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.
82 This chapter presents an interpretation of the findings of research performed under NCHRP Project 3-78A in the con- text of application to engineering practice. The chapter ini- tially presents a synthesis of the study approach, followed by a summary of the field studies performed during this project. It then discusses specific implications for facility design and accessibility considerations at the three tested facility types: channelized turn lanes, single-lane roundabouts, and two- lane roundabouts. The chapter concludes with a discussion of future research needs and recommendations for follow-up work to this project. Synthesis of Approach The results of NCHRP Project 3-78A have three principal components: (1) the formulation of a framework for describ- ing accessibility, (2) field study results of blind pedestrian cross- ing studies including treatment installation, and (3) approaches for extension of the findings to other locations. First, the project devised an analysis framework for assess- ing the accessibility and usability of different crossings to pedestrians who are blind. While the framework was developed in this project in light of specific challenges to this group of pedestrians, it has broader application to other pedestrian pop- ulations. The accessibility framework quantifies the crossing performance at a tested location in four dimensions: 1. Crossing opportunity criterion â Are there sufficient crossing opportunities in the form of yields or crossable gaps? 2. Opportunity utilization criterion â Are the crossing opportunities utilized by the pedestrian? 3. Delay criterion â Is a crossing opportunity taken within a reasonable time? 4. Safety criterion â Does the crossing interaction occur without a significant degree of risk? This framework allows the analyst to isolate different com- ponents of the crossing task at crosswalks. Any treatments intended to improve the accessibility are specifically targeted to one or more of the criteria. On the basis of these four analysis dimensions, the report describes the development and imple- mentation of a set of performance measures. The performance measures include the probability of encountering crossable gaps, P(CG_ENC), and yields, P(Y_ENC), and the probability of utilizing these crossing opportunities, P(GO|CG) and P(GO|Y). These performance measures were used in field studies to quantify the first two accessibility criteria. The per- formance measures are further hypothesized to affect criteria 3 and 4, the delay and safety experienced by the pedestrian during the crossing. Chapter 4 of this report provides detailed descriptions and definitions of all performance measures. The present research effort serves to introduce a structured and measurable framework for quantifying the principal operational parameters of accessibility. Second, the research conducted a total of 11 field studies at multiple test locations involving crossing trials of blind study participants. These included four studies at an intersection with CTLs, two of which were posttests after installing sound strips, lane delineators, and a pedestrian-actuated flashing beacon (one location only). Four studies were performed at two-lane roundabout crosswalks. Again, two were posttests after installing a raised crosswalk and a pedestrian hybrid beacon at the two studied crosswalks. The remaining three studies were performed at two single-lane roundabouts. One of the single-lane roundabouts was evaluated in a pretest and posttest design, but without treatment installation, to test for a learning effect in study participants with repeated exposure. A twelfth dataset, originally collected for a prior research effort, was also included in the analysis. This dataset was also for a single-lane roundabout. In total the field studies evalu- ated approximately 3,300 individual crossing attempts by 56 different blind study participants in 12 studies at five different sites. C H A P T E R 7 Interpretation and Application
83 In the studies, pedestrians who were blind repeatedly crossed the street at roundabouts and CTLs. The participants were at all times accompanied by a certified O&M special- ist who was trained to prevent collisions by intervening in the crossing attempt in the event that participants misjudged traffic patterns. The data these studies yielded were very com- plex. A multitude of factors were observed and recorded both in the field in real-time and in post-processing from video recordings. Details on the study methodology are given in Chapter 3. The data analysis involved formulation of a new analysis framework for quantitatively describing pedestrian accessibility that is described in Chapter 4. Detailed results are presented in Chapter 5 with supplemental results given in Appendix A. Third, the research presented two approaches for exten- sion of the analysis framework and the field study results to other locations and traffic patterns (Chapter 6). Using the pedestrian and driver behavioral attributes defined in the analysis framework, the team developed regression-based mixed-priority models to predict pedestrian delay at unsignal- ized crosswalks. The term âmixed-priorityâ refers to a cross- ing situation where some pedestrians cross in gaps between moving vehicles while others cross in front of yielding vehicles. The models are sensitive to the fact that there is variability in the rates of occurrence of gaps and yields as well as in the rates of utilizing these opportunities. Separate delay mod- els were developed for each of the three facility types, and guidance was provided on how the models may be applied to other sites. The extension piece further provided guid- ance for how microsimulation models can be used to analyze pedestrianâvehicle interaction at roundabouts and channel- ized turn lanes. The discussion demonstrated that the analy- sis framework is compatible with a simulation environment. Special attention was given to a sensitivity analysis of different pedestrian signalization options at single-lane and two-lane roundabouts. Modeling was not done with regard to the safety of pedes- trians who are blind. The direct measure of safety used in this research (O&M interventions) was not frequent enough to sup- port modeling. Nonetheless, interventions were numerous, representing times where there was a high likelihood of the blind pedestrian being injured or killed without evasive action on the part of a driver or the pedestrian. In the worst test con- dition (CTL pretest), the probability of intervention was 1 in 18. Viewing this statistic from the perspective of a blind traveler who commuted across that CTL, this level of risk is unaccept- able. Intervention rates were also high at other sites, including the two-lane roundabout at pretest and one of the single-lane roundabouts. Consequently, the delay modeling effort should only be interpreted as one piece of guidance in the broader question of accessibility of a site. Implications of Field Study Results There are many challenges inherent in translating accessi- bility research into policy guidance. Roundabouts and CTLs, along with the individuals who use them, are diverse. The facilities differ in terms of vehicle volume, level of ambient noise, driver culture, and physical design. The individuals who