Superelevation Criteria for Sharp Horizontal Curves on Steep Grades (2014)

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.

20 S E C T I O N 3 Three types of field studies were conducted as part of this research. Results of the field studies were used as inputs into the vehicle dynamics simulation models and/or served to val- idate the model outputs. Section 4 describes in more detail how the results from the field studies were used in the vehicle dynamics simulation modeling portion of the research. The field studies conducted during this research con- sisted of: • Speed and vehicle maneuver studies, • Instrumented vehicle studies, and • Friction testing. The field studies were conducted in mountainous regions in the eastern and western parts of the United States. This section of the report provides a brief description of the site selection process to identify sites for inclusion in one or more of the field studies, presents the general characteristics of the sites, describes the field studies, and presents the primary results. 3.1 Site Selection The goal of the site selection process was to identify sharp horizontal curves on grades of 4% or more, on a range of roadway types (freeways, other divided highways, and un- divided highways) including high- and low-speed facilities in both rural and urban areas. For site selection purposes, a sharp horizontal curve was defined as a horizontal curve that, under current AASHTO policy, would require super elevation of at least 6% when designed with criteria applicable to a maximum superelevation rate of 8% (i.e., sites with above- minimum-radius curves were included in the field studies). Sharp horizontal curves were also identified for inclusion in the field studies based on the presence of curve warning signs and/or advisory speed signs. It was also desirable to col- lect data in different geographical locations throughout the United States. Initially, the research team identified the states of Pennsylvania, Maryland, and West Virginia in the eastern United States and California, Colorado, Utah, and Washing- ton in the western United States, as potential locations for the field studies. Several steps were taken to identify candidate data collec- tion sites: • Where available, roadway inventory data were obtained to find areas with sharp curves on steep grades in the selected states whose geometrics fit the selection criteria. • Crash data were obtained where available to conduct a system-wide review to find sites with concentrations of lane departure and rollover crashes involving trucks and/ or passenger vehicles. • An online survey was distributed to state trucking asso- ciations in the respective states requesting that their safety offices and/or drivers identify locations which they were familiar with that have sharp horizontal curves on steep grades. • The transportation agencies from the respective states were also contacted for suggestions of candidate data collection sites. Through these various means, close to 100 candidate data collection sites were identified. The research team then con- ducted site selection trips in the states of California, Maryland, Pennsylvania, Washington, and West Virginia to gather detailed geometric data in the field and to select the final sites for inclu- sion in the field studies. Twenty sites were selected for inclusion in one or more of the field studies. Table 4 presents location information, grade, and horizontal curve data for each of the sites. The grade and curve data were obtained from a combination of roadway inventory files, plan and profiles sheets, and field measurements. Seventeen of the sites were located on down- grade sections, while three of the sites were on upgrades. Most of the sites were on freeways, but several sites were on two-lane or multilane highways, and one site was a freeway- to-freeway ramp. The grade represents the maximum grade Field Studies

21 either approaching the curve or in the curve. Similarly, the super elevation represents the maximum superelevation on the curve. In a few cases, the selection criteria were relaxed to include sites in the field studies. Three types of field studies were conducted as part of this research. Table 5 provides a matrix indicating if data from the respective site were used for the given field study. Table 5 also shows if crash data from the site were included in the crash analysis. Section 5 of this report provides details on the crash analysis. 3.2 Speed and Vehicle Maneuver Studies The primary purpose of the speed portion of the studies was to determine, at each data collection site, the distribution of vehicle speeds on the approach tangent and on the curve for both passenger vehicles and trucks. These speed distribu- tions were used in the vehicle dynamics simulation modeling. The primary purpose of the vehicle maneuver portion of the studies was to determine the duration of lane-change maneuvers at sharp horizontal curves on steep grades and the proportion of vehicles that change lanes. The data on duration of lane-change maneuvers were used in the vehicle dynamics simulation modeling, and the proportion of vehi- cles that change lanes indicates the extent or the frequency of such maneuvers. 3.2.1 Data Collection Methodology Speed data were collected using laser guns. Laser guns col- lect speeds and distances of subject vehicles in a continuous fashion. By comparing distances to benchmark locations/ distances, speeds were determined at specific locations along the study site such as upstream of the curve, the beginning of the curve (i.e., PC), and the mid-point of the curve. In general, speed data were collected beginning at least 500 ft upstream of the curve and at least through the mid-point of the curve. Depending on the geometry and available sight dis- tance, one or two laser guns were used to collect speed data for vehicles over the length of the study area. The laser guns were operated by a researcher inside of a vehicle parked on the side of the roadway in a location chosen based on several criteria: • Location was safe; • Data collectors and equipment were situated as incon- spicuously as possible such that they had no (or minimal) impact on driver behavior or desired operating speeds; and Site State Route (direction) County MP Nearest city Roadway type Grade (%) Length of grade (mi) Curve radius (ft) Curve length (mi) emax (%) Spiral Curve direction CA1 CA I-5 (NB) Kern 1.6-2.1 Lebec Freeway −3.1 >1.0 2,000 0.47 2 Absent Left CA2 CA SR 17 (NB) Santa Clara 2.0-3.0 Los Gatos Multilane −6.2 0.25 537 0.21 12 Absent Right CA3 CA SR 17 (SB) Santa Cruz 10.3-9.7 Scotts Valley Multilane −6.3 0.25 575 0.13 8.8 Absent Left MD1 MD I-68 (WB) Garrett 5.5-7.0 Friendsville Freeway −4.1 0.78 1,909 0.31 6 Absent Left MD22 MD I-68 (WB) Washington 74.5-75.0 Hancock East Freeway 6.0 >1.0 1,909 0.42 5.5 Absent Right MD3 MD I-68 (WB) Washington 72.5-73.5 Hancock West Freeway −5.7 0.21 1,900 0.32 4.5 Absent Right PA1 PA I-79 (NB) Washington Interchange I-70/I-79 Washington Ramp −5.0 1.0 Comp 1 0.19 6.25 Absent Right PA2 PA I-80 (EB) Jefferson 79.5-80.5 Brookville Freeway −4.0 0.67 1,637 0.27 8.3 Present Left WA1 WA I-90 (WB) Grant 137.5-138 Vantage Freeway −4.9 >1.0 955 0.23 9.3 Present Right WA2 WA I-82 (WB) Kittitas 15.14-15.94 Ellensburg Freeway −5.0 >1.0 1,600 0.24 10 Absent Left WA3 WA I-82 (WB) Kittitas 4.00-4.63 Ellensburg Freeway −5.0 >1.0 2,400 0.19 7 Absent Right WA4 WA I-82 (EB) Kittitas 21.75-22.5 Ellensburg Freeway −3.8 0.6 1,600 0.33 5.8 Absent Right WA52 WA US 97 (NB) Kittitas 162.7-163 Ellensburg Two-lane 6.0 0.86 1,637 0.19 2 Absent Left WA6 WA I-90 (EB) Kittitas 131.48-31.69 Ellensburg Freeway −2.9 >1.0 2,800 0.33 7 Absent Right WA72 WA US 2 (EB) King 60.0-60.7 Skykomish Multilane 5.9 >1.0 577 0.25 10 Present Left WV1 WV I-77 (SB) Mercer 20.6-21.4 Camp Creek Freeway −4.9 >1.0 1,206 0.50 8 Present Left WV2 WV I-68 (WB) Monongalia 9.9-10.6 Cheat Lake Freeway −5.7 >1.0 1,909 0.49 7.8 Present Left WV3 WV I-79 (SB) Kanawha 2.05-2.5 Mink Shoals Freeway −3.7 0.75 1,146 0.05 8 Present Left WV4 WV I-77 (NB) Kanawha 76.5-78.0 Cabin Creek Freeway −5.2 >1.0 1,041 0.26 8 Present Right WV5 WV I-64 (EB) Kanawha 49.7-50.5 Institute Freeway −5.0 0.58 1,637 0.33 7.2 Present Left 1 Compound curve with four radii: 430 ft, 230 ft, 150 ft, and 310 ft. 2 Upgrade sites. Table 4. Data collection sites and site characteristic information.

