Cover Image

Not for Sale

View/Hide Left Panel
Click for next page ( 102

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 101
101 PART 1 Detailed Channelized Turn Lane Results Introduction controlled by protected signal phases. In the PRE condition, all right-turn lanes were free flowing (no signal) and were con- This section describes detailed analysis results of data trolled only by a downstream yield sign. As a result, vehicle collected at the channelized right turn lane in Charlotte, NC, speeds through the turn lane were relatively fast. Vehicle at the intersection of Providence Road and Pineville-Matthews movements were entirely uninhibited and free flowing during Road (Exhibit 1). The focus is on pedestrian-related measures, signal phases where the downstream lanes were clear, which including the availability and utilization of yield and gaps, as well were the signal phases serving the adjacent through movements as pedestrian delay and O&M interventions. on Providence Road and the opposing left turns on Pineville- Two turn lane crosswalks at the CTL sites were studied in Matthews Road. this project: the crosswalks in the southeast (SE) and northwest The treatments tested at the CTL site were intended to (NW) corner. Both turn lanes served the right-turn movements provide a relatively low-cost solution to make the site accessible from Providence Road onto Pineville-Matthews Road. Similar to and usable by pedestrians who are blind. The high-end treat- to other data collection sites, the CTL was studied in a pre and ment would have been a pedestrian signal, which was not tested post study design with treatment installation. The treatments because its effects on crossing performance are predictable. were (1) sound strips that were intended to increase the aware- The treatment tested was sound strips that enhance the auditory ness of pedestrians of approaching vehicles at the NW corner sound patterns of approaching vehicles. The hypothesis was and (2) sound strips in combination with a pedestrian-actuated that sound strips enhance the rate of opportunity utilization flashing beacon that was intended to increase driver yielding of pedestrians and therefore reduce delay. Presumably, sound behavior at the SE corner. In the following discussion, the cross- strips would also help reduce the rate of O&M interventions walks will be identified by treatments installed as SS-ONLY and if the subjects have increased awareness of the presence of SS+FB, respectively. Both turn lanes were further supplemented a vehicle in the turn lane. The sound strips were tested in with lane delineators that were intended to prevent late merges isolation at the SS-ONLY corner; Exhibit 2 shows a photo of into the turn lane. All treatments, including the lane delin- the installation along with the mentioned lane delineators. eators, were installed between pre and post studies. At the SS+FB corner, the sound strip and lane delineator The pre study was completed in May 2008; a total of 16 blind treatment was supplemented with a pedestrian-actuated flash- travelers participated. Fourteen of the original 16 participants ing beacon (Exhibit 3). The FB was intended to increase driver returned for the post experiment in November 2008. The treat- awareness of the pedestrian's intent to cross and thus increase ments were installed in early October 2008, allowing six weeks the rate of yielding. When activated, the FB would transmit for driver adaptation. an audible speech message saying "Flashing Beacon Is On" for the duration of the flashing mode. The flashing mode would terminate after 20 s. CTL Analysis Results Site Description Crossing Statistics for CTL Site and Treatments The CTL site is located at the intersection of two major arterial streets in southeast Charlotte. Providence Road has a The analysis of crossing performance focuses on aspects four-lane cross-section, and Pineville-Matthews Road has a of pedestrianvehicle interaction following the NCHRP six-lane cross-section in the vicinity of the intersection. All Project 3-78A analysis framework. The first analysis component four left-turn movements have dual left-turn lanes and are thus describes the availability and utilization of yields in both the

