ($100 million) that could introduce a range of new technologies/strategies. The facility’s managed lane is currently a single HOV lane in each direction. HOV operating restrictions apply during the traffic peak period in the peak direction only. The HOV lane is concurrent with other lanes and is distinguished by special pavement markings. Vehicles can access the managed lane from the GPL throughout the entire facility with no restrictions. The simulation platform used to develop the model was Vissim. No existing ITS strategies were included in the existing model. 4.1.9 I-95 Express Lanes, Miami, Florida Interstate 95 (I-95) is a key component of the Interstate Highway System, running along the east coast of the country from Miami, Florida, to the U.S.-Canada border in eastern Maine. The study segment of this facility is an urban freeway with directional commuter traffic flows that runs through the densely urban cores within Miami–Dade and Broward Counties in South Florida Source: NCHRP 20-102(08) project team; base map data © 2018 Google. Figure 4.10. I-270 case study site coverage.

58 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles (see Figure 4.11). The managed lane section currently runs approximately 20 miles between Davie Road, near Downtown Ft. Lauderdale, to just north of Downtown Miami at SR-836. I-95 operates at high levels of congestion throughout most of the day, with concentrated congestion at various bottlenecks during off-peak hours while distributed throughout the facility during peak periods. The closest parallel facility is the signalized arterial of US-1/ Federal Highway/Biscayne Boulevard. Another parallel alternative route located along the northern portion of the facility is the Florida Turnpike. The various transportation modes include passenger cars, heavy vehicles, and buses. Heavy vehicles are restricted from using the express lanes. Existing ITS strategies used along the I-95 corridor include VMS, video monitoring, and various detectors. The I-95 express lanes represent a conversion from HOV to HOT operation and were imple- mented to provide more reliable trip times for corridor users. The facility allows for toll-free access for HOV3+ users and transit but requires carpools to pre-register. SOVs can access the lane by paying a toll that is assessed on a dynamic basis in response to congestion. As volumes increase, so does the price for access. Currently, the maximum rate for an SOV is $1.50 per mile or $10.50 over the full length of the express lanes. This cap may be raised if the LOS on the facility Source: NCHRP 20-102(08) project team; base map from Florida DOT (www.95express.com). Figure 4.11. I-95 case study coverage.

Case Study Site Selection 59 consistently declines below 45 mph over a 90-day period, a policy that is largely the result of the project’s initial funding through the Federal Urban Partnership Agreement. The Florida DOT estimates that about 2% to 3% of the traffic in the express lanes is travelling toll free. The managed lanes are separated from the GPLs by flexible delineator posts. The model cur- rently includes four northbound entrances, five southbound entrances, four northbound exits, and four southbound exits between the managed lanes and GPLs (see Figure 4.12). The simulation platform used to develop this network was Vissim. The Vissim modeling included the GPLs, the managed lanes, and the individual interchange operations. 4.2 Evaluation Criteria This section describes the evaluation criteria used for modeling the initial candidate case study sites. These evaluation criteria were ranked as being of low, medium, or high priority based on their relevancy in assessing DL conditions. For example, model availability is considered a high-priority criterion, whereas having a moderately sized facility is of low priority. Based on the relative importance of each evaluation factor, the team used weighted scoring when ranking case study sites. 4.2.1 Case Study Site Characteristics The team used eight evaluation criteria to identify characteristics and rank the case study sites. This evaluation included characterization of the geographic and operational conditions that exist in the test sites. 4.2.1.1 Geographic Characteristics Managed lanes generally are an urban/suburban roadway feature; hence, it is desirable that the final selected case study sites represent reasonable use of dedicated/managed lanes in or near metropolitan area conditions. Drivers in larger metropolitan areas will be more accustomed to regularly encountering recurring or nonrecurring congestion. In larger metro politan areas, congestion will tend to be more ubiquitous and bidirectional. For this criterion, the characteristics assessed reflected diverse sites that ranged from less urban to more urban in terms of number of lanes, AADT, and location. This evaluation criterion was given medium priority in the case study selection process because managed lanes are mostly an urban feature (Figure 4.13). 4.2.1.2 Availability of Data/Case Study Site Model Successful modeling of CACC in DLs depends on the model’s closeness to the real world. Hence, availability of a calibrated case study site model is of extreme importance. The team selected case study site models that were available for use in a calibrated state. To evaluate the impacts of DLs for CACC-equipped vehicles, the case study sites needed to be validated and cali- brated using historical, near real-time, and real-time data. The data had to represent a case study site’s geographic and temporal scope as well as characteristics such as existing ITS infrastructure and managed lane configurations. The availability of case study models and the associated cali- bration data was a high-priority evaluation criterion, given the importance of a fully calibrated simulation model in assessing realistic and credible benefits and sensitivity parameters of CACC application on a DL facility. The research team gave preference to models that were available in an open-source portal such as the U.S.DOT’s Open Source Application Development Portal (OSADP) or the U.S.DOT Data Repository, as well as models that were available upon request from local agencies (see Figure 4.14).