22 • Subject vehicles tracked from the rear as they drove away from the laser gun. Figure 9 illustrates the general field setup for the speed studies. At each site speed data were collected over the course of a single day. Speed data were collected for both passenger vehicles and trucks under free-flow conditions. During post- processing of the data, vehicles were grouped into vehicle classes as follows: • Passenger vehicles: – Sedan – Sport utility vehicle (SUV) – Pickup – Van Site Speed data Vehicle maneuver data Instrumented vehicle data Friction testing data Crash data CA1 X X X CA2 X X X CA3 X X X MD1 X X X X X MD2 X X X X X MD3 X X X X X PA1 X X X PA2 X X X WA1 X X WA2 X X X WA3 X X X WA4 X X X WA5 X X X WA6 X X X WA7 X X WV1 X X X X WV2 X X X X WV3 X X X X WV4 X X X X WV5 X X X X Table 5. Data collection sites, field studies, and crash analysis matrix. PC PT 500 ft P1 LG1 LG2 VR1 Data Collection Setup LG1 : Laser gun 1 LG2 : Laser gun 2 (as necessary) VR1: Video recorder 1 VR2: Video recorder 2 (as necessary) P1: Initial point of data collection upstream of curve PC: Point of curvature PT: Point of tangency Figure 9. General data collection setup for speed and vehicle maneuver studies.

23 • Trucks: – Single-unit truck – Tractor semi-trailer truck – Tractor semi-trailer/full-trailer truck (double) While collecting speed data, one or two video cameras were also positioned on the roadside to record vehicle maneuvers at the sites. The field of view for each camera was as follows: • Camera 1—approach and upstream end of the horizontal curve • Camera 2—mid-point and downstream end of the hori- zontal curve The videos from the cameras were reviewed in the office to document the number of vehicles and types (e.g., passenger vehicles and trucks) at the site, the number of vehicles chang- ing lanes, and the duration and direction of the lane-change maneuvers. Figure 10 shows a tractor semi-trailer maneuver from the left to the right lane at one of the data collection sites. At a few sites, the perspective of the camera did not pro- vide a sufficient view to document lane-change information. 3.2.2 Analysis Results of Speed Data Figure 11 shows the locations on the approach tangent and horizontal curve at which speed data were collected. The zero point of each measurement distribution represents the begin- ning of the curve (i.e., PC). In most cases, a maximum of 3% to 6% of the observations were obtained at a specific loca- tion along the study site. At a few sites (e.g., CA1, CA2, and WV5), the geometrics and roadside characteristics prohibited collecting speed data over the desired coverage area. Table 6 provides summary statistics of the speed data for passenger vehicles located 500 ft upstream of the curve, at the beginning of the curve (i.e., PC), and 500 ft downstream of the PC at each data collection site. The Table also provides the posted speed limit at each site and the advisory speed (if posted). The third column provides the average vehicle count (i.e., number of observations) at the three respective loca- tions included in the table. At some sites, passenger vehicle speeds decreased going from 500 ft upstream of the curve to the beginning of the curve, while at other sites speeds increased. At most sites passenger vehicle speeds decreased going from the beginning of the curve to 500 ft downstream of the curve. Table 7 provides the corresponding summary statistics for trucks. At most sites truck speeds decreased going from 500 ft upstream of the curve to the beginning of the curve. Similarly, at most sites truck speeds decreased going from the beginning of the curve to 500 ft downstream of the beginning of the curve. Table 8 provides detailed speed information collected at Maryland site MD1 for both passenger vehicles and trucks at 100 ft intervals. These speed data were entered into the simulation models (i.e., CarSim and TruckSim) to determine friction supply curves (and the corresponding lateral friction and rollover margins) for passenger vehicles and trucks at each of the data collection sites based upon actual operating speeds measured at the sites. 3.2.3 Analysis Results of Lane-Change Maneuver Data The primary measures of interest from the lane-change analysis consisted of the frequency and duration of the maneu- vers. Table 9 provides summary statistics on the frequency and percentage of lane-change maneuvers observed at each site by vehicle type and grade direction (i.e., downgrade and upgrade). The Table provides data on total vehicles by vehicle type and whether the lane change consisted of a maneuver from the right lane to the left lane (identified as left in the table) or from the left lane to the right lane (i.e., right in the table). As long as the lane-change maneuver occurred within the field of view of the video camera, the lane-change maneu- ver was documented. Thus, in some cases the lane change may have occurred on the approach tangent, on the approach tangent and into the curve, or entirely within the curve. At Figure 10. Video of tractor semi-trailer maneuvering from left lane to right lane.

24 0 4 8 C A 1 0 4 8 C A 2 0 4 8 C A 3 0 4 8 M D 1 0 4 8 M D 2 0 4 8 M D 3 0 4 8 PA 1 0 4 8 PA 2 0 4 8 W A 1 -2440 -2040 -1640 -1240 -840 -440 -40 360 760 1160 1560 1960 2360 2760 0 4 8 Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t W A 2 Distance from Point of Curvature (ft) Figure 11. Speed measurement coverage at data collection sites.

25 0 4 8 W A 3 0 4 8 W A 4 0 4 8 W A 5 0 4 8 W A 6 0 4 8 W A 7 0 4 8 W V 1 0 4 8 W V 2 0 4 8 W V 3 0 4 8 Pe rc en t W V 4 -2440 -2040 -1640 -1240 -840 -440 -40 360 760 1160 1560 1960 2360 2760 0 4 8 Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t Pe rc en t W V 5 Distance from Point of Curvature (ft) Figure 11. (Continued).