OCR for page 101
102 Exhibit 1. Aerial view of CTL site. Exhibit 3. FB installation at CTL. Photo by Bastian Schroeder Photo by Google on the basis of all pedestrianvehicle events and not just poten- tial yielders. Chapter 3 provides additional discussion on these pre and post treatment conditions. Two yield measures were and other performance measures, including examples on the used in the analysis: difference between the yielding measures. Treatments at the SS-ONLY crosswalks were sound strips P(Y_ENC): The probability of encountering a yield event, and lane delineators only; the SS+FB corner was further sup- defined as the number of yields divided by the total of all plemented with a pedestrian-actuated flashing beacon. The fig- events encountered by the pedestrian until he/she completes ures shown represent the mean results considering all subjects. the crossing. Each subject completed ten crossing trials at the roundabout, P(GO|Y): The probability of yield utilization, defined by with each trial consisting of two lane crossings (e.g., curb to the number of crossings in a yield divided by total number splitter island and island to curb). For example, a subject in the of yields encountered by the pedestrian. pre condition would have crossed each crosswalk 20 times (twice in each of 10 trials) and would have performed a total of The P(Y_ENC) measure is somewhat different from the 40 crossings at the site. The average performance for each traditionally used probability of yielding since it is calculated crosswalk in the pre condition was then calculated from the mean of these 20 crossing for all 16 subjects. In the post con- ditions, 14 subjects participated in the experiment. In total, 30 subjects were included in the study (16 pre, 14 post) and Exhibit 2. Sound strip installation at CTL. each performed 40 lane crossings, resulting in a theoretical total of 1,200 crossing attempts at this site. However, several subjects appeared to struggle with too many crossing attempts, and the number of trials per crosswalk was therefore capped at six for some participants. Overall, 993 crossings were com- pleted by the participants. Exhibit 4 shows the statistics for the studied crosswalks pre and post treatment installation. Photo by Bastian Schroeder Exhibit 4 shows that the probability of encountering a yield, P(Y_ENC), was not significantly different at the two studied crosswalks in the pre condition (p = 0.2728).The installation of the sound strips and lane delineators at the SS-ONLY corner did not result in a notable change in yield encounters. The added installation of the flashing beacon increased the likelihood of encountering a yield from 15.2% to 22.0%, which is small but significant at p = 0.0363. From a driver perspective

OCR for page 101
103 Exhibit 4. Yield availability and utilization statistics for CTL crosswalks. a) P(Y_ENC) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 18.4% 4.2% 37.5% 8.3% SS+FB (n = 16) 15.2% 6.0% 36.4% 7.9% Post SS-ONLY (n = 14) 18.6% 10.0% 75.0% 20.6% SS+FB (n = 14) 22.0% 0.0% 35.7% 8.9% b) P(GO|Y) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 50.8% 0.0% 100.0% 31.0% SS+FB (n = 16) 53.1% 8.0% 100.0% 28.5% Post SS-ONLY (n = 14) 40.5% 10.0% 75.0% 20.6% SS+FB (n = 14) 64.6% 20.0% 100.0% 28.2% [P(Yield), which is not shown in Exhibit 4], the rate of yield- The analysis next considered the availability and utilization ing increased from 24.1% to 43.1% (p = 0.0123). This means of crossable gaps. For the purpose of this analysis, a crossable that with the installation of the FB, 43% of drivers stopped at gap was defined as a gap greater than 6.5 s, which was sufficient the beacon, but these events still only represented 22.0% to cross the 16-ft crosswalk at a walking speed of 3.5 ft/s while of the encountered vehicle events. The remaining events are allowing for a 2-s buffer. These 2 seconds allowed for some in the form of gaps. pedestrian reaction time before initiating the crossing, as well The exhibit further shows the rates of yield utilization, as a safety buffer between a completed crossing and the next P(GO|Y), defined as the rate of yields that resulted in a pedes- vehicle arrival. Similar to the yield statistics, two gap-related trian crossing the roadway. The yield utilization rates at the parameters are defined: SS-ONLY corner actually appeared to decrease from 50.8% to P(CG_ENC): The probability of encountering a CG event 40.5%, although that change is not statistically significant (p = (gap greater than 6.5 s), defined as the number of crossable 0.2878) due to a high standard deviation. Similarly, the appar- gaps divided by the total of all events encountered by the ent increase in P(GO|Y) at the SS+FB corner from 53.1% to pedestrian. 64.4% is not significant at the given sample size (p = 0.2769). P(GO|CG): The probability of crossable gap utilization, The high standard deviations in the yield utilization measure defined by the number of crossings in a CG divided by total suggest great inter-subject variability. In the range of P(GO|Y), number of CGs encountered by the pedestrian. it is evident that some pedestrians had perfect yield utilization, while others utilized only 8% to 10% of yield opportunities. Exhibit 5 shows the statistics for the studied crosswalk. A fraction of yields further fell into the "forced yield" cat- The results in Exhibit 5 show a slightly higher P(CG_ENC) egory, which is defined as the pedestrian stepping out into the at the SS+FB crosswalk, which is significant at p = 0.0554. With roadway before the vehicle initiated the yielding process. The the installation of the treatments, the rate of crossable gap degree of risk associated with these events depends on the rel- encounter increases for both the SS-ONLY and SS+FB cross- ative position and speed of the vehicle at the time of crossing walks, but neither increase is significant given the high standard initiation. Forced yield events should therefore not necessar- deviations across subjects (p = 0.1666 and 0.4440, respectively). ily be interpreted as poor or risky decisions. In the pre condi- The rates of gap utilization are again comparable between tion, 11.3% and 11.5% of yields were forced at the SS-ONLY SS+FB and SS-ONLY crosswalks in the pre condition. While the and SS+FB crossings, respectively. In the post condition, the SS-ONLY treatments did not significantly affect gap utiliza- corresponding forced yield percentages were reduced to 6.3% tion (p = 0.4238), the added installation of the flashing bea- and 5.9%. This reduction in the percentage of forced yields con increased P(GO|CG) from 63.2% to 89.3% at the SS+FB was not statistically significant due to high standard devia- crosswalk (p = 0.0011). The effect may be attributable to an tions (p = 0.3520 and p = 0.1902 for SS-ONLY and SS+FB, increased level of confidence resulting from the speech message respectively). emitted from the beacon.