60 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles Source: NCHRP 20-102(08) project team; base map data from Florida DOT (2018). Figure 4.12. I-95 express lane configurations and access points.

Case Study Site Selection 61 Figure 4.14. Case study characterization based on model availability. Figure 4.13. Case study characterization based on geographic characteristics. 4.2.1.3 Diversity in Demand/Operational Conditions Operating demand of a corridor facility determines the operational conditions for the drivers. For this case study selection, we assess the demand in terms of traffic volumes over the entire case study site. Traffic demand for low (uncongested), medium (near capacity), and high (congested) levels will yield different traffic patterns and a wide range of cases to assess and compare their performances. Although low-demand conditions do not present challenging conditions for the deployment of CAV applications, having a variable demand would allow assessment of impacts under different saturation rates. Hence, the selection included a case study site with varying traffic demand, or multiple case study sites that represent different demands. Having differing demand conditions is important to analyze the sensitivity of DLs under various saturation rates, but the demand can be scaled easily from existing models. Therefore, this criterion was consid- ered medium priority (see Figure 4.15). 4.2.1.4 Length of Facility The length of a DL facility relative to the overall case study site is an important factor in gaug- ing its influence on the overall network, parallel corridor, and parallel arterials. The length of the facility corresponds directly to the proportion of benefits or disbenefits imposed on the assess- ment boundary, which is defined as the limits of the roadway facility that have been included in the assessment. Effects like the proportion of a given trip utilizing the DL versus not using the DL can be compared between facilities with longer DLs versus shorter DLs. At the same time, modeling the CACC application entails computationally intensive driver-behavior capture to an external interface and trajectory implementation, and larger models can become difficult to Figure 4.15. Case study characterization based on operational demand.

62 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles Figure 4.16. Case study characterization based on routing features. model. The team selected medium-sized facilities to enable full evaluation of trip-based perfor- mance measures as well as manage the computation size. This evaluation criterion was given a medium priority in the case study selection. 4.2.1.5 Availability of Alternate Routes One consideration in assessing CACC DLs’ impact on non-users is the availability of alternate routes, such as parallel arterials. The case study sites were assessed to determine whether a parallel route was explicitly included in the model and whether vehicle rerouting was possible through these alternate routes. For the case study selection, the team prioritized models that included alternate routes. Because most of the candidate models had been developed for corridor analysis, however, only a few might have included alternate parallel routes. This evaluation criterion was given a medium priority in the case study selection due to this limitation (see Figure 4.16). 4.2.1.6 Diversity in Modes Varying modes of transportation within the traffic stream composition is an important consideration due to its impact on traffic flow characteristics. Freeways and interstates in urban environments have a significant composition of heavy and transit vehicles unless heavy and/or transit vehicle access is restricted. Heavy and transit vehicles have different accel- eration and deceleration profiles compared to passenger cars. For evaluation purposes, the selected case study sites needed to have varying vehicle type composition or be restricted by time-of-day access to heavy and transit vehicles so that their impacts could be assessed. This evaluation criterion was given a medium priority in the case study selection (see Figure 4.17). 4.2.1.7 Existence of Managed Lanes Managed lanes commonly are used within urban metropolitan areas. Types of managed lanes may include HOV lanes, HOT lanes, and express toll lanes (ETLs). One important factor for consideration is the traffic and safety impacts of using several types of managed lanes on the same corridor. Other impacts include mixed use of managed lanes, which may include dedicated CACC with HOV, HOT, and ETLs. These scenarios could be compared to a scenario with com- plete conversion of existing managed lanes to dedicated CACC lanes. For this study, the research team gave preference to case study sites with existing managed lanes or where managed lanes had been proposed for deployment in the near future. This evaluation criterion was given a medium priority in the case study selection (see Figure 4.18). Figure 4.17. Case study characterization based on modal diversity.

