Design Guidance for Freeway Mainline Ramp Terminals (2012)

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.

103 This portion of the research focused on investigating driver behaviors while performing merge and diverge maneuvers onto and off of freeways. There are a number of ways to explore behavioral tendencies and patterns, both directly and indirectly. This study focused on several indirect behavioral measures, specifically looking at speed, acceleration/deceleration, use of throttle and brake, glance activity, and presence of a leading vehicle during the merge/diverge maneuver. The behavioral study was designed to collect detailed information on a limited number of drivers, in contrast to more generalized information gathered on a large number of vehicles in the observational study. By observing drivers in close detail, it is possible to identify behavioral patterns and influences that determine how drivers operate their vehicles on freeway ramps, thus providing further information to determine whether the assumptions and data used to sup- port existing design guidelines are appropriate for current conditions. This section describes the processes used to collect and reduce the data for the behavioral study, and it describes the various types of data collected and the analysis results. 6.1 Data Collection The behavioral study took place in the Dallas-Ft. Worth Metroplex area of Texas, using a data collection protocol approved by institutional review boards at Texas A&M Uni- versity and MRIGlobal. Flyers were distributed to various agencies/organizations to recruit 12 subjects to participate in the study. Subjects were paid for their participation. The Texas A&M Transportation Institute (TTI) has devel- oped an instrumented vehicle to facilitate robust data collec- tion in both test-track and public road environments. A 2006 Toyota Highlander is equipped with multiple integrated sys- tems to record various data relating driver behavior, traffic conditions, and vehicle performance. All on-board equip- ment is managed by a data acquisition system on a central computer. This computer is responsible for integrating the many streams of data that can be collected through the vehi- cle. The computer stores basic driving data such as brake and throttle position and steering wheel angle, gathered through potentiometers located on the pedals and steering column. A global positioning system (GPS) provides real-time data on the exact position of the vehicle, enabling the calculation of location, distance traveled, and velocity. A series of video cameras provided information on adja- cent traffic conditions and in-vehicle driver behaviors. One in-vehicle camera was positioned to monitor drivers’ head turns and glance direction. Another camera was positioned to monitor foot activity on the pedals. While the main source of pedal activity came from the potentiometers, the video record of the feet provided an opportunity to check the source of any anomalies in the pedal potentiometer data. Researchers recruited 12 subjects to drive the instrumented vehicle through the predetermined course. The 12 subject drivers were recruited from the general population in the Dallas/Ft. Worth area; they were not affiliated with TTI, nor had they previously driven the instrumented vehicle. Subjects met at a designated hotel in the area for a briefing session to complete an informed consent form and a demo- graphics questionnaire and receive pre-driving instruc- tions. The script for the verbal instructions to each subject is provided in Appendix C (available on the TRB website at http://www.trb.org/Main/Blurbs/167516.aspx). Three sub- jects drove for approximately 90 minutes each, during each day of the experiment; two subjects drove in the morning and one subject drove in the afternoon. A typical day’s schedule had subjects beginning their tasks at 8:00 AM, 10:30 AM, and 1:30 PM. This allowed the subjects to avoid a majority of the daily commuter traffic. The subjects were given instructions during the brief- ing, prior to beginning the driving course; they were told before they started driving and were reminded after they started driving that they were in complete control of the vehicle at all times and were responsible for its safe operation. Two researchers were inside the vehicle during S e c t i o n 6 Behavioral Study of Freeway Mainline Ramp Terminals

104 the study to offer additional instructions as needed: one researcher sat in the front passenger seat, giving directions and acting as a safety observer, while the other researcher served as the data recorder, sitting in the rear passenger seat and operating the computer. Subjects were instructed to drive normally, obey the speed limit, and follow the driving directions offered during the course, but they were not told that freeway merging and diverging were the focus of the study. At the conclusion of the driving course, subjects were paid $50 for their time. The subjects drove from the hotel parking lot to I-635 via neighborhood roads. During this time the subjects accli- mated themselves to the instrumented vehicle. The pre- defined route (see Figure 55) took the subjects on I-635 as far west as Royal Lane and on I-35E as far south as Medical Dis- trict Drive. On the freeway, the subjects were given directions on when to exit as they passed certain pre-selected state or federal guide signs. On the surface streets, subjects were given advance driving directions and repeated driving directions, if needed. Data were collected at nine entrance ramps and nine exit ramps; a summary of site characteristics for these ramps is provided in Table 43. All subjects drove all 18 ramps except for Subject 7, who was instructed to bypass Ramps 16 and 17 due to an incident blocking both ramps. As the subjects drove through each ramp, two notes were input into the computer; the first note indicated which ramp was approaching and the second note indicated which ramp was just completed. These notes helped to identify the boundar- ies of each ramp within the data file. In addition, four keystroke data points were recorded on each ramp. For entrance ramps these points were: 1. Beginning of ramp or end of controlling feature, 2. Point at which the ramp edgeline changes from yellow to white, 3. Merge location, and 4. End of taper. For exit ramps they were: 1. Beginning of taper, 2. Diverge location, Figure 55. Behavioral study driving route (Image Credit: Google EarthTM Mapping Service).