Site Posted speed limit/ advisory speed (mph) Avg veh count 500 ft upstream of curve Beginning of curve (i.e., PC) 500 ft downstream of PC Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph CA1 65 27 66.0 73.4 7.0 30 NA NA NA NA NA NA NA NA CA2 40/45 53 51.3 55.9 5.0 90 53.0 57.1 4.8 93 NA NA NA NA CA3 50/40 55 53.9 58.0 4.1 33 NA NA NA NA 49.1 52.0 3.4 0 MD1 65 70 65.4 71.0 5.2 17 65.5 70.8 4.7 20 65.0 69.5 4.7 14 MD2 65 65 63.9 70.0 7.5 3 63.2 69.9 7.8 12 61.2 68.6 9.9 9 MD3 65 76 68.4 73.5 4.9 32 68.0 73.1 5.2 26 67.1 72.5 5.0 20 PA1 40/25 61 NA NA NA NA 36.6 40.4 4.0 2 NA NA NA NA PA2 65 66 63.5 67.0 4.5 9 66.8 72.0 4.6 20 67.0 72.5 4.5 19 WA1 70/50 79 64.4 70.8 5.1 0 62.4 68.8 5.3 0 57.3 62.9 6.0 0 WA2 70 42 69.4 72.6 4.0 2 69.2 74.0 4.4 2 68.6 72.4 4.6 2 WA3 70 73 NA NA NA NA 70.2 74.6 4.4 11 69.5 74.4 4.5 8 WA4 70 66 67.8 71.5 3.8 2 67.8 71.9 4.0 0 66.3 70.6 4.3 0 WA5 60 55 57.9 62.5 4.3 24 58.4 64.0 5.8 33 56.8 64.2 6.5 24 WA6 70 54 68.2 73.0 3.6 0 68.9 72.9 3.6 3 68.5 72.6 3.8 5 WA7 60/40 114 NA NA NA NA 51.0 56.0 5.8 82 48.2 52.8 4.7 72 WV1 70/50 53 64.3 70.4 5.2 0 66.9 72.0 5.6 6 67.1 71.2 5.2 6 WV2 70/50 45 67.8 74.1 6.8 13 68.4 73.5 5.9 13 68.9 76.4 4.8 15 WV3 70/50 86 67.7 73.2 5.4 7 67.2 73.0 5.0 7 65.3 71.3 5.2 1 WV4 60/50 94 64.2 69.7 5.6 35 62.6 68.0 5.3 26 NA NA NA NA WV5 60 69 NA NA NA NA NA NA NA NA 68.5 73.5 4.4 72 Table 6. Summary of speed data near curves (passenger vehicles). Site Posted speed limit/ advisory speed (mph) Avg veh count 500 ft upstream of curve Beginning of curve (i.e., PC) 500 ft downstream of PC Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph Mean speed (mph) 85th %tile speed (mph) Std dev (mph) Percent exceed posted speed limit >5 mph CA1 551 47 53.3 56.9 3.8 3 51.9 55.5 3.3 0 NA NA NA NA CA2 351/45 32 42.5 48.1 5.1 56 41.8 48.1 5.3 51 NA NA NA NA CA3 50/40 23 39.8 48.0 6.6 4 NA NA NA NA 38.0 45.0 5.1 0 MD1 65 48 63.2 66.3 3.7 0 63.0 66.4 3.1 0 61.7 65.8 3.7 0 MD2 65 65 43.0 60.0 13.4 1 41.5 59.0 13.4 1 39.1 54.6 13.2 0 MD3 65 63 64.5 69.1 5.5 7 64.1 68.0 5.4 11 64.3 69.9 5.7 14 PA1 40/25 46 33.7 38.6 5.5 2 26.2 30.4 3.8 0 NA NA NA NA PA2 65 54 64.5 67.2 3.7 4 65.0 68.9 4.6 9 65.3 69.1 4.3 11 WA1 601/50 54 56.8 60.4 5.1 0 54.5 58.7 3.7 0 50.0 54.1 4.0 0 WA2 601 44 60.8 65.1 5.0 14 60.3 65.1 4.4 7 60.1 64.7 4.4 7 WA3 601 38 NA NA NA NA 58.7 65.0 6.1 17 58.3 63.8 6.0 11 WA4 601 34 58.8 62.9 4.4 0 58.5 62.0 4.0 6 57.8 62.0 4.2 3 WA5 60 52 45.9 56.0 11.4 2 44.4 56.0 11.9 3 41.1 54.0 12.0 0 WA6 601 40 61.2 66.2 3.5 19 61.8 65.4 3.0 16 61.1 63.6 2.5 4 WA7 60/40 15 NA NA NA NA 36.7 48.2 11.2 47 36.1 47.2 10.6 36 WV1 70/50 48 61.5 67.3 5.3 0 62.5 67.0 5.0 0 62.0 66.3 5.0 0 WV2 501/50 37 55.8 62.7 7.4 40 54.6 58.6 6.8 38 55.5 66.0 8.0 37 WV3 70/50 49 63.9 69.1 4.7 0 63.4 69.0 4.8 0 62.5 66.5 4.9 0 WV4 60/50 71 58.6 63.3 4.1 7 57.3 61.9 3.8 0 NA NA NA NA WV5 60 49 NA NA NA NA NA NA NA NA 65.8 69.9 3.9 53 1 Dual speed limits for passenger vehicles and trucks. Table 7. Summary of speed data near curves (trucks).