Case Study Site Selection 63 4.2.1.8 Existence of ITS Strategies ITS strategies are implemented to maximize roadway carrying capacity and increase safety. Concurrent implementation of ITS strategies with CACC DLs may have either synergistic or conflicting effects on roadway capacity and driver safety. For example, CACC is expected to work synergistically with dynamic speed limits because it improves the string stability of CACC platoons. The research team assessed the case study sites to determine whether ITS strategies existed and were modeled in the available simulation model. These existing ITS strategies could then be screened for conditions that can cause synergies or conflicts with CACC applications. This evaluation criterion was given a low priority in the case study site selection because currently implemented ITS strategies may or may not exist in conjunction with CACC implementation. 4.2.2 Managed Lane Characteristics The existing or proposed characteristics of the managed lanes for each of the case study sites also were assessed. Specifically, the following five characteristics were used for case study site scoring: managed lane geometry, user types, operating rules, physical barrier types, and diversity in access point configurations. 4.2.2.1 Managed Lane Geometry The number of lanes available for use as managed lanes is a critical factor to assess the capacity benefits of additional lanes. Capacity impacts are an important determining factor in deciding on the implementation of additional lanes due to roadway widening or hard running shoulder uses. Addi- tional lanes mitigate the “snail” effect by which the slowest-moving vehicle in the managed lane can govern the speed of the entire lane. Additional lane design should complement the access manage- ment strategy to accommodate traffic safety and capacity due to lane changes. Case study sites with a diverse number of DLs and varying roadway geometries were preferred so that these impacts could be assessed. This evaluation criterion was given a low priority in the case study selection. 4.2.2.2 User Types Within the selected case study sites, existing managed lanes (e.g., HOT lanes, HOV lanes, and ETLs) with a mix of user-types (e.g., SOVs, HOVs, transit vehicles, and heavy vehicles) were pre- ferred. To assess the benefits and disbenefits of imposing future restrictions on current user types (e.g., through conversion of existing managed lanes to dedicated CAV lanes) the team identified case study sites with a diverse user base. Among the project objectives, a major consideration was to evaluate the feasibility of mixed lane use by CAV vehicles and non-CAV vehicles. Hence, this evaluation criterion was given a medium priority in the case study selection (see Figure 4.19). 4.2.2.3 Operating Rules Managed lanes can have a variety of operating rules to manage the facility for both operational and safety reasons. For example, time-of-day and vehicle-class access restrictions commonly are used along certain managed lane facilities. These operating rules influence traffic patterns Figure 4.18. Case study characterization based on managed lanes.