OCR for page 101
104 Exhibit 5. Crossable gap availability and utilization statistics for CTL crosswalks. a) P(CG_ENC) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 34.9% 16.9% 64.7% 11.3% SS+FB (n = 16) 44.7% 27.9% 84.6% 16.1% Post SS-ONLY (n = 14) 41.2% 24.7% 62.1% 12.8% SS+FB (n = 14) 49.2% 29.1% 75.0% 15.6% b) P(GO|Gap>Min) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 60.3% 4.0% 100.0% 28.9% SS+FB (n = 16) 63.2% 10.0% 100.0% 24.9% Post SS-ONLY (n = 14) 68.2% 5.8% 100.0% 24.4% SS+FB (n = 14) 89.3% 58.3% 100.0% 13.4% The combined effect of gap and yield availability and uti- roundabout sites, the observed ranges and standard deviations lization is reflected in the delay experienced by pedestrians. of the delay estimates are large, suggesting great variability Delay statistics in Exhibit 6 are provided for two delay measures: across subjects. The highest average delay in the pre study was 80.6 s for one subject, while another one had an average delay Observed Delay per Leg (s): The average pedestrian delay of only 3.7 s across all trials. in seconds, defined as the time difference between when the With installation of the treatments, the SS-ONLY delay trial started and when the pedestrian initiated the crossing. dropped from 26.2 s to 18.5 s, which is not statistically signifi- Delay>Min (s): The delay beyond the first opportunity cant at p = 0.1898. However, it is evident that both the range of (Delay>Min), defined as the time difference between first observed delays and the standard deviation of the estimate yield or crossable gap encountered by the pedestrian and showed corresponding reductions, suggesting at least some the actual crossing initiation. impact from the sound strip installation. The SS+FB corner saw a higher delay reduction, from 23.4 s to 12.2 s, which is Statistics for all measures are for crossing one lane of significant at p = 0.0453. Again, both the range and standard channelized turn lane. deviation of the delay estimate show a reduction, suggesting Exhibit 6 shows similar delays and Delay>Min at both turn more consistent behavior across subjects after treatment lanes in the pre condition (p = 0.6972). As was the case for the installation. The single highest delays for any subject in a trial were 119.0 and 113.1 s in the pre and post conditions, respectively, Exhibit 6. Average pedestrian delay statistics for studied crosswalk. excluding events that were capped at the 2-min time-out limit. Overall, at the SS-ONLY crosswalk the 2-min limit was a) Observed Delay per Leg (s) reached 19 times in the pre and 16 times in the post study Pre Avg. Min. Max. Std. Dev. (with two fewer subjects participating). At the SS+FB crosswalk SS-ONLY (n = 16) 26.2 3.7 80.6 20.7 the time-out was reached 10 and 2 times in the pre and post, SS+FB (n = 16) 23.4 4.1 75.7 19.6 respectively. So despite average delay improvements, isolated Post SS-ONLY (n = 14) 18.5 5.3 34.5 9.2 trials still performed very poorly after treatment installation. SS+FB (n = 14) 12.2 3.2 36.0 8.0 The results for Delay>Min also show corresponding trends. b) Delay>Min (s) There was no significant difference between SS+FB and Pre Avg. Min. Max. Std. Dev. SS-ONLY in the pre condition (p = 0.9089). A small but SS-ONLY (n = 16) 15.6 0.3 65.4 17.4 statistically insignificant drop was evident for the SS-ONLY SS+FB (n = 16) 14.9 0.6 63.4 16.9 corner (15.6 to 11.7 s, p = 0.4224), while the SS+FB corner saw Post a significant reduction from 14.9 to 4.9 s (p = 0.0342). The SS-ONLY (n = 14) 11.7 2.1 27.9 7.5 difference between delay and Delay>Min suggests that quite SS+FB (n = 14) 4.9 0.0 20.4 5.7 a few participants missed crossing opportunities, especially