27 Dist (ft) Passenger vehicles Trucks # of Obs Avg speed (mph) Std dev (mph) Speed percentiles (mph) # of Obs Avg speed (mph) Std dev (mph) Speed percentiles (mph) Min 25th Median 85th Max Min 25th Median 85th Max −500 69 65.4 5.2 51.4 61.6 65.1 71.0 76.0 49 63.2 3.7 51.0 60.6 63.9 66.3 69.0 −400 70 65.7 4.9 53.2 61.9 65.5 71.1 75.5 48 63.2 3.6 51.0 61.2 64.0 66.4 69.1 −300 72 65.9 4.7 54.7 62.0 65.5 72.0 75.4 48 63.2 3.5 51.0 60.9 64.0 66.4 69.1 −200 73 65.7 4.8 55.2 62.1 65.1 72.0 75.5 47 63.3 3.1 55.6 60.9 64.1 66.4 69.1 −100 72 65.7 4.8 54.7 62.1 64.9 72.0 75.6 47 63.2 3.1 56.2 60.9 63.9 66.5 69.2 0 71 65.5 4.7 54.0 62.2 65.0 70.8 75.5 47 63.0 3.1 56.4 60.8 63.8 66.4 69.2 100 71 65.4 4.7 53.3 62.0 65.0 70.8 75.3 46 62.8 3.2 55.8 60.7 63.1 66.7 69.2 200 72 65.4 4.8 52.8 62.0 65.2 70.9 75.1 46 62.5 3.3 55.3 59.9 62.9 66.4 69.2 300 71 65.4 4.7 52.6 61.7 65.5 70.7 75.0 45 62.3 3.4 54.8 59.9 62.7 66.1 69.2 400 72 65.3 4.7 52.9 61.9 65.4 70.4 74.8 47 61.9 3.7 52.7 59.4 62.4 66.0 69.0 500 71 65.0 4.7 53.5 61.5 65.2 69.5 74.7 48 61.7 3.7 52.9 59.5 62.2 65.8 68.9 600 69 64.9 4.7 54.2 61.5 65.2 69.9 74.6 49 61.5 3.7 53.0 59.4 61.8 65.7 68.8 700 68 64.9 4.8 54.5 61.7 65.2 70.0 75.2 48 61.2 3.7 53.0 59.0 61.6 65.8 68.5 800 66 64.9 4.8 54.0 61.7 65.1 70.5 75.0 47 60.9 3.7 53.0 58.5 61.1 65.6 68.1 900 65 64.9 4.7 53.8 61.9 64.8 70.0 75.2 46 60.6 3.7 52.9 58.3 60.7 65.4 67.7 1,000 61 64.8 4.5 53.8 61.8 64.8 69.7 73.3 46 60.5 3.7 52.8 58.0 60.7 65.2 67.4 1,100 55 64.5 4.4 53.8 61.4 64.9 68.5 74.6 42 60.7 3.9 52.4 57.9 60.9 65.0 67.0 1,200 49 64.3 4.6 53.7 61.3 64.4 69.6 75.2 39 60.4 4.1 52.1 57.7 61.0 64.9 66.7 1,300 35 65.3 4.6 53.9 61.5 65.2 70.6 76.0 38 60.3 4.2 51.7 57.6 61.0 64.9 66.4 1,400 19 64.4 4.1 54.0 61.2 65.1 69.6 72.7 31 59.6 4.2 51.1 56.7 60.1 64.5 65.9 NOTE: Posted speed limit at site is 65 mph. Table 8. Speed distribution data for passenger vehicles and trucks at Maryland site MD1. Site Passenger vehicles Trucks All vehicles combined Total vehicles Lane-change count (%) Total vehicles Lane-change count (%) Total vehicles Lane-change count (%) Left Right Total Left Right Total Left Right Total Downgrade CA1 2,432 25 (0.01) 20 (0.01) 45 (0.02) 1,271 5 (0.00) 19 (0.01) 24 (0.02) 3,703 30 (0.01) 39 (0.01) 69 (0.02) CA2 2,344 57 (0.02) 39 (0.02) 96 (0.04) 141 3 (0.02) 4 (0.03) 7 (0.05) 2,485 60 (0.02) 43 (0.02) 103 (0.04) CA3 2,804 30 (0.01) 37 (0.01) 67 (0.02) 148 1 (0.01) 0 (0.00) 1 (0.01) 2,952 31 (0.01) 37 (0.01) 68 (0.02) MD1 321 27 (0.08) 36 (0.11) 63 (0.20) 88 12 (0.14) 4 (0.05) 16 (0.18) 409 39 (0.10) 40 (0.10) 79 (0.19) MD3 924 37 (0.04) 12 (0.01) 49 (0.05) 208 4 (0.02) 3 (0.01) 7 (0.03) 1,132 41 (0.04) 15 (0.01) 56 (0.05) PA2 944 20 (0.02) 38 (0.04) 58 (0.06) 439 10 (0.02) 39 (0.09) 49 (0.11) 1,383 30 (0.02) 77 (0.06) 107 (0.08) WA1 669 12 (0.02) 10 (0.01) 22 (0.03) 262 1 (0.00) 8 (0.03) 9 (0.03) 931 13 (0.01) 18 (0.02) 31 (0.03) WA2 426 25 (0.06) 20 (0.05) 45 (0.11) 138 6 (0.04) 1 (0.01) 7 (0.05) 564 31 (0.05) 21 (0.04) 52 (0.09) WA3 610 8 (0.01) 34 (0.06) 42 (0.07) 121 4 (0.03) 4 (0.03) 8 (0.07) 731 12 (0.02) 38 (0.05) 50 (0.07) WA4 488 19 (0.04) 43 (0.09) 62 (0.13) 119 4 (0.03) 5 (0.04) 9 (0.08) 607 23 (0.04) 48 (0.08) 71 (0.12) WA6 475 5 (0.01) 13 (0.03) 18 (0.04) 168 0 (0.00) 7 (0.04) 7 (0.04) 643 5 (0.01) 20 (0.03) 25 (0.04) WV1 953 8 (0.01) 103 (0.11) 111 (0.12) 278 2 (0.01) 122 (0.44) 124 (0.45) 1,231 10 (0.01) 225 (0.18) 235 (0.19) WV3 957 53 (0.06) 38 (0.04) 91 (0.10) 102 10 (0.10) 11 (0.11) 21 (0.21) 1,059 63 (0.06) 49 (0.05) 112 (0.11) WV4 625 9 (0.01) 17 (0.03) 26 (0.04) 380 7 (0.02) 5 (0.01) 12 (0.03) 1,005 16 (0.02) 22 (0.02) 38 (0.04) WV5 1,687 6 (0.00) 13 (0.01) 19 (0.01) 328 6 (0.02) 7 (0.02) 13 (0.04) 2,015 12 (0.01) 20 (0.01) 32 (0.02) Upgrade MD2 1,204 27 (0.02) 36 (0.03) 63 (0.05) 257 13 (0.05) 3 (0.01) 16 (0.06) 1,461 40 (0.03) 39 (0.03) 79 (0.05) WA5 188 4 (0.02) 3 (0.02) 7 (0.04) 86 1 (0.01) 3 (0.03) 4 (0.05) 274 5 (0.02) 6 (0.02) 11 (0.04) Table 9. Summary of lane-change maneuvers by vehicle type and grade direction.

28 most sites, less than 10% of the vehicles changed lanes near or on the curve. At two of the sites (MD1 and WV1), nearly 20% of the vehicles changed lanes. This was most likely due to entrance/exit ramps located in the vicinity of these curves. Table 10 presents summary statistics of lane-change duration data for passenger vehicles on both downgrades and upgrades. Lane-change duration was defined to be the amount of time from when the right tires of a vehicle crossed the lane lines to when the left tires crossed the lane lines for a right maneuver and when the left tires of a vehicle crossed the lane lines to when the right tires crossed the lane lines for a left maneuver. Thus, the actual lane-change duration from when the driver initiated the maneuver when positioned near the center of one travel lane until the time the driver completed the maneuver to the center of the other travel lane was longer than what is reported here, but for consistency and an objective measure for determining the start and end times of the maneu- vers, the definition above was used. From Table 10 it is assessed that on downgrades, passenger vehicles had similar mean durations for maneuvers to the left (2.85 s) and to the right (2.94 s). On upgrades, passenger vehicles took slightly longer to maneuver to the right (3.25 s) compared to the left (2.95 s). Table 11 presents summary statistics of lane-change dura- tion data for trucks on both downgrades and upgrades. On downgrades, trucks had similar mean durations for maneu- vers to the left (4.00 s) and to the right (4.09 s). On upgrades, trucks took longer to maneuver to the right (5.81 s) than to the left (4.47 s). Tables 12 and 13 provide lane-change summary statistics for passenger vehicles and trucks by curve direction to assess whether lane-change duration is affected by whether the maneuver is made with the curve (i.e., left maneuver on a curve to the left or a right maneuver on a curve to the right) or against the curve (i.e., left maneuver on a curve to the right or a right maneuver on a curve to the left). A split-plot model was used to estimate the statistical dif- ferences between mean lane-change durations for: • Two grade directions (i.e., upgrade and downgrade); • Two curve directions (i.e., left and right); • Two vehicle types (i.e., passenger vehicles and trucks); and • Two lane-change directions (i.e., left and right). The 17 field sites were included in the model as random effects, assuming that the sites were chosen from a larger popu- lation of potential sites. This allows for estimation of the main effects on lane-change duration accounting for the added variability associated with using data from multiple sites. Site Left maneuver (s) Right maneuver (s) Veh count Mean Std dev Min Max Veh count Mean Std dev Min Max Downgrade CA1 25 2.84 0.80 1.00 4.00 20 3.35 1.27 2.00 8.00 CA2 57 2.63 0.98 1.00 7.00 39 2.28 0.76 1.00 4.00 CA3 30 3.13 0.43 2.00 4.00 37 3.35 0.59 2.00 4.00 MD1 27 3.03 0.55 2.15 4.53 36 3.24 0.72 1.94 5.28 MD3 37 2.57 0.60 1.65 4.16 12 2.55 0.59 1.56 3.53 PA2 20 2.70 0.68 1.69 4.84 38 2.44 0.58 1.53 3.87 WA1 12 2.67 0.65 2.00 4.00 10 3.00 1.33 1.00 5.00 WA2 25 2.92 0.81 2.00 5.00 20 3.10 1.25 2.00 7.00 WA3 8 2.88 0.83 2.00 4.00 34 3.00 0.85 2.00 5.00 WA4 19 3.21 0.71 2.00 5.00 43 2.72 0.85 1.00 5.00 WA6 5 3.00 1.22 2.00 5.00 13 3.69 1.44 2.00 6.00 WV1 8 3.12 0.65 2.25 3.97 103 2.92 0.60 1.69 4.57 WV3 53 2.95 0.80 1.72 5.38 38 3.12 0.75 1.62 4.66 WV4 9 2.88 0.52 2.34 3.81 17 3.37 0.60 2.69 4.65 WV5 6 2.26 0.29 1.91 2.63 13 2.47 0.48 1.90 3.31 Downgrade average 341 2.85 0.76 1.00 7.00 473 2.94 0.86 1.00 8.00 Upgrade MD2 27 3.02 0.53 2.15 4.53 36 3.24 0.72 1.94 5.28 WA5 4 2.50 1.29 1.00 4.00 3 3.33 1.53 2.00 5.00 Upgrade average 31 2.95 0.67 1.00 4.53 39 3.25 0.78 1.94 5.28 Table 10. Lane-change duration statistics for passenger vehicles by grade direction.