105 3. Point at which the ramp edgeline changes from white to yellow, and 4. End of ramp or beginning of controlling feature. Figure 56 illustrates generalized entrance and exit ramps with the four key points highlighted. While the subjects drove the course, sensors on the vehicle recorded cumulative elapsed time, distance traveled, displacement of the throttle and brake pedals, X-Y-Z coordinates from GPS, velocity, and direction traveled. In addition, video cameras recorded the driver’s facial and head movements, the driver’s foot move- ments and pedal activity, and the driver’s view of the environ- ment in front of the vehicle. 6.2 Data Reduction After post processing the spreadsheet file for each subject, researchers reviewed the data in the spreadsheets and the cor- responding video recordings to quantify glance activity and pedal activity, determine the presence of lead vehicles, and develop speed-distance plots for each of the 18 ramps on the driving course. Technicians reviewed the video recordings of the subjects’ behavior while merging onto the nine entrance ramps to observe glances that they made in the center or side mirror or over their shoulder. For each merge maneuver, a technician watched the movement of the driver’s head, shoulders, and torso, to clas- sify four different categories of glances, as shown in Table 44. Ramp Ramp type Merge/ diverge type Length (ft) Design speed (mi/h) No. Location Type Painted nose to ctrl feat SCL Taper Freeway Ramp 1 I-635/Josey Ln WBEntrance Straight Parallel 680 710 0* 65 0*** 2 I-635/Freeport Pkwy WB Exit Loop Taper 400 100 180 70 25 3 I-635/Royal Ln EBEntrance Straight Taper 1,660 345 490 70 0*** 4 I-635/Freeport Pkwy EBExit Loop Taper 410 110 160 70 25 5 SH 114/Freeport Pkwy EBEntrance Straight Parallel 1,850 675 280 70 0*** 6 SH 114/Belt Line Rd EBExit Straight Taper 1,010 310 585 70 0*** 7 SH 114/Belt Line Rd EBEntrance Straight Taper 810 650 385 70 0*** 8 SH 114/Rochelle Blvd EBExit Straight Taper 580 70 160 70 0*** 9 SH 114/Rochelle Blvd EBEntrance Straight Parallel 1,010 1,120 0* 70 0*** 10 SH 183/Regal Row EBExit Straight Parallel 665 355 0** 70 0*** 11 I-35E/Mockingbird Ln SBEntrance Straight Taper 660 180 250 65 0*** 12 I-35E/Medical District Dr SB Exit Straight Taper 235 50 100 65 45 13 I-35E/Inwood Rd NBEntrance Straight Parallel 230 200 450 65 45 14 I-635/Marsh Ln EBExit Straight Taper 995 40 135 65 35 15 I-635/Midway Rd WBEntrance Straight Taper 700 145 530 65 0*** 16 I-635/Webb Chapel Rd WB Exit Straight Taper 650 80 185 65 0*** 17 I-635/Marsh Ln EBEntrance Straight Taper 1,115 300 630 65 0*** 18 I-635/Midway Rd EBExit Straight Taper 980 35 170 65 0*** * The path of these entrance ramps carried through to the adjacent downstream exit ramp; therefore, there was no lane departure taper on these ramps. ** This exit ramp was an “exit-only” freeway lane that was carried through to the ramp proper; therefore, there was no lane addition taper on this ramp. *** For the purposes of comparison with Green Book Exhibits 10-70 and 10-73, these ramps have an estimated design speed of 0 mi/h, corresponding to a stop condition. Table 43. Characteristics of ramps used in behavioral study driving route.