64 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles throughout the day at the imposed area. Other operating rules may include the enforcement of left-lane passing only laws, which may involve safety concerns for vehicles that must pass mul- tiple platooned vehicles to find an acceptable gap for a lane change. The team categorized the testbed operating rules as: a. Time-of-day operation, wherein lanes operate as managed lanes only during peak hours. During non-peak hours, no lanes are dedicated to special vehicle categories such as HOVs or toll-paying SOVs. b. Time-of-day pricing, wherein lanes always operate as managed lanes, but the pricing depends on the time of day and follows a schedule. This category includes managed lanes for which off-peak usage may be free. c. Dynamic congestion pricing, wherein the usage fee for the managed lanes is determined based on existing travel conditions. This evaluation criterion was given a medium priority in the case study selection because the variance in these factors is somewhat limited (see Figure 4.20). 4.2.2.4 Physical Barrier Types Managed lanes that are separated from the GPLs by physical barriers like flexible delineator posts or concrete median barriers may have different posted speed limits from the GPLs. The potential difference in speed limits distinguishes managed lanes with physical barriers from managed lanes that are separated only by pavement striping. Differences in posted speed limits will have considerable effects on roadway capacity, traffic characteristics, and driver behaviors. The impacts on driver behaviors and traffic characteristics caused by varying physical barrier separations can be assessed and compared to the impacts on driver behaviors and traffic charac- teristics at managed lanes with no physical barrier separations. Accordingly, the research team gave preference to a testbed portfolio that included varying barrier types. This evaluation crite- rion was given a medium priority in the case study selection (see Figure 4.21). 4.2.2.5 Diversity in Access Point Configurations Access points to and from the DLs have a significant impact on the roadway capacity. The frequency of available access points along a DL directly correlates to drivers’ wayfinding and Figure 4.19. Case study characterization based on managed lane user characteristics. Figure 4.20. Case study characterization based on managed lane operating rules.

Case Study Site Selection 65 access to the facility. Driver lane-changing behaviors on both GPLs and DLs will be affected by advanced knowledge of access availability. Treatments that mediate the impacts of traffic turbu- lence caused by weaving vehicles making lane changes to enter or exit the DLs also are impor- tant factors to consider. The two types of access point configurations categorized by the team were continuous access and restricted access, with the latter type defined by access point frequency, strictness of access point location, and access section length. Access point frequency and weave management treatments, such as shorter access lengths (which challenge weaving movements), could be compared to assess the treatments’ impacts on both the GPLs and DLs. This evaluation criterion was given a medium priority in the case study selection (see Figure 4.22). 4.2.3 CAV Modeling Feasibility The feasibility of modeling CAVs also represented an important set of scoring criteria. Spe- cifically, the case study site needed to be modeled in an environment that permitted model- ing of customized vehicle and driver behavior. Specific feasibility criteria considered by the research team were the possibility of external programming interface and available driver behavior calibration data. 4.2.3.1 Possibility of External Programming Interface For the purposes of this project, the simulation environment needed to allow for modeling CACC driver and automatic car-following behaviors. The environment needed to allow for the inclusion of external API or a software-in-the-loop-system, if the CACC driver behavior was not already readily available with the model. External API also was required to query and receive vehicle parameters that were not already catalogued for analysis. This evaluation criterion was given a high priority in the case study selection because modeling CACC applications without external API was not possible (see Figure 4.23). 4.2.3.2 Available Driver Behavior Calibration Data The case study sites selected would need driver behavior calibration data specific to the local environments to allow for a detailed replication of conventional local driving behavior. A case study model that closely mimicked existing driver behavior would provide for a high-fidelity rep- resentation of the case study area and better comparisons among analyzed scenarios. Assessing Figure 4.21. Case study characterization based on managed lane separation. Figure 4.22. Case study characterization based on managed lane access features.