OCR for page 101
105 when considering the range and standard deviation of the of the pre and post curves for both crosswalks, with a bigger estimates. Similar to delay, the Delay>Min parameter saw effect expectedly at the SS+FB corner. The 85th percentile some tightening in these variability measures after treatment overall delay was reduced from 40.9 to 32.7 s at the SS-ONLY installation. corner, and from 38.6 s to 17.9 s at the SS+FB crosswalk. Exhibit 7 shows the cumulative distribution of delay for Exhibit 8 shows the 85th percentile delay estimate by all subjects in the pre and post conditions for both SS-ONLY subject for SS-ONLY (a) and SS+FB (b) crosswalks. The exhibit (a) and SS+FB (b) crosswalks. The figures show a relative shift makes evident that there is a lot of inter-subject variability. Exhibit 7. Cumulative delay distribution all subjects channelized turn lane. a) SS-ONLY 100 95 90 85 80 75 70 65 60 Percentile 55 50 45 40 35 30 25 PRE 20 15 10 POST 5 0 0 10 20 30 40 50 60 70 80 90 100 110 120 Delay (sec.) b) SS+FB 100 95 POST 90 85 80 PRE 85%ILE DELAY 75 70 65 60 Percentile 55 50 45 40 35 30 25 PRE 20 15 POST 10 5 0 0 10 20 30 40 50 60 70 80 90 100 110 120 Delay (sec.)

OCR for page 101
106 Exhibit 8. 85th percentile delay by subject channelized turn lane. a) SS-ONLY 120 85th Percentile Delay (sec.) 100 PRE POST 80 60 40 20 0 10 13 14 1 3 5 8 11 15 6 12 16 2 4 7 9 8:30am 10:30am 1:30pm 3:30pm Subject b) SS+FB 120 100 PRE 85th Percentile Delay (sec.) POST 80 60 40 20 0 10 13 14 1 3 5 8 11 15 6 12 16 2 4 7 9 8:30am 10:30am 1:30pm 3:30pm Subject The results of the SS-ONLY crosswalk show that some partic- constant (subjects 1 and 3) and one actually experienced slightly ipants seemed to benefit from the sound strips (subjects 1, 4, 7, higher delay after the treatment was installed (subject 6). 9, 11, 12, 13, and 15), while others experienced no difference Subjects 5 and 14 did not participate in the post study. (subjects 2, 6, 8, 10, and 16). One subject experienced greater The results are further arranged by time of day. It appears delay after treatment installation (subject 3). Note that sub- that participants in the 8:30 a.m. time slot experienced lower jects 5 and 14 did not participate in the post study. delay than those participating later in the day, when traffic For the SS+FB crosswalk, Exhibit 8 shows that several volumes were higher. However, given the low sample of obser- participants experienced reductions in 85th percentile delay vations in each category, no effect can be isolated. Most par- after installation of the treatments (subjects 2, 4, 7, 8, 9, ticipants were in the 10:30 a.m. time slot, during which a range 10, 11, 12, 13, 15, and 16), while others stayed approximately of delay times was observed.