106 Technicians also reviewed the video to classify three types of pedal activities related to foot movements, using the terms in Table 45. Using the locations of the key points on each ramp, research- ers identified when each driver passed the beginning of the ramp, passed the change in edgeline, made the lane change to initiate the merge/diverge maneuver, and passed the end of the ramp. Using those four key points on each ramp, researchers divided each ramp into four stages and then plotted the data from 10 s prior to Point 1 to 10 s after Point 4. An example of such a plot for an entrance ramp is shown in Figure 57. The zero distance along the x-axis (i.e., cumulative distance in feet) corresponds to the painted nose of each ramp. Figure 57 shows the speed of the instrumented vehicle as it increases from 25 mi/h at a point 1,250 ft upstream of the painted nose to 60 mi/h at a point 1,250 ft downstream of the (a) Entrance Ramp (b) Exit Ramp • 1 – Beginning of taper • 2 – Diverge point, defined as the location where the vehicle’s right tire s crossed the lane line into the SCL • 3 – Point at which ramp edgeline changes from white to yellow • 4 – End of ramp or beginning of controlling feature Painted Nose • 1 – Beginning of ramp or e nd of controlling feature • 2 – Point at which ramp edgeline changes from yellow to white • 3 – Merge point, defined as the location where the vehicle’s left tires crossed the lane line into the freeway mainlane • 4 – End of taper Painted Nose Figure 56. Key points along mainline freeway-ramp terminals. Code Definition A Relaxed body, head turn (i.e., subject’s back against the driver’s seat, only the subject’s head is turned; typical glance into a mirror) B Relaxed body, shoulder turn (i.e., subject’s back against the driver’s seat, the subject’s head and shoulders are turned; may be a glance into a mirror or through the window) C Non-relaxed body, head turn (i.e., subject’s back turned away from the driver’s seat, only the subject’s head is turned; typical glance through the window) D Non-relaxed body, shoulder turn (i.e., subject’s back against the driver’s seat, the subject’s head and shoulders are turned; typical glance through the window) Table 44. Head movement codes and definitions. Code Definition Accel Gas pedal (i.e., throttle) is depressed Decel Brake pedal is depressed Coast Foot is positioned between pedals for greater than 0.3 s Table 45. Foot movement codes and definitions.

107 painted nose. It also shows the following driver and vehicle characteristics corresponding to those times and locations: • The stage of the ramp being traversed, expressed as a numerical code: – 5 = within 10 s before passing Point 1, – 10 = Stage 1, between Points 1 and 2, – 20 = Stage 2, between Points 2 and 3, – 30 = Stage 3, between Points 3 and 4, and – 40 = Stage 4, within 10 s after passing Point 4; • Throttle and brake pedal use; • The presence of a lead vehicle ahead of the instrumented vehicle near enough to affect the subject’s desired speed; • The occurrence of any glances by the subject into a mirror or through the side window. Figure 57 shows that this subject entered the ramp approxi- mately 750 ft upstream of the painted nose, passed the change in edgeline about 500 ft later, and began the lane change to merge about 250 ft downstream of the painted nose, with the end of the taper approximately 700 ft downstream of the painted nose. Within that distance, the subject depressed the brake pedal once, with as much as 32 percent activation, in proximity to the start of the ramp. During the remainder of the merge activity, the subject activated the throttle, between 18 percent and 40 percent for most of the ramp distance. The driver completed four glances, almost exclusively prior to merging onto the freeway. There was a lead vehicle ahead of the instrumented vehicle for a distance of approximately 400 ft, roughly corresponding to Stage 3; this indicates that the subject accepted a gap in traffic and merged onto the free- way behind another vehicle already in the freeway mainlane. Researchers generated a plot for each subject and each ramp; the 216 plots and their corresponding data led to the sub- sequent analyses discussed in the remainder of this section. 6.3 Entrance Ramp Analyses 6.3.1 Glance Activity Researchers observed a total of 308 glances by the 12 sub- jects on the nine entrance ramps. For glances of all types shown in Table 44, researchers measured the time duration of each glance and calculated the corresponding distance traveled and change in speed. Of the 308 glances, 145 of them began prior to the painted nose, as shown in Figure 58. Tables 46 and 47 present summaries of glance data by ramp and by subject, respectively. The glance summary data indicate that the average glance by a merging driver is typically about 2.5 to 3.0 s long, though some glances are very small and others are very lengthy, as drivers occasionally look into their mirrors for extended peri- ods of time with imperceptible breaks. The first glances on a given ramp were frequently longer than subsequent glances, suggesting that later glances were commonly used to confirm 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 -1250 -750 -250 250 750 1250 Th ro tt le o r B ra ke U se (% ) S pe ed (m i/h ) a nd E ve nt C od es Cumulative Distance (ft) Ramp 1: 635 WB/Josey/Entrance Speed Ramp Pos Throttle Brake Lead Veh Glance Painted Nose Figure 57. Speed-distance plot for freeway entrance ramp. -1500 -1000 -500 0 500 1000 1500 2000 0 5 10 15 20 B eg in ni ng D is ta nc e (ft ) Glance Duration (s) Figure 58. Glances by subject drivers in relative position to the painted nose.

108 the appropriateness of a gap identified previously. Based on the data summarized in the tables, a merging driver traveled 100 to 200 ft and increased speed by 2 to 3 mi/h during an average glance. 6.3.2 Use of Speed-Change Lane This analysis investigated the proportion of the SCL each subject used at each ramp. For example, if a subject’s merge point was 350 ft downstream of the painted nose, and the SCL length was 500 ft, the amount of SCL used was (350/500) = 70 percent. Figure 59 illustrates the SCL usage data at entrance ramps on the driving course. Each marker represents the portion of SCL used by one subject on an entrance ramp; the solid line connects the average SCL percent values for each ramp. The data table attached to the chart shows the summary statistics for each ramp. The data in Figure 59 suggest that on ramps with SCLs less than 350 ft in length (Ramps 3, 11, 13, 15, and 17), subject drivers used all of the provided SCL length and completed their merge in the taper. For the four ramps with SCLs longer than 600 ft (Ramps 1, 5, 7, and 9), most subjects completed their merge within the SCL. Ramp 1 has a noticeably lower average than the other eight ramps. This suggests that drivers treated this ramp differently. Given that it was the first ramp encountered on the driving course for each subject, earlier merging on Ramp 1 may have been a function of the subjects adjusting to driving on the freeway in the instrumented vehicle. The negative sign on the minimum value for Ramp 1 indicates that this driver merged onto the freeway early, upstream of the painted nose. 6.3.3 Acceleration Researchers used the speed and time data from each sub- ject to develop speed profiles for each subject on each ramp, which could then be used to evaluate acceleration patterns. Figure 60 shows the recorded acceleration rates for each sub- ject on each ramp, represented as constant acceleration from the beginning of the entrance ramp (Point 1 in Figure 56a) to the merge point (Point 3 in Figure 56a). Each marker rep- Ramp no. Average no. glances/ subject Min glance duration Max glance duration Average glance duration Average distance (ft) Average speed increase per glance for all subjects (mi/h) 1 2.7 0.6 15.3 3.0 177 2.8 3 4.3 0.4 6.9 2.3 149 2.7 5 2.7 0.9 11.6 2.8 212 2.6 7 2.3 0.8 7.5 2.9 196 3.9 9 3.5 0.2 5.9 2.3 160 3.2 11 3.2 0.8 9.7 3.0 100 2.3 13 2.4 0.9 10.7 2.5 176 2.2 15 3.0 0.8 6.3 2.2 127 2.2 17 1.8 0.7 5.8 2.5 119 0.4 All 2.9 0.2 15.3 2.6 155 2.5 (s) (s) (s) Table 46. Summary data for glance behavior by entrance ramp. Subject no . Average. glances/ ramp Min glance duration Max glance duration Average glance duration Average distance (ft) Average speed increase per glance for all ramps (mi/h) 1 0.4 1.5 2.6 2.2 140 2.8 2 2.8 1.1 8.1 2.9 178 2.9 3 3.7 1.1 11.7 3.5 229 4.0 4 2.9 0.6 5.8 2.5 154 2.6 5 2.6 1.3 6.7 3.0 176 2.5 6 5.0 0.8 11.6 2.7 140 2.2 7 2.6 0.2 15.3 4.0 234 4.0 8 4.2 0.9 5.5 2.1 115 1.7 9 3.3 0.9 6.3 2.2 140 2.2 10 2.6 0.8 3.0 1.5 95 1.6 11 2.2 0.4 3.3 1.6 106 1.5 12 2.0 0.9 5.8 2.7 164 3.2 All 2.9 0.2 15.3 2.6 155 2.5 (s) (s) (s) no. Table 47. Summary data for glance behavior by subject.

109 -50% 0% 50% 100% 150% 200% 250% 300% 350% 1 3 5 7 9 11 13 15 17 % o f S CL Ramp Number Subject Data Average SC L le ng th (f t) 710 345 675 650 1120 180 200 145 300 Ta pe r/ pa ra ll el Ta pe r Ta pe r Pa ra ll el Ta pe r Pa ra ll el Ta pe r Pa ra ll el Ta pe r Ta pe r St ra ig ht /c ur ve St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht Mi n SC L % –5 .3 % 112. 0% 64. 1% 70. 4% 49. 5% 109. 8% 115. 7% 195. 1% 121. 3% Ma x SC L % 55. 0% 192. 8% 83. 5% 144. 1% 82. 5% 163. 7% 191. 0% 330. 4% 240. 9% Av g SC L % 21. 0% 153. 6% 72. 7% 88. 7% 58. 7% 131. 7% 161. 1% 267. 3% 186. 5% St d. de vi at io n 19. 7% 24. 1% 6. 5% 25. 1% 8. 7% 18. 9% 26. 0% 42. 7% 34. 6% Figure 59. SCL use at entrance ramps. -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 3 5 7 9 11 13 15 17 A cc el er at io n (ft /s2 ) Ramp Number Av g Ob se rv ed A cce l (f t/ s 2 ) 1. 07 1. 13 1. 28 1. 86 1. 56 2. 87 1. 22 1. 38 1. 57 Av g A cce l Di st an ce (f t) 914 1, 441 1, 340 908 1, 270 734 585 1, 116 1, 358 Figure 60. Acceleration rates at entrance ramps.