participated in the NCHRP Project 3-78A studies dif- fered in their propensity for risk taking, hearing ability, travel experience, and physical and cognitive abilities. It was a chal- lenge in this project to draw appropriate, general conclusions from the results from a program of research that was limited in comparison to the range of factors that might influence street crossing. The program was limited by resources and by the availability of adequate numbers of participants near roundabout and CTL facilities. Because of these factors, the team narrowed the focus by studying relatively skilled, func- tionally blind individuals and by evaluating only two treat- ments at roundabouts (PHB and raised crosswalk) and two at CTLs (sound strips and flashing beacon). Other limitations in the work included the fact that sample sizes were modest and participants were recruited rather than selected randomly from a population. Despite the limitations noted above, the team derived con- clusions from the studies where possible. Given that there are numerous explanations for the results obtained, the team also hypothesized about the possible explanations or interpreta- tions of the findings. The information that follows is organ- ized by type of facility (CTL, single-lane roundabout, and two-lane roundabout). In regard to the CTL, the studies showed high rates of inter- vention at pretest for both CTLs (9.4% and 5.6%, respec- tively). Since the pretests at the two locations were completed by the same participants on the same day, and no treatment was in place, it is likely that the differences observed in the two locations in regard to intervention rate are due to ambi- ent noise levels or vehicle speeds and volumes. Traffic counts were higher at the SS-ONLY installation, which was the CTL with the higher intervention rate. Regardless of the cause, intervention rates at both CTL locations were very high and represent an unacceptable level of risk for the participants. The pretest at the two CTLs further showed high average delay per subject at 26.2 s and 23.4 s, respectively, to cross a single lane of traffic. However, individual delays were much higher, and some trials even timed out after participants were not able to cross within 2 min. The posttest data revealed that both tested treatments yielded a significant reduction in intervention rates and reduced aver- age delay to some extent. The combination of sound strips and flashing beacons reduced the intervention rate and delay more than the sound strips alone, although the difference between the two posttest intervention rates is not statistically
significant (p = 0.3483). However, the team concluded that even the lower intervention rate (1.4%) was still unacceptably high and that additional treatments are needed at this CTL to achieve an acceptable level of crossing safety. The research team did not evaluate any treatments at single-lane roundabouts. This research conducted pretests at a single-lane roundabout in Charlotte, NC, and Golden, CO. Also included in this report are findings from a study con- ducted at a single-lane roundabout in Raleigh, NC. The team did not pursue treatment installation as originally planned at the Charlotte single-lane roundabout, partly because it was believed that the treatment would be unlikely to result in a measurable effect of pedestrian safety, based on an intervention assessment. The pretest intervention rate was low, at 0.8%, so there was not much room for improvement from a safety per- spective. It was therefore concluded that project resources would be better spent elsewhere. However, at this particular single-lane roundabout, partic- ipants experienced relatively long pedestrian delay, probably due to the very low yielding rate by drivers and the tendency of participants to wait for long gaps before initiating cross- ing rather than forcing yields. The fact that interventions were relatively low at the Charlotte location led the team to believe that acoustic conditions at this intersection facilitated the use of a gap detection strategy, that participants recog- nized this fact and used this strategy, and that the use of this strategy, while resulting in delay, yielded an acceptable level of risk. It is concluded that a treatment is not necessary at this roundabout under prevailing traffic conditions when crossed by blind pedestrians who report they cross streets independ- ently. It is unclear if and how behavior and performance measures at this site would change with higher traffic volumes and fewer (large) gaps. In regard to the Golden single-lane roundabout, there was not a statistically significant difference between the interven- tion rate at pretest (1.4% of crossings) and posttest (0.5%). Both studies further showed a relatively low average delay for participants of 11.9 s and 12.1 s, respectively. No treatment was installed at the single-lane roundabout, and the same par- ticipants participated in both rounds. The decision for two studies without treatment installation was motivated by the goal to test for a learning effect in participants with repeated (twice) exposure to the same site and the same study protocol. The same participants who crossed at this site also crossed at the studied two-lane roundabouts to allow for a direct within-participant comparison of a single-lane and a two-lane roundabout. The fact that no learning effect was evident at the single-lane roundabout from a safety or delay perspective gives confidence that any pretestâposttest difference at the two- lane roundabout can in fact be attributed to a treatment effect. The team concludes that the Golden single-lane roundabout did not pose extraordinary crossing challenges under prevail- ing traffic conditions, but some risk remains. Similar to the Golden site, it is unclear if and how behavior and performance measures would change under different traffic conditions. The Raleigh single-lane roundabout data was collected as part of an earlier research effort, and the data was reanalyzed using the analysis framework developed as part of this project. The intervention rate was the highest of all roundabouts stud- ied (although the rate was lower than the rate at the CTL site studied). The team considers the 3.9% intervention rate at this intersection to be unacceptable and suggests that a treat- ment is needed at this intersection to reduce crossing risk. It needs to be noted that this prior study actually included some participants with normal vision and that these sighted partic- ipants did not experience any interventions. It is unclear what factors contributed to the high rate of interventions for par- ticipants who were blind at this location. There was a rela- tively high proportion of yielding vehicles, and it may be that participants crossed in what they perceived as a gap but what was actually a yield that was not detected by the blind pedes- trian. Individuals may have initiated crossing at the same time a yielded vehicle began to move. In this scenario, the O&M specialist would likely have intervened. Anecdotal observations seem to support this conclusion. At the Golden two-lane roundabout, the intervention rate dropped from 2.8% and 2.4% of trials prior to installation of treatments (for the raised crosswalk and the pedestrian hybrid beacon, respectively) to zero after installation of treatments. