Cover Image

Not for Sale

View/Hide Left Panel
Click for next page ( 134

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement

Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 133
133 Exhibit 53. Summary of availability and For the intended crossing phase (Red/Walk), 39.3% of cross- utilization statistics, PHB crosswalk, post. ings were classified as correct utilizations of crossing oppor- tunities, and 23.4% of events were correctly rejected events. Post (n = 242) Near Lane Far Lane Further, 4.7% were classified as missed opportunities and Availability Statistics inefficient behavior, and none were observed in the potentially P(Y_Enc) 76.4% 68.6% risky category. Similar to the early phases, many events (32.7%) P(CG_Enc) 70.2% 19.8% fell into the inconclusive category and were associated with a Utilization Statistics forced yield in either the near or far lane. Those pedestrians that P(GO|Y) 92.4% 97.6% waited to initiate crossing during the "Flashing Red" phase P(GO|CG) 100.0% 100.0% mostly made correct "GO" decisions, but more than half were once again associated with forced yields. A notable difference between Exhibits 49 and 53 for pre and Those pedestrians who rejected opportunities in the "Walk" post data is a drastic reduction in the rejected opportunities. phase ultimately crossed in the "Flashing Don't Walk" phase. With the introduction of the signal, drivers tended to yield Again, the majority of events here are related to yield events much more frequently, and many of these yields resulted in with drivers stopped at the signal. No rejected opportunities crossings. The proportion of inefficient decisions was reduced or inefficient events were observed, but again many events fall slightly, as was the rate of potentially risky events. The rate of into the inconclusive category. Exhibit 53 summarizes the avail- inconclusive events saw a large increase. As discussed above, ability and utilization statistics for the post treatment installa- these events are associated with forced yields, where none tion data. Since it appeared from the analysis above that most resulted in an O&M intervention. It is unclear whether a pedestrians crossed independently of the signal indication, the forced yield at a signal can truly be classified as a risky event results are presented in light of the nearfar lane framework dis- since drivers are presumably prepared to stop given that the cussed above. This also ensures that the numbers are directly signal has been activated. However, in combination with red- comparable to the pre condition results. light running events, some risk may remain. The discussion The summary statistics suggest a large increase in the below examines in more detail the risk and delay performance availability of both yields and gaps from the pre condition measures. (Exhibit 49), as well as more efficient utilization of these crossing opportunities. Exhibit 54 shows a summary of all Performance Statistics events for the post condition by signal phase. at the PHB Crosswalk Exhibit 54 shows that crossing performance in the early phases ("Flashing Yellow" and "Solid Yellow") was generally The changed pedestrian and driver behavior that took place characterized by mostly correctly accepted crossing opportu- with the introduction of the PHB affects the delay and risk per- nities as well as a large portion of forced yields (inconclusive formance measures. Delay statistics in Exhibit 55 are provided events). Virtually no risky or inefficient events were observed for pedestrian delay in seconds, defined as the time differ- during these phases. ence between the time a trial started and when the pedestrian Exhibit 54. Summary of pedestrian behavior post condition, PHB crosswalk. Pedestrian Crosswalk Condition Decision Crossable/Safe Non-Cross./Unsafe Inconclusive Phase = Flashing Yellow/Don't Walk (n = 23) GO 5 21.7% 0 0.0% 18 78.3% NoGO 0 0.0% 0 0.0% 0.0% Phase = Yellow/Don't Walk (n = 82) GO 39 47.6% 1 1 .2% 42 51.2% NoGO 0 0.0% 0 0.0% 0.0% Phase = Red/Walk (n = 107) GO 42 39.3% 0 0.0% 35 32.7% NoGO 5 4.7% 25 23.4% 0.0% Phase = Flashing Red/Flashing Don't Walk (n = 28) GO 12 42.9% 0 0.0% 16 57.1% NoGO 0 0.0% 0 0.0% 0.0%