110 resents the acceleration rate of one subject on an entrance ramp; the solid line connects the average rate for each ramp. The data table attached to the chart shows the summary sta- tistics for each ramp. The first row of the data table shows the average observed acceleration rate for all subjects to acceler- ate from the beginning of the ramp (Point 1) to the merge point (Point 3). The average distance between Points 1 and 3 on each ramp is listed in the second row. The average accel- eration rates were typically between 1.0 and 1.5 ft/s2, though rates at Ramps 7 and 11 were somewhat higher. Ramp 11 is unique because it lies just upstream of another entrance ramp and its merging area; thus, it is necessary for drivers entering at Ramp 11 to ensure they have attained the prevailing free- way speed to facilitate merging prior to the subsequent ramp. This feature encourages drivers to accelerate at a higher rate than on other ramps. Causes for the higher acceleration rate at Ramp 7 are not as clear; it could be related to the fact that drivers entered the ramp immediately after exiting the free- way, with no driving time on local roads. However, as great an effect is not shown in Ramp 9, which also was a part of an “off-and-on” sequence of events in the driving route. Table 48 shows a comparison between the observed acceleration rates and distances and minimum acceleration lengths and assumed acceleration rates from Exhibit 10-70 of the Green Book. Because the entrance ramps on the driv- ing route were straight ramps, the controlling feature at the beginning of the ramp was always the crossroad terminal. At some sites, the terminal was at a signal-controlled intersec- tion, while at others the terminal was located on a one-way frontage road. For the latter condition, it was not necessary for vehicles to stop before entering the ramp. Thus, the speed of the instrumented vehicle at Point 1 was not always equal to zero. The observed rates are less than the Green Book rates at all locations except Ramp 11. This indicates that drivers are typically more casual in their acceleration under uncongested conditions than the Green Book rates suggest. The observed acceleration distances are also shorter than those offered by the Green Book, which initially seems coun- terintuitive in conjunction with lower acceleration rates. A review of the data indicates that subjects in this study com- monly merged at speeds lower than the merge speed assumed by the Green Book. In addition, the Green Book length extends to the point at which the width of the SCL decreases below 12 ft, while the observed distances were measured to the point of merge, which were often upstream of the point at which the 12-ft threshold was crossed. Thus, while some subjects used much of the SCL, many did not, producing average accel- eration distances shorter than those listed in the Green Book. This suggests that the Green Book distances are sufficient to accommodate typical merge maneuvers by this population of 12 subject drivers. 6.4 Exit Ramp Analyses 6.4.1 Coasting The objective of this analysis was to examine the valid- ity of the assumption in the Green Book methodology that drivers decelerate in gear (i.e., coast) for 3 s prior to apply- ing the brake when exiting a freeway. The 1965 Blue Book (AASHO, 1965) states that deceleration is a two-step process: first, the accelerator pedal is released (for a length of time assumed for 3 s) and the vehicle slows in gear without the use of brakes, and second, the brakes are applied. Two graphs were included in Figure VII-15 of the 1965 Blue Book to provide these distances. Previous research (Fitzpatrick and Zimmerman, 2007) concluded that the graphs were based on data from studies conducted in the 1930s and 1940s, but the underlying methodology was carried through to the current edition of the Green Book. Using the reduced and processed data, researchers further examined the details of the data for exit ramp coasting. In the review of coasting data, researchers examined three time values for each subject on each ramp: • Throttle Release Time: The elapsed time between the occurrence of peak speed and the deactivation of throttle (i.e., the time spent to remove the foot from the pedal). Ramp no. Parallel/ taper Design speed (mi/h) Observed average Green Book Exhibit 10-70 Freeway Ramp Accel (ft/s2) Accel (ft/s2) Distance (ft) Distance (ft) 1 Taper 65 Stop Condition1 .07 914 1.92 1,410 3 Taper 70 Stop Condition 1.13 1,441 1.87 1,620 5 Parallel 70 Stop Condition 1.28 1,340 1.87 1,620 7 Taper 70 Stop Condition 1.86 908 1.87 1,620 9 Parallel 70 Stop Condition 1.56 1,270 1.87 1,620 11 Taper 65 Stop Condition 2.87 734 1.92 1,410 13 Parallel 65 45 1.22 585 1.62 600 15 Taper 65 Stop Condition 1.38 1,116 1.92 1,410 17 Taper 65 Stop Condition 1.57 1,358 1.92 1,410 Table 48. Comparison of constant acceleration rates and distances.