66 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles the traffic flow performance under a mixed-use case including CAV and non-CAV was critical to determining their traffic impacts, so this criterion was given a high priority. 4.3 Selected Case Study Sites The project team used the scoring criteria discussed in the preceding sections of this chapter to score and rank the nine candidate case study sites. 4.3.1 Case Study Site Scoring The study team developed a comprehensive scoring process to rank the initial candidate case study sites and select a portfolio of case study sites that could be used to effectively model the CAV applications and determine the implications on dedicating lanes to such vehicles. The selected case study sites needed to be able to define guidelines that agencies can use to determine whether their specific applications would merit lane dedication. These guidelines should include different levels of traffic congestion, network connectivity, availability of alternate routes and modes, spacing of access/egress points, truck traffic, and traffic patterns (e.g., core focused versus dispersed). Selecting a single case study site would not be sufficient to model the diversity in conditions that needed to be assessed, whereas modeling numerous test sites would be resource-intensive. Consequently, the team used the evaluation criteria scoring process to select two case study sites. 4.3.1.1 Mapping of Evaluation Criteria The team used the evaluation criteria it had developed to identify a set of 15 parameters (see Table 4.2). The nine initial candidate testbeds were then evaluated based on these 15 parameters. For each parameter, the team identified corresponding site-specific value(s), which are shown in Table 4.2 and Table 4.3. Multiple values were selected for certain parameters that involved a mix of different values. For example, the case study from St. Paul, Minnesota involved a mix of rural, suburban, and urban geographical areas. 4.3.1.2 Scoring of Case Study Sites Once the site-specific value for each parameter had been assessed, the team scored the param- eter based on whether it was least preferred (0) to most preferred (3), as shown in Table 4.4. For example, for availability of model and data, the Northern Virginia test site received a score of 3 because the model is available as open source, whereas case study sites such as the Chicago site received a score of 2. A weighted factor to indicate the priority of that specific evaluation factor was assigned. A weight value of 1 through 3 was used for factors with priority low to high, respectively. The final score of each testbed was calculated as a sum-product of each of the evaluation scores and their corresponding weights (w). Thus, for a testbed (i), the final score (Si) was calculated as follows: ∑= ,S s wi ij jj where j represents the variable evaluation scores. Figure 4.23. Case study characterization based on modeling interface.

Northern Virginia San Mateo, California San Diego, California St. Paul, Minnesota Minneapolis, Minnesota Chicago, Illinois Detroit, Michigan Maryland Miami, Florida Ca se S tu dy S ite C ha ra ct er is tic s Characteristics Urban ● ● ● ● ● ● Suburban ● ● ● ● ● Rural ● Availability of Model and Data Open-source ● ● Available on Request ● ● ● ● ● ● ● Unavailable Demand Levels Low ● ● ● ● ● Medium ● ● ● ● ● ● ● High ● ● ● ● ● ● ● ● ● Size of Model Length (Miles) 13 8.5 22 15 14 14.5 18.5 26 20 Alternate Routing Unavailable Available, but not modeled ● ● ● ● ● ● ● Available and modeled ● ● Modal Diversity Cars ● ● ● ● ● ● ● ● ● Trucks ● ● ● ● ● ● ● ● ● Transit ● ● ● Existing ITS Strategies None Available ● ● ● ● ● ● ● ● ● Managed Lanes Existing ● ● ● ● ● Proposed ● ● ● ● Unavailable Table 4.2. Modeling feasibility given site-specific values for case study site and managed lane characteristics (Part 1 of 2).

Northern Virginia San Mateo, California San Diego, California St. Paul, Minnesota Minneapolis, Minnesota Chicago, Illinois Detroit, Michigan Maryland Miami, Florida M an ag ed L an e Ch ar ac te ri sti cs Characteristics Number of Lanes 1 2 2 1 1 1 1 1 2 User Types HOV ● ● ● ● ● ● ● ● ● HOT ● ● ● ● ● Transit ● ● ● ● ● ● ● ● ● Trucks ● Operating Rules Time of Day Operation ● ● ● ● ● ● ● ● Time of Day Pricing Congestion Pricing ● ● ● Physical Barriers None ● Lane-marking ● ● ● ● ● ● Delineators ● ● Separated Access Point Throughout ● ● Limited ● ● ● ● ● ● ● Fe as ib ili ty Modeling Platform API Unavailable ● API Available ● ● ● ● ● ● ● ● Driver Behavior Not Calibrated Calibrated ● ● ● ● ● ● ● ● ● Table 4.3. Modeling feasibility given site-specific values for case study site and managed lane characteristics (Part 2 of 2).