29 Site Left maneuver (s) Right maneuver (s) Veh count Mean Std dev Min Max Veh count Mean Std dev Min Max Downgrade CA1 5 5.60 0.89 4.00 6.00 19 6.37 1.46 3.00 8.00 CA2 3 4.00 1.00 3.00 5.00 4 3.00 0.82 2.00 4.00 CA3 1 3.00 3.00 3.00 0 MD1 12 4.31 0.45 3.75 5.16 4 5.21 1.25 3.35 5.97 MD3 4 3.65 0.45 2.99 4.03 3 3.10 0.46 2.79 3.63 PA2 10 3.13 0.62 2.50 4.78 39 3.35 0.66 2.22 5.18 WA1 1 6.00 6.00 6.00 8 6.63 1.06 5.00 8.00 WA2 6 5.33 1.63 3.00 7.00 1 7.00 7.00 7.00 WA3 4 4.75 1.50 4.00 7.00 4 5.50 1.29 4.00 7.00 WA4 4 4.50 1.00 4.00 6.00 5 4.60 0.89 3.00 5.00 WA6 0 6 7.17 1.60 5.00 9.00 WV1 2 3.94 1.73 2.72 5.16 122 3.72 0.80 2.28 5.50 WV3 10 3.28 0.74 2.13 4.15 11 3.50 1.00 2.03 4.90 WV4 7 3.84 0.77 2.88 4.83 5 3.57 1.01 2.24 4.72 WV5 6 2.82 0.62 2.16 3.84 7 3.18 0.57 2.31 4.18 Downgrade average 75 4.00 1.18 2.13 7.00 238 4.09 1.41 2.00 9.00 Upgrade MD2 13 4.43 0.63 3.75 5.97 3 4.96 1.40 3.35 5.84 WA5 1 5.00 5.00 5.00 3 6.67 2.08 5.00 9.00 Upgrade average 14 4.47 0.63 3.75 5.97 6 5.81 1.84 3.35 9.00 Table 11. Lane-change duration statistics for trucks by grade direction. Site Left maneuver (s) Right maneuver (s) Veh count Mean Std dev Min Max Veh count Mean Std dev Min Max Curve left CA1 25 2.84 0.80 1.00 4.00 20 3.35 1.27 2.00 8.00 CA3 30 3.13 0.43 2.00 4.00 37 3.35 0.59 2.00 4.00 MD1 27 3.03 0.55 2.15 4.53 36 3.24 0.72 1.94 5.28 PA2 20 2.70 0.68 1.69 4.84 38 2.44 0.58 1.53 3.87 WA2 25 2.92 0.81 2.00 5.00 20 3.10 1.25 2.00 7.00 WA5 4 2.50 1.29 1.00 4.00 3 3.33 1.53 2.00 5.00 WV1 8 3.12 0.65 2.25 3.97 103 2.92 0.60 1.69 4.57 WV3 53 2.95 0.80 1.72 5.38 38 3.12 0.75 1.62 4.66 WV5 6 2.26 0.29 1.91 2.63 13 2.47 0.48 1.90 3.31 Left curve average 198 2.92 0.72 1.00 5.38 308 3.00 0.80 1.53 8.00 Curve right CA2 57 2.63 0.98 1.00 7.00 39 2.28 0.76 1.00 4.00 MD2 27 3.02 0.53 2.15 4.53 36 3.24 0.72 1.94 5.28 MD3 37 2.57 0.60 1.65 4.16 12 2.55 0.59 1.56 3.53 WA1 12 2.67 0.65 2.00 4.00 10 3.00 1.33 1.00 5.00 WA3 8 2.88 0.83 2.00 4.00 34 3.00 0.85 2.00 5.00 WA4 19 3.21 0.71 2.00 5.00 43 2.72 0.85 1.00 5.00 WA6 5 3.00 1.22 2.00 5.00 13 3.69 1.44 2.00 6.00 WV4 9 2.88 0.52 2.34 3.81 17 3.37 0.60 2.69 4.65 Right curve average 174 2.78 0.79 1.00 7.00 204 2.90 0.94 1.00 6.00 Table 12. Lane-change duration statistics for passenger vehicles by curve direction.

30 The degrees of freedom were calculated using the Welch- Satterthwaite equation, and variance components were used for the variance-structure of the split-plot model. Main effect results are shown in Table 14, and statistically signifi- cant interaction effects are shown in Table 15. There is no evidence of a statistically significant difference in vehicle lane-change duration means at upgrade sites com- pared to downgrade sites (p-value = 0.2385) or at sites with a curve to the left compared to a curve to the right (p-value = 0.7898). There is a statistically significant difference in lane- change duration means between passenger vehicles and trucks (p-value < 0.0001) where passenger vehicles execute the lane- change maneuver about 1.4 s quicker than trucks. There is also a statistically significant difference in mean lane-change duration for vehicles maneuvering into the left lane compared to vehicles maneuvering into the right lane (p-value = 0.0066), but for practical purposes, this difference in lane-change dura- tion (i.e., 0.27 s) is minimal or insignificant. Interactions between main effects were also important rela- tionships to examine, because one main effect can vary greatly Site Left maneuver (s) Right maneuver (s) Veh count Mean Std dev Min Max Veh count Mean Std dev Min Max Curve left CA1 5 5.60 0.89 4.00 6.00 19 6.37 1.46 3.00 8.00 CA3 1 3.00 3.00 3.00 0 MD1 12 4.31 0.45 3.75 5.16 4 5.21 1.25 3.35 5.97 PA2 10 3.13 0.62 2.50 4.78 39 3.35 0.66 2.22 5.18 WA2 6 5.33 1.63 3.00 7.00 1 7.00 7.00 7.00 WA5 1 5.00 5.00 5.00 3 6.67 2.08 5.00 9.00 WV1 2 3.94 1.73 2.72 5.16 122 3.72 0.80 2.28 5.50 WV3 10 3.28 0.74 2.13 4.15 11 3.50 1.00 2.03 4.90 WV5 6 2.82 0.62 2.16 3.84 7 3.18 0.57 2.31 4.18 Left curve average 53 3.94 1.24 2.13 7.00 206 3.95 1.27 2.03 9.00 Curve right CA2 3 4.00 1.00 3.00 5.00 4 3.00 0.82 2.00 4.00 MD2 13 4.43 0.63 3.75 5.97 3 4.96 1.40 3.35 5.84 MD3 4 3.65 0.45 2.99 4.03 3 3.10 0.46 2.79 3.63 WA1 1 6.00 6.00 6.00 8 6.63 1.06 5.00 8.00 WA3 4 4.75 1.50 4.00 7.00 4 5.50 1.29 4.00 7.00 WA4 4 4.50 1.00 4.00 6.00 5 4.60 0.89 3.00 5.00 WA6 0 6 7.17 1.60 5.00 9.00 WV4 7 3.84 0.77 2.88 4.83 5 3.57 1.01 2.24 4.72 Right curve average 36 4.28 0.90 2.88 7.00 38 5.13 1.85 2.00 9.00 Table 13. Lane-change duration statistics for trucks by curve direction. Main effect Group Vehicle count Mean (s) Std dev (s) Difference in means (s) p-valuea Grade Upgrade 90 4.14 1.07 0.67 0.2385 Downgrade 1,127 3.47 1.08 Curve Left 765 3.45 1.04 −0.32 0.7898 Right 452 3.77 1.20 Vehicle type Passenger vehicles 884 2.90 0.82 −1.43 <0.0001 Trucks 333 4.33 1.34 Lane-change maneuver Left 461 3.48 0.82 −0.27 0.0066 Right 756 3.75 1.28 a p-values below 0.05 indicate statistical significance at the 95% confidence level. Table 14. Analysis results of lane-change durations (main effects).