111 • No Pedal Time: The elapsed time between the deactivation of throttle and the activation of brake (i.e., the amount of time when neither throttle nor brake was in use). • Throttle Release 1 No Pedal: The sum of the previous two time values. Visual exploration of the data reveals that the times for each ramp follow a lognormal distribution rather than a nor- mal distribution (skewed to the right) and the center differs by ramp. This finding is intuitive since the minimum times bounded below by 0 and 50 percent of the Throttle Release + No Pedal times are less than 3.5 s, but values as high as 13.3, 16.1, and 32.4 s were also observed. The geometric mean and 95 percent confidence intervals for the three times are pre- sented in Table 49 for each ramp. Observed times less than 0.1 s were rounded up to 0.1 s. The results for all ramps are based on the random effects model described below. Review of the data in Table 49 indicates that the No Pedal times of subject drivers were noticeably shorter than the 3.0 s assumed by the Green Book, never averaging more than 1.54 s for any ramp. The minimum No Pedal time recorded was less than 0.1 s, and the maximum was 8.8 s. Adding the Throttle Release time to the No Pedal time produces results larger than 3 s on all but two ramps. To determine the statistical significance of the results as compared to the assumed 3-s average coasting time, a ran- dom effects model was estimated for each of the times based on the log transformed values. The model includes an inter- cept term and a random effect for each ramp. The model does not account for the correlation between measurements from the same driver on exit ramps due to insufficient degrees of freedom available for testing. The variability in time within each ramp is assumed to be equivalent for all ramps. The esti- mated overall average time was compared to the 3 s assumed time. The results of the t-test for the No Pedal time are pre- sented in Table 50, and the results for the Throttle Release + No Pedal times are presented in Table 51. The results show that average No Pedal time of 0.98 s is statistically different from 3 s, while the average Throttle Release + No Pedal time of 3.07 s is not statistically different from 3 s. Further investi- gation of the coasting data indicates 2 s of the coasting time typically occurs prior to the diverge maneuver, and 1 s of the coasting time occurs within the SCL following the diverge maneuver. 6.4.2 Use of Speed-Change Lane This analysis investigated how much of the SCL each sub- ject used at each exit ramp. This was calculated by dividing the measured distance between Point 2 and the painted nose by the length of the SCL. Figure 61 shows the location at which drivers entered the SCL on exit ramps on the driving course, expressed as a percentage of distance into the SCL. Each marker in Figure 61 represents the distance into the Throttle release time (s) No pedal time (s) Throttle release +no pedal (s) Ramp No. Average 95% CI Average 95% CI Average 95% CI 2 1.47 (0.70,3.10) 1.00 (0.60,1.66) 3.21 (2.17,4.74) 4 1.08 (0.47,2.45) 1.27 (0.72,2.23) 3.02 (1.92,4.73) 6 1.42 (0.98,2.07) 1.26 (0.69,2.31) 3.29 (2.45,4.41) 8 2.56 (1.12,5.83) 0.78 (0.54,1.12) 3.86 (2.22,6.71) 10 1.76 (0.95,3.25) 0.80 (0.46,1.42) 3.25 (2.25,4.69) 12 0.88 (0.33,2.33) 0.89 (0.52,1.53) 2.09 (1.01,4.32) 14 2.74 (1.08,6.95) 1.27 (0.66,2.47) 6.17 (3.93,9.69) 161 0.61 (0.16,2.29) 0.40 (0.17,0.94) 0.91 (0.23,3.57) 18 2.51 (1.56,4.04) 1.54 (0.80,3.00) 4.55 (2.88,7.17) All 1.53 (1.05,2.23) 0.98 (0.74,1.30) 3.07 (2.07,4.55) 1 Only six times were recorded; 12 were recorded for all other ramps. Table 49. Summary statistics for coasting data. Average log time (log s) Average time (s) Standard Error (log s) DF t-statistic P-value -0.02 0.98 0.12 8 -9.18 <0.01 Note: DF = degrees of freedom. Table 50. Results of random effects model test for average no pedal time 5 3 s (log–scale transformed).