OCR for page 133
134 Exhibit 55. Average pedestrian delay statistics for Exhibit 55 shows that the average pedestrian delay per leg in PHB crosswalk. the post condition decreased significantly from that in the pre a) Observed Delay per Leg (s) condition, from 16.0 s to 4.2 s (p = 0.0007). There was no sig- Pre Avg. Min. Max. Std. Dev. nificant difference between the delay experienced at the entry Entry (n = 18) 14.9 2.3 36.5 10.1 and exit portions of the crossing in either study. In addition to Exit (n = 18) 17.1 3.6 46.5 12.0 reporting the average delay for all participants, it is important Overall (n = 36) 16.0 2.3 46.5 11.0 to emphasize that some individual participants experienced Post much larger delays. The highest average delay for a participant Entry (n = 13) 5.9 2.6 14.6 3.2 was 46.5 s in the pre and 14.6 s in the post case. However, the Exit (n = 13) 5.8 3.5 11.7 2.4 single highest delay experienced by a study participant was Overall (n = 26) 5.8 2.6 14.6 2.8 100.2 s (not shown in the exhibit). The single highest delay in b) Delay>Min (s) the post condition was 56.3 s, indicating that some pedestrians Pre Avg. Min. Max. Std. Dev. did not cross during the first "Walk" phase. The reported delay Entry (n = 18) 4.5 0.2 11.9 3.5 figures further do not include trials that were terminated when Exit (n = 18) 3.9 0.0 11.8 3.9 the subject's wait time exceeded the 2-min time-out limit. Overall (n = 36) 4.2 0.2 11.9 3.7 Overall, the 2-min time-out limit was reached in 3 of 288 lane Post crossings for all subjects in the pre condition and never with Entry (n = 13) 1.4 0.2 3.1 1.0 the PHB present (208 lane crossings). Exit (n = 13) 1.4 0.2 3.0 0.9 The results for Delay>Min also show a significant reduc- Overall (n = 26) 1.4 0.2 3.0 0.9 tion between the pre and post conditions from 4.5 s to 1.4 s (p = 0.0044). Overall, the Delay>Min results suggest that the blind pedestrians did not miss many crossing opportunities. initiated the crossing. The exhibit further shows the delay Despite these low averages, some pedestrians experienced beyond the first opportunity (Delay>Min), which was defined Delay>Min of up to 33.0 s in the pre condition and up to 11.4 s as the time difference between the first yield or crossable gap the post case (not shown). The highest average Delay>Min encountered by the pedestrian and the actual crossing initia- values were 11.9 s and 3.1 s, respectively. tion. The crossable gap definition assumed crossing of two Exhibit 56 shows the cumulative distribution of pedestrian lanes at one leg of the roundabout. All statistics shown are cal- delay at the PHB. The 85th percentile delay is highlighted. culated from the average performance of each individual sub- The exhibit clearly shows a shift in the delay distribution, ject. The sample sizes in the pre and post conditions are 18 and with pedestrians in the post condition experiencing much lower 13 participants, respectively. delays. The 85th percentile delay was reduced from 29.8 s to Exhibit 56. Cumulative distribution of pedestrian delay at PHB crosswalk. PHB 100 POST 90 8.7 sec. 80 PRE 85%ILE 29.8 sec. DELAY 70 Percentile 60 50 40 30 PRE 20 POST 10 0 0 10 20 30 40 50 60 70 80 90 100 110 120 Delay (sec.)