31 at different levels of another main effect. All inter action effects were tested in this model, but only two were found to be sta- tistically significant. For the interaction between vehicle type and grade type, lane-change durations for trucks are much higher if they occur on an upgrade compared to a downgrade (difference in means = 1.09 s) compared to lane-change dura- tions for passenger vehicles along an upgrade compared to a downgrade (difference in means = 0.23 s). For the interaction between curve direction and lane-change direction, there is also some evidence that curve direction has less of an effect on a vehicle making a left maneuver (difference in means between left curve sites and right curve sites = 0.10 s) than on a vehicle making a right maneuver (difference in means between left curve and right curve sites = 0.96 s; p-value = 0.0551). 3.3 Instrumented Vehicle Studies At five data collection sites (see Table 5 for specific sites), the research team collected a range of data using an instrumented vehicle. The purposes were to: 1. Measure the road geometry (i.e., grade, curvature, and cross slope) of each site to confirm whether the vehicle-based road measurements were in agreement with information from roadway inventory files, plan and profile sheets, and field measurements; 2. Obtain in-vehicle dynamics measurements for compari- son with simulation outputs to check the fidelity of the vehicle simulation software; and 3. Measure the continuous speed profiles of vehicles travers- ing the entire lengths of the data collection site (i.e., along the entire grade and curve) since laser gun measurements were collected primarily on the tangent approaching the curve and through the curve, and not along the entire length of the downgrade or upgrade. 3.3.1 Data Collection Methodology Roadway geometry, cross-slope, and vehicle dynamic data were collected at five sites from in-vehicle sensors while the test vehicle followed free-flow vehicles through the sites. At each site data were collected while following behind five sepa- rate passenger vehicles and two tractor semi-trailers. The instrumented vehicle was a 2010 Dodge Durango. This vehicle was chosen because of its capacity to hold the data collection equipment and because the vehicle’s inertial and kinematic parameters align well with those of a stan- dard full-size SUV as defined within CarSim. The vehicle was instrumented with a defense-grade global positioning sys- tem (GPS) coupled to a ring-laser-gyro inertial measurement unit (IMU) that gives accurate absolute measures of position and orientation. GPS/IMU data were collected at 100 Hz. In addition, a roof-mounted light detection and ranging (LIDAR) sensor was mounted to a gantry behind and above the vehicle. This road-scanning system was installed to look down perpendicular to the road to give 180° cross-section measurements of the road surface at 0.5° intervals, out to a distance of 260 ft from the sensor, for a total of 361 points per sweep. Each LIDAR sweep obtained 361 data points at 37.5 Hz while capturing the intensity of the LIDAR return. A camera was mounted to the dashboard of the vehicle and manually aligned so that the vanishing point of straight-line driving corresponds roughly to the center of the image. And finally, a steering angle sensor was installed to capture the driver’s steering inputs directly. All sensor inputs were col- lected using Player/Stage software, and each measurement was time-stamped with the computer’s local clock. At the beginning of each day of testing, the vehicle was calibrated. A diagram of the data collection system is shown in Figure 12, and an example screenshot from the forward-facing camera on the dashboard is shown in Figure 13. Interaction effect Group Vehicle count Mean (s) Std dev (s) Difference in means (s) p-valuea Grade direction and vehicle type Upgrade, passenger vehicles 70 3.13 0.63 0.23 0.0039 Downgrade, passenger vehicles 814 2.90 0.81 Upgrade, trucks 20 5.14 1.38 1.09 Downgrade, trucks 313 4.05 1.30 Curve direction and lane-change direction Left curve, left maneuver 251 3.43 1.01 −0.10 0.0551 Right curve, left maneuver 210 3.53 0.85 Left curve, right maneuver 514 4.98 1.06 0.96 Right curve, right maneuver 242 4.02 1.47 a p-values below 0.05 indicate statistical significance at the 95% confidence level. Table 15. Analysis results of lane-change durations (interaction effects).

32 The instrumented vehicle was driven behind vehicles in the traffic stream chosen randomly but selected such that the vehicles were not following other vehicles that would influ- ence their speed. The instrumented vehicle was maintained at a constant following distance—approximately 300 ft behind the lead vehicle as shown in Figure 13. Selected vehicles were followed beginning at the top/bottom of the grade and fol- lowed down/up the entire grade and through the curve. The data collection system provided a range of data, including the following: • Vehicle data – Velocities on each of the three axes – Acceleration/deceleration on each of the three axes – Steering angle – Roll, pitch, and yaw angles and rates about each axis – Position of the vehicle in latitude, longitude, and elevation • Roadway data – Vertical alignment – Horizontal alignment – Normal cross slope – Transition from normal cross slope to full superelevation – Full superelevation in curve 3.3.2 Analysis of Results Roadway geometry data were obtained from the LIDAR mea- surements. Standard coordinate transformations were used to convert from LIDAR coordinates, to vehicle coordi nates, and finally to globally referenced coordinates (see Vemulapalli and Brennan [2009] for details). The resulting point-cloud data were filtered to develop a smoothed road profile that provided grade, horizontal alignment, and cross-slope information for each site (Varunjikar, 2011). An example illustration of the resulting road profile after processing is shown in Figure 14. One of the first confirmations conducted on the mea- sured data was to verify that the measured grades matched Figure 12. Instrumented vehicle data collection system. Figure 13. Screenshot from instrumented vehicle during data collection. Figure 14. Three-dimensional point cloud obtained by instrumented vehicle data (site MD1).