112 SCL for one subject on an exit ramp, and the solid line con- nects the average SCL distance values for each ramp. The data table attached to the chart shows summary statistics for each ramp. Diverging drivers in this study tended to complete their maneuvers within the second half of the SCL or later. The later drivers did not travel in the provided SCL, but instead completed their diverge maneuver beyond the painted nose; they are represented in Figure 61 as greater than 100 percent of SCL. The latter finding is somewhat related to the SCL length, since drivers on the two ramps with SCLs longer than 300 ft (Ramps 6 and 10), generally completed their maneu- vers within the SCL, while the incidences of late diverges were more common on the ramps 110 ft or shorter. Ramp 10 is also unique because the approach to Ramp 10 is a through lane that changes to an exit-only lane. Thus, there is no taper upstream of the SCL at Ramp 10; only a change in lane line from a dotted or “skip-stripe” line to a solid line denotes the beginning of the SCL. It appears that there may be a bit of a learning process as drivers see ramps that are similar in type. Ramps 16 and 18 have two of the shortest SCLs in the study, but they have the lowest average diverge point. The averages are influenced by two drivers at each ramp who began their diverge in the taper, upstream of the SCL and represented in Figure 61 as less than 0 percent of SCL. However, there is a lower frequency of late merges at these two ramps than in other ramps with SCLs of similar length. Average log time (log s) Average time (s) Standard Error (log s) DF t-statistic P-value 1.12 3.07 0.17 8 0.13 0.90 Table 51. Results of random effects model test for average throttle release 1 no pedal time 5 3 s (log–scale transformed). -200% -150% -100% -50% 0% 50% 100% 150% 200% 250% 2 4 6 8 10 12 14 16 18 % o f S CL Ramp Number Subject Data Ave rage SC L Le ng th (f t) 100 110 310 70 355 50 40 80 35 Ta pe r/ Pa ra lle l Ta pe r Ta pe r Ta pe r Ta pe r Pa ra ll el Ta pe r Ta pe r Ta pe r Ta pe r St ra ig ht /C ur ve Cu rv e St ra ig ht Cu rv e St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht St ra ig ht Mi n SC L % 39. 0% 45. 8% 77. 0% 34. 4% 17. 7% 58. 7% 10. 2% – 67. 4% –1 81 .2 % Ma x SC L % 160. 6% 149. 1% 103. 5% 146. 7% 99. 8% 167. 0% 183. 2% 120. 7% 113. 9% Av g SC L % 108. 3% 80. 2% 94. 8% 86. 1% 82. 1% 110. 6% 86. 0% 39. 6% 40. 9% St d. De vi at io n 42. 9% 30. 8% 8. 9% 30. 1% 27. 3% 28. 5% 43. 9% 58. 4% 86. 2% Figure 61. SCL use at exit ramps.

113 6.4.3 Deceleration Similar to the acceleration analysis for entrance ramps, researchers used the speed and time data from each subject to develop speed profiles for each subject on each exit ramp, which were then used to evaluate deceleration patterns. Fig- ure 62 shows the recorded deceleration rates for each subject on each ramp, represented as constant deceleration from the diverge point (Point 2 in Figure 56b) to the critical feature (Point 4 in Figure 56b). Each marker represents the decel- eration rate of one subject on an individual exit ramp; the solid line connects the average rate for each ramp. The data table attached to the chart shows the summary statistics for each ramp. The first row of the data table shows the average observed deceleration rate for all subjects to decelerate from the diverge point (Point 2) to the end of the ramp (Point 4). The average distance between Points 2 and 4 on each ramp is listed in the second row. The average deceleration rates were typically between –1.5 and –3.0 ft/s2, though the rate at Ramp 16 is somewhat less. Ramp 16 is unique because the ramp terminates at a one-way frontage road approximately 600 ft upstream of the signalized intersection of the adjacent cross- road. Thus, in the typically uncongested conditions in which subjects traveled through this ramp, there was an added dis- tance of approximately 600 ft in which to decelerate, which reduced the need for more pronounced deceleration on the ramp. Ramp 12 was the only other exit ramp to terminate at a frontage road, and the frontage road volumes at that loca- tion were much higher, reducing the possibility of using the frontage road for substantial deceleration. Table 52 shows a comparison between the observed decel- eration rates and distances and minimum deceleration lengths and assumed deceleration rates from Exhibit 10-73 of the Green Book. Deceleration rates based on Green Book criteria are provided assuming a constant deceleration and the assumed deceleration rates during primary braking. Comparison of the observed rates and Green Book rates shows that the observed rates are less than the Green Book rates at all locations. A review of the speed profiles indicates that every subject began deceleration prior to entering the SCL at every location; this means that the needed decel- eration was accomplished over a greater distance than the length of the SCL and the distance to the controlling feature. As a result, drivers decelerated more gradually than the rates implied by the Green Book. Further review of the data indicates that, while average deceleration rates are useful for summarizing the data, the actual deceleration rates are variable, commonly increasing as the driver approaches the end of the ramp. While the final deceleration rates are not necessarily equal to those obtained from Green Book values, they are typically closer in value than the average rates shown in Table 52. Given these deceleration characteristics of subject drivers, the deceleration length val- ues provided in the Green Book appear sufficient to accom- modate typical diverge maneuvers. Although the observed Av g Ob se rv ed Dece l (f t/ s 2 ) –2 .9 0 –2 .9 9 –1 .9 0 –2 .3 1 –2 .4 8 –2 .1 9 –1 .8 2 –0 .5 4 –1 .6 4 Av g Dece l Di st an ce (f t) 575 474 682 587 808 236 615 669 965 -6 -5 -4 -3 -2 -1 0 1 D ec el er at io n (ft /s2 ) 2 4 6 8 10 12 14 16 18 Ramp Number Figure 62. Deceleration rates at exit ramps.