OCR for page 101
107 The analysis further investigated two parameters that let a pedestrian cross, unless they experience a downstream were intended to describe the efficiency with which crossing conflict and therefore do not incur any additional delay. At opportunities were utilized: the CTL, there may be a downstream queue resulting from vehicles yielding to cross-street traffic. At the same time, these Latency (s): Latency is defined as the time between when high YLTs prove that there are some determined yielders that the previous vehicle went through the crosswalk and the time disrupt traffic operations similar to the way a signalized cross- the pedestrian initiated the crossing. ing would, or even more so. For the 16-ft crossing, a pedestrian Yield Lost Time (s): The YLT is defined as the time between signal would likely be timed as 4 s of "Walk" followed by 5 s when a driver first yields and the time the crossing is initiated. of "Flashing Don't Walk" (16 ft/3.5 fps = 4.6 s), which is Note that in some cases, pedestrians may prefer to cross significantly less than the time some drivers yielded. Pre- only after a car has come to a full stop (stopped yield), and sumably, a PHB or HAWK phasing scheme would further so some inherent yield utilization time is expected. reduce the signal impact. Finally, the analysis evaluates the rate of O&M interven- Exhibit 9 shows statistics for both measures. tions, a measure of pedestrian risk during the crossings. The The latency results in Exhibit 9 suggest that on average study participants were at all times accompanied by a certi- pedestrians waited 8 to 10 s into a crossable gap before ini- fied O&M specialist, who was directed to stop the participants tiating the crossing in the pre study. This suggests a lot of if the crossing decision would have resulted in undue risk to inefficiency in decision-making and likely contributes to the pedestrian and/or driver. The resulting rate of O&M inter- low overall rate of gap utilization at the site. Individual sub- vention is defined as follows: jects even experienced average latency times up to 32 s. With installation of the treatments, the average latency dropped at Intervention Rate (%): The intervention rate is defined by the SS+FB crosswalk from 10.0 to 6.8 s (p = 0.0986). It appears the number of times the O&M specialist intervened for a that the combination of sound strips and flashing beacon either particular subject divided by the total number of lanes gave pedestrians more confidence in their actions or that the crossed for a particular condition. For example, one inter- sound strips helped with identifying gap crossing opportunities vention over a set of 20 lane crossings at one turn lane cor- (in this case through the absence of sound). responds to an intervention rate of 5%. For the yield lost time measure, there was no measurable difference between SS-ONLY and SS+FB corners and no The summary statistics for O&M interventions are given in detectable impact with treatment installation. The average YLT Exhibit 10. was in the range of 3 to 4 s, which again points to inefficiencies The results show that a total of 44 O&M interventions in the utilization of yields. Similar to observations at round- were observed across all 16 participants in the pre case, 30 at abouts, isolated YLTs reached a maximum average of 18.3 s. the SS-ONLY, and 14 at the SS+FB crosswalk. On average, It is expected that few drivers are willing to wait that long to each participant experienced 1.9 interventions at the SS_ONLY Exhibit 9. Latency and yield lost time statistics for CTL crosswalks. a) Latency (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 8.2 2.1 15.9 3.8 SS+FB (n = 16) 10.0 3.1 32.0 7.0 Post SS-ONLY (n = 14) 7.6 2.2 12.5 2.9 SS+FB (n = 14) 6.8 1.7 11.8 2.5 b) Yield Lost Time (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 3.6 0.9 13.7 3.7 SS+FB (n = 16) 3.4 5.1 18.3 5.1 Post SS-ONLY (n = 14) 4.1 0.5 13.0 3.9 SS+FB (n = 14) 3.8 1.1 9.6 2.3