Case Study Site Selection 69 Parameter Scoring and Criteria 1. Case Study Characteristics Geographic Characteristics 3 = Urban region. 2 = Suburban region. 1 = Rural region. Availability of Model and Data 3 = All models that were available as open-source. 2 or 1 = The score was lowered based on the increasing difficulty of obtaining the model. Demand Levels 3 = Sites that replicate low, medium, and high demand conditions. 2 or 1 = The score was lowered depending on the model’s inability to mimic certain demand conditions. Size of Model 3 = Sites between 7 miles and 14 miles in length. 2 or 1 = The score was lowered for smaller or larger sites owing to the relative increase in complexity/computational intensity of modeling CAV applications at these sites. Alternate Routing 3 = Sites with an available alternate route that also could be modeled. 2 or 1 = The score was lowered when an alternate route was not available for modeling. Modal Diversity 3 = Sites with a diverse modal set (including cars, trucks, and transit). 2 or 1 = The score was lowered when the number of modes was reduced. Existing ITS Strategies 3 = Sites with existing ITS strategies (e.g., ramp metering, hard-shoulder running, variable speed limits). 1 = Sites without existing ITS strategies.* 2. Managed Lane Characteristics Existence of Managed Lanes 3 = Sites with existing managed lanes. 2 = Sites with proposed managed lanes. 1 = Sites with no managed lanes. User Types 3 = Sites that allow all types of users in the managed lanes. 2 = Sites with some restrictions on vehicle types allowed in the managed lanes. 1 = Sites with the greatest restrictions on vehicle types allowed in the managed lanes. Operating Rules 3 = Sites with an operating rule. 1 = Sites without an operating rule.* Physical Barriers 3 = Sites with separation or barriers. 1 = Sites without separation or barriers.* Access Options 3 = Sites with limited-entry managed lanes. 1 = Sites with continuous access.* 3. CAV Modeling Feasibility CAV Modeling Ability 3 = Sites with available API. 1 = Sites without available API.* Driver Behavior 3 = Sites that can be calibrated to realistic driving behavior. 1 = Sites not calibrated to realistic driving behavior.* *Scoring for this parameter did not include a score of 2. Table 4.4. Case study site scoring criteria.

70 Dedicating Lanes for Priority or Exclusive Use by Connected and Automated Vehicles With regard to parameters for which variety was preferred, the case study sites that represented a diverse set of values were given higher scores. For example, St. Paul, Minnesota, received a high score for demand levels because the case study site is subject to varying demand levels. These scores were generated for each parameter and a total score was assessed as shown in Table 4.5. Based on the case study site scores provided in Table 4.5, the top-ranking testbeds were: 1. Northern Virginia, and 2. San Mateo, California. Chapter 5 provides a detailed description of these two testbeds along with details on their calibration data and operational conditions in terms of traffic demand, weather conditions, and occurrence of incidents. W ei gh ts N or th er n Vi rg in ia Sa n M at eo , Ca lif or ni a Sa n Di eg o, Ca lif or ni a St . P au l, M in ne so ta M in ne ap ol is, M in ne so ta Ch ic ag o, Ill in oi s De tr oi t, M ic hi ga n M ar yl an d M ia m i, Fl or id a Ca se S tu dy S ite C ha ra ct er isti cs Geographic Characteristics 2 2 3 2 3 3 3 3 2 3 Availability of Model and Data 3 3 3 2 2 2 2 2 2 2 Demand Levels 2 3 3 2 3 3 3 3 1 1 Size of Model 2 3 3 2 2 3 2 2 1 2 Alternate Routing 2 2 3 3 2 2 2 2 2 2 Modal Diversity 2 3 3 3 2 2 2 2 2 2 Existing ITS Strategies 1 3 3 3 3 3 3 3 3 3 M an ag ed L an e Ch ar ac te ris tic s Existence of Managed Lanes 2 3 2 3 3 2 2 2 3 3 User Types 2 2 2 3 3 3 3 2 2 2 Operating Rules 2 3 3 3 3 3 3 3 3 3 Physical Barriers 2 3 3 3 3 3 3 3 3 3 Access Options 2 3 1 3 3 3 3 1 1 3 M od el in g Fe as ib ili ty CAV Modeling Ability 3 3 3 3 3 1 3 3 3 3 Driver Behavior 3 3 3 3 3 3 3 3 3 3 TOTAL CASE STUDY SITE SCORE 84 82 81 81 75 79 73 67 75 Table 4.5. Site-specific scoring matrix.