114 deceleration values were less than assumed in the Green Book, none of the drivers appeared uncomfortable and likely could have tolerated greater deceleration if necessary. 6.5 Summary of Significant Findings The most significant findings from the examination of behavioral data are as follows: • In uncongested or lightly congested conditions, a typical glance into a mirror or over the shoulder by a driver merging onto the freeway is typically about 2.5 to 3.0 s, but the driver tends to take three such glances on a given entrance ramp. During the typical glance, the driver travels between 100 and 200 ft and increases speed by approximately 2.5 mi/h. • Merging drivers in uncongested or lightly congested con- ditions tended to use at least half of the SCL provided when entering the freeway, and often used the taper area to complete their merging maneuvers. • Observed acceleration rates were lower than the assumed acceleration rates from the Green Book. Additionally, observed acceleration distances were shorter than those offered by the Green Book. This seems counterintuitive, but subjects in this study commonly merged at speeds lower than the merge speed assumed by the Green Book. • The assumption in the Green Book that drivers deceler- ate in gear (i.e., coast) for 3 s was not statistically different from the observations for the drivers in this study, if the Green Book definition of coasting includes the time used to remove the driver’s foot from the throttle. If coasting is defined as the activation of neither throttle nor brake, the appropriate value for this driver population is approxi- mately 0.98 s. • Diverging drivers tended to use the freeway through lane for a large portion of their deceleration when exiting the freeway, and seldom entered the SCL within the first 50 per- cent of the provided length. Drivers on the two ramps with SCLs longer than 300 ft, generally completed their maneu- vers within the SCL, while the incidences of late diverges were more common on the ramps 110 ft or shorter. • Every subject began deceleration prior to entering the SCL at every location; this means that the needed deceleration was accomplished over a greater distance than the length of the SCL and the distance to the controlling feature. As a result, drivers decelerated more gradually than the rates implied by the Green Book. • Actual deceleration rates are variable, commonly increas- ing as the driver approaches the end of the ramp. • Deceleration length values provided in the Green Book appear sufficient to accommodate typical diverge maneu- vers by this subject driver population. Although the observed deceleration values were less than assumed in the Green Book, none of the drivers appeared uncomfort- able, and there was no indication that the drivers could not tolerate greater deceleration if necessary as assumed in the Green Book criteria. Ramp no. Diverge type Design speed (mi/h) Observed average Green Book Exhibit 10-72 Freeway Ramp Distance (ft) Decel (ft/s2) Distance (ft) Constant decel (ft/s2) Decel during braking (ft/s2) 2 Taper 70 25 575 –2.90 550 –5.63 –7.77 4 Taper 70 25 474 –2.99 550 –5.63 –7.77 6 Taper 70 StopCondition 682 –1.90 615 –5.88 –7.40 8 Taper 70 StopCondition 587 –2.31 615 –5.88 –7.40 10 Parallel 70 StopCondition 808 –2.48 615 –5.88 –7.40 12 Taper 65 45 236 –2.19 340 –4.51 –7.66 14 Taper 65 35 615 –1.82 440 –5.19 –7.49 16 Taper 65 StopCondition 669 –0.54 570 –5.71 –7.55 18 Taper 65 StopCondition 965 –1.64 570 –5.71 –7.55 Table 52. Comparison of constant deceleration rates and distances.