OCR for page 133
135 Exhibit 57. 85th percentile delay by subject PHB Exhibit 58. Latency and yield lost time statistics for crosswalk. PHB crosswalk. PHB a) Latency (s) 120 Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 5.0 1.4 10.1 2.9 85th Percentile Delay (sec.) 100 PRE Exit (n = 18) 7.0 3.0 14.9 3.5 POST Overall (n = 36) 5.9 1.4 14.9 3.3 80 Post 60 Entry (n = 13) 4.4 2.3 7.9 1.8 Exit (n = 13) 5.2 2.9 8.2 1.7 40 Overall (n = 26) 4.8 2.3 8.2 1.7 20 b) Yield Lost Time (s) Pre Avg. Min. Max. Std. Dev. 0 Entry (n = 18) 2.7 -1.7 9.9 3.2 1 3 6 9 11 16 4 7 10 12 14 17 2 5 8 13 15 18 Exit (n = 18) 1.2 -4.9 8.7 3.3 Overall (n = 36) 1.9 -4.9 9.9 3.3 9:00am 11:30am 3:30pm Post Subject Entry (n = 13) -0.4 -3.2 3.3 1.9 Exit (n = 13) -0.3 -5.1 2.7 2.4 Overall (n = 26) -0.4 -5.1 3.3 2.1 8.7 s. The difference is also evident when examining the cross- ing performance of individual participants. Exhibit 57 shows the 85th percentile delay for all participants in the pre and post condition. Note that subjects labeled 1, 5, 10, 15, and 16 did not For the YLT measure, pedestrians in the pre condition participate in the post study. waited an average of 2.7 s before crossing in front of an Exhibit 57 shows that the 85th percentile delay was reduced already yielding vehicle. However, the maximum average for every participant in the posttest condition. Further, the YLT was 9.9 s, and individual YLT observations were even effect appeared to be greatest for those subjects who experi- higher. In many cases, drivers may not be willing to wait this enced high delays in the pre study. So, in addition to reduc- long and a high YLT will therefore translate to an increased ing the overall delay, the PHB also created a more uniform percentage of missed yields [lower P(GO|Y)] or even an unsafe distribution of delay, even for participants with presumably condition where both driver and pedestrian proceed simulta- modest travel skills. The data in Exhibit 57 are arranged by neously. Note that the YLT can be negative, suggesting that the time of day the subjects participated in the study. A visual some pedestrians forced vehicles to yield. After installation of comparison does not show a significant effect on performance the PHB, many pedestrians crossed with the signal and thus by time of day. before the vehicles had yielded, resulting in an average YLT of The team further investigated two parameters intended to 0.4 s. The maximum average YLT also decreased to 3.3 s, describe the efficiency with which a crossing opportunity is suggesting a quicker response to the yielding vehicle. utilized. For utilized gaps, the latency is defined as the time dif- The above measures primarily focus on the efficiency of ference between a vehicle entering the crosswalk and the time crossing and largely ignore the explicit risk experienced by the pedestrian initiated the crossing. For utilized yields, the pedestrians. While delay and other efficiency measures are YLT is defined as the time difference between the driver first used frequently by engineers, they fail to capture the human slowing down for a yield and the time the crossing is initiated. element of crossing risk. The selected surrogate risk measure Note that in some cases, pedestrians may prefer to cross only for this study is the number of times the O&M specialist had after a car has come to a full stop (stopped yield) and so some to intervene in the crossing. Exhibit 59 shows the frequency inherent yield utilization time is expected. Exhibit 58 shows and rate of O&M interventions for all trials. statistics for both measures in the pre and post cases. Exhibit 59 shows a drastic reduction in the occurrence of The latency results in Exhibit 58 suggest that on average interventions. The percentage of trials that resulted in an pedestrians wait 5.9 s into a crossable gap before initiating the O&M intervention is reduced from 2.4% to zero in the post crossing, suggesting inefficiency in decision-making. With condition. In the pre case, the entry lane actually had a higher the installation of the PHB, the average latency decreases intervention rate than the exit, which is contrary to findings slightly to 4.8 s; however, that difference is not statistically sig- at other multi-lane roundabouts (Guth et al. 2005). Follow- nificant (p = 0.2363). Both the range and standard deviation ing discussion in Ashmead et al. (2005), a 2.4% likelihood of of the latency estimate are reduced in the post condition. a risky decision will result in a cumulative risk of 62.2% after