33 the grades as reported on profile sheets for the sites. As an example, the measured grades were inferred from the height (z) versus horizontal alignment measurements as shown in Figure 15 for the WV2 site. In Figure 15 (and all subsequent figures), the zero point on the horizontal align- ment depicts the beginning of the curve (i.e., PC). Positive values for the horizontal alignment represent the relative position along the length of the curve, and negative values represent the relative position on the approach tangent to the curve. In Figure 15 the inferred grade is -5.6% which is consistent with the grade obtained from the profile sheets for this same site. A similar level of consistency between measured grades and grades obtained from profile sheets was found across all five sites in the instrumented vehicle study (Table 16). The second level of consistency checks focused on hori- zontal alignment. The measured horizontal alignment of the WV2 site is shown Figure 16. The Figure illustrates a curve to the left. Through visual inspection, comparisons were made of the collected horizontal geometry, as shown in Figure 16, and CAD drawings created by the research team of the road plans. Additionally, the collected horizontal vehicle trajec- tory was compared to Google Earth satellite images to further confirm geometric consistency. These comparisons indicated Figure 15. Measured elevations and horizontal distance for all traversals (site WV2). Site Percent grade Measured using instrumented vehicle Obtained from profile sheets MD1a −4.07 ± 0.27 −4.1 MD2 +6.17 ± 0.33 6.0 MD3 −5.61 ± 0.25 −5.7 PA1a −5.19 ± 0.16 −5.0 WV2 −5.62 ± 0.22 −5.7 aSlope for approach is different than the curve. The values shown here are for the approach geometry, not the curve itself. Table 16. Comparison of grades from instrumented vehicle data and profile sheets.

34 a high level of agreement between the instrumented vehicle data and the actual roadway plans. After it was confirmed that the in-vehicle geometric mea- surements agreed well with horizontal and vertical alignment information obtained from roadway plans and profiles, the horizontal and vertical alignment data were imported into CarSim to simulate the vehicle’s dynamics to compare simula- tion results to instrumented vehicle measurements. An exam- ple of this comparison is shown in Figure 17 for the WV2 site. Note, the horizontal and vertical alignment data and cross- section data (i.e., cross-slope and superelevation data) used to represent the site geometry in CarSim were based upon -1000 -500 0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000 X position (ft) Y po sit io n (ft) Start of Curve End of Curve Figure 16. Horizontal alignment from instrumented vehicle for all traversals (site WV2). -500 0 500 1000 1500 2000 2500 3000 3500 -10 -5 0 5 10 Horizontal Distance (ft) Ve hi cl e R ol l A ng le (d eg ) -500 0 500 1000 1500 2000 2500 3000 3500 -6 -4 -2 0 Horizontal Distance (ft) Ve hi cl e Pi tc h An gl e (de g) Start of Curve End of Curve IMU CarSim Figure 17. Comparison of CarSim simulation results and instrumented vehicle measurements for all traversals (site WV2).

35 information obtained from plans and profiles (and/or a com- bination of roadway inventory files and field measurements). In Figure 17, from a visual perspective, the simulation outputs closely agree with the measured data, including many of the transient effects such as oscillations in the entry and exit of the curve. Confirmation that the simulation outputs closely agreed with the data from the instrumented vehicle was important in several respects. First, it confirmed the fidelity and/or accu- racy of the CarSim model for use in subsequent phases of this research. Second, it provided a reasonable level of confirma- tion that horizontal and vertical alignment and cross-slope/ superelevation data obtained from combinations of plans and profiles, roadway inventory files, and field measurements could be used to accurately model the geometrics of the 20 data col- lection sites within CarSim, without the need to use the instru- mented vehicle to collect this information. 3.3.3 Continuous Speed Profiles One of the purposes of the instrumented vehicle study was to measure the continuous speed profiles of vehicles travers- ing each field site for the entire length of the site (i.e., from the top of the grade through the curve or from the bottom of the grade through the curve), since roadside laser gun measure- ments of speed are limited in their coverage to shorter seg- ments of the sites. Figure 18 shows all the speed traces from the instrumented vehicle traversals versus the mean speeds measured using laser guns for the WV2 site. This particular site had full coverage of the curve from the roadside laser gun locations. The Figure illustrates some of the phenomenon of a typical passenger vehicle. For example, the instrumented vehicle study showed that most vehicles that were followed maintained relatively constant speed through the curve, punctuated by areas of short changes. This behavior was readily observed in most of the traversals. For some vehicles, however, there are very large speed changes within the curve. For example, Figure 18 shows a situation where one followed vehicle changed speed from approximately 80 mph before the curve, to 50 mph within the curve, and then back to 80 mph after the curve. Figure 19 shows the corresponding acceleration/ deceleration of the subject vehicles while traversing the data collection site, as measured from the instrumented vehicle. Shown in this figure are the individual data traces for each vehicle traversal, the mean acceleration at each point in the curve, and the upper and lower bounds created from two standard deviations from the mean at each location. Prior work by Bonneson (2000b) suggested that vehicles slow down slightly on the entrance to a curve, with very minor decel- eration rates of -3 ft/s2. This deceleration on the entrance to a curve was not conclusively or consistently seen in the speed data collected from the instrumented vehicle; indeed, sev- eral of the followed vehicles actually accelerated rather than decelerated upon entrance to the curve. However, the upper and lower bounds on the accelerations throughout the curve are approximately bounded by 3 ft/s2 deviations from zero acceleration (e.g., constant speed). -500 0 500 1000 1500 2000 2500 3000 3500 30 40 50 60 70 80 90 Horizontal Distance (ft) Ve hi cl e Sp ee d (m ph ) Start of Curve End of Curve In-vehicle speed measurement Vehicle speed study mean speed Figure 18. Comparison of speed profiles from instrumented vehicle and mean speeds from laser guns (site WV2).

36 Several general findings regarding the speed data collected from the instrumented vehicle are as follows: • Overall, the mean speed profiles measured by the instru- mented vehicle agreed with the speed data collected from the laser guns. • The variability in the vehicle acceleration within a curve was approximately between 3 and -3 ft/s2; this magnitude is consistent with the curve-entry deceleration reported by Bonneson (2000b). Hereafter, this deceleration level is denoted as “curve-entry deceleration,” even though the field data indicate that the deceleration may occur throughout the curve. 3.3.4 Summary of Instrumented Vehicle Study Consistent with the main goals of the instrumented vehicle study, several observations can be inferred from the analy- sis results presented above. First, the horizontal and verti- cal alignment and cross-slope/superelevation data obtained from combinations of plans and profiles, roadway inventory files, and field measurements agreed with the corresponding data measured from the instrumented vehicle. Because plans and profiles, roadway inventory files, and field measurements were available for all 20 data collection sites, and the instru- mented vehicle results were available at only 5 sites, horizontal and vertical alignment and cross-slope/superelevation data obtained from combinations of plans and profiles, roadway inventory files, and field measurements were used for all site- specific simulations. Second, the outputs from the vehicle dynamics simulations agreed closely with the instrumented vehicle data. This agree- ment gives confidence in the fidelity of the simulation results. Third, the speed profiles of the instrumented vehicle study were found to be in agreement with the speed data collected from the laser guns. In addition, the magnitude of the decel- erations observed from the instrumented vehicle speed data is consistent with the findings of NCHRP Report 439 (Bonneson, 2000b). Thus, for the simulations (see Section 4), some scenarios were performed assuming minor decelerations of -3 ft/s2 as curve-entry deceleration levels. 3.4 Friction Testing 3.4.1 Purpose The purpose of the friction testing was to establish fric- tion values for tires on both passenger vehicles and trucks suitable for modeling a vehicle’s expected behavior on steep grades and through sharp horizontal curves where both lat- eral and longitudinal forces must be generated. The friction data were used in conjunction with the simulation analy- ses (see Section 4) to determine the difference between the AASHTO design friction curves and the friction supply and demand on representative grades and horizontal curves. This -500 0 500 1000 1500 2000 2500 3000 3500 -6 -4 -2 0 2 4 Horizontal Distance (ft) Ve hi cl e Ac ce le ra tio n (ft/ s2 ) Start of Curve End of Curve Mean Mean+2 Mean-2 Figure 19. Longitudinal acceleration profiles from instrumented vehicle (site WV2).

37 section of the report (1) explains the data collection meth- odology used to collect friction data in the field, (2) presents general processing procedures to translate the raw field data into friction values for use in the simulation modeling, and (3) summarizes the results focusing on how the friction data were utilized in the simulation modeling. 3.4.2 Data Collection Methodology A dynamic friction (DF) tester was used to evaluate the skid resistance of the pavements at the field study sites. Testing was performed in accordance with American Society for Testing and Materials (ASTM) E1911-09a, Standard Test Method for Measuring Pavement Surface Frictional Properties Using the Dynamic Friction Tester. The DF tester measures the necessary torque to turn three small rubber pads in a circular path on the measured surface at different speeds. The apparatus consists of a horizontal spinning disk fitted with three spring-loaded rub- ber sliders that contact the paved surface as the disk rotational speed decreases due to the friction supplied between the sliders and the paved surface. A water supply unit delivers water to the paved surface being tested. The torque generated by the slider forces measured during the spin down is used to calculate the friction supply as a function of speed. Typical test speeds range from 55 to 3 mph. The DF tester is shown in Figure 20. The device was manually placed on the pavement surface at each testing site location. A laptop computer was used to control the test and record the data. When a test was initiated, the disk was accelerated to the standard spinning speed of Figure 20. DF tester. 55 mph. The spinning disk was then dropped to the ground, at which time automated data acquisition began. The test was complete when the disk stopped. A circular track (CT) meter was used with the DF tester to measure road surface texture characteristics. The CT meter measures surface texture on the same circular track as the DF tester. The CT meter calculates the mean profile depth (MPD) of the road surface and the International Friction Index (IFI). Raw data from the DF tester and CT meter were filtered to calculate the friction supply at the tire–pavement interface. The data were used to prepare friction supply curves for wet and dry pavements similar to those presented in the AASHTO Green Book as shown in Figure 21 (e.g., see the “New tires– wet concrete pavement” curve). The following protocol was used during field testing: 1. Each test section was divided into two segments: – The first segment consisted of 450 ft of the approach tangent upstream of the horizontal curve. – The second segment was the entire length of the hori- zontal curve. 2. The first segment was subdivided into three equal lengths (i.e., 150 ft sectors). Friction measurements were collected at the beginning and end points of these sectors using the DF tester and CT meter devices. This yielded four total friction supply measurement locations on the approach tangent to the horizontal curve. The intent of measuring friction at multiple locations on the approach tangent was to understand variability in pavement friction for areas

38 of normal travel and slight deceleration. All friction mea- surements on the approach tangent were taken in the left wheel path. 3. The second segment (i.e., horizontal curve) was similarly divided into three equal length sectors yielding four physi- cal measurement locations. The intent of measuring fric- tion supply at four locations within the horizontal curve was to provide information about the variability in friction supply within limits of the curve. 4. Each measurement location within a sector was defined as a 6 ft long straight line. The beginning, middle, and end points of the 6 ft line were separately measured using the DF tester and CT meter devices, producing three individ- ual measurement points for each location. 5. Within each horizontal curve segment, the measurement location was determined as follows: – On curves to the right, measurements were recorded in the left wheel path as this location will experience more polishing and therefore will supply less friction than the right wheel path. – On curves to the left, friction supply was measured in the right wheel path. A diagram of the testing points/locations on the approach tangent and horizontal curve is shown in Figure 22. Friction data were collected at eight field sites (see Table 5 for specific sites). The travel lanes at each of the sites consisted of asphalt pavement that appeared to be in good condition. The resulting DF tester and CT meter values are presented in Table 17. 3.4.3 Summary of Friction Testing Friction measurements were recorded at 21 locations approaching and within a curve at each of eight field sites. These field measurements were then processed to obtain tire force curves for representative passenger vehicle and truck tires on the roads where friction measurements were taken. Section 4.2 describes the general procedures for taking the field measurements and generating tire force curves. Figure 21. AASHTO side friction factors for low-speed design and friction supply curves (AASHTO, 2011).

39 1,2,3 450 ft Close-up of measurement locations 150 ft 150 ft 150 ft First Segment (approach) Second Segment (curve) Curve is divided into 3 equal lengths 6 ft 1 2 3 4,5,6 10,11,12 7,8,9 13,14,15 16,17,18 19,20,21 Figure 22. Friction measurement locations at a site. Site Dynamic friction tester—DFT20 (coefficient of friction) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 MD1 0.70 0.69 0.70 0.67 0.67 0.65 0.67 0.66 0.66 0.64 0.64 0.65 0.67 0.65 0.66 0.65 0.66 0.67 0.63 0.63 0.64 MD2 0.40 0.44 0.43 0.46 0.44 0.43 0.44 0.42 0.43 0.42 0.44 0.44 0.44 0.46 0.44 0.46 0.43 0.42 0.49 0.47 0.52 MD3 0.53 0.52 0.51 0.56 0.56 0.52 0.38 0.50 0.50 0.50 0.50 0.50 0.44 0.45 0.45 0.48 0.49 0.48 0.45 0.44 0.43 WV1 0.38 0.41 0.45 0.48 0.46 0.48 0.48 0.48 0.48 0.50 0.51 0.47 0.47 0.48 0.47 0.50 0.50 0.51 0.51 0.51 0.53 WV2 0.52 0.58 0.59 0.59 0.59 0.59 0.57 0.58 0.57 0.57 0.57 0.56 0.52 0.52 0.52 0.52 0.52 0.51 0.47 0.44 0.46 WV3 0.51 0.59 0.60 0.60 0.61 0.60 0.61 0.63 0.59 0.57 0.59 0.60 0.47 0.50 0.48 0.44 0.47 0.45 0.57 0.53 0.54 WV4 0.51 0.54 0.57 0.56 0.57 0.58 0.60 0.56 0.60 0.62 0.58 0.59 0.58 0.58 0.53 0.57 0.56 0.57 0.55 0.56 0.35 WV5 0.50 0.52 0.54 0.58 0.56 0.57 0.55 0.55 0.54 0.54 0.53 0.54 0.47 0.48 0.48 0.38 0.46 0.48 0.53 0.54 0.58 Site Circular track meter—Mean profile depth (in.) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 MD1 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 0.02 MD2 0.03 0.03 0.03 0.02 0.02 0.02 0.03 0.03 0.03 0.02 0.02 0.02 0.02 0.03 0.03 0.03 0.02 0.02 0.02 0.02 0.02 MD3 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.02 0.03 0.03 0.04 0.03 0.05 0.04 0.04 0.03 0.03 0.03 WV1 0.02 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.02 0.02 0.03 0.03 0.03 0.03 0.03 0.03 WV2 0.03 0.03 0.02 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.06 0.04 0.03 0.03 0.03 0.03 WV3 0.05 0.05 0.05 0.05 0.04 0.05 0.05 0.05 0.05 0.05 0.05 0.05 0.06 0.08 0.09 0.07 0.07 0.06 0.06 0.06 0.05 WV4 0.03 0.02 0.03 0.03 0.02 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.02 0.02 0.03 0.02 0.02 0.03 0.02 0.03 0.02 WV5 0.03 0.04 0.03 0.03 0.04 0.04 0.03 0.03 0.03 0.03 0.03 0.03 0.04 0.04 0.04 0.04 0.05 0.05 0.04 0.05 0.05 Table 17. DF tester and CT meter values for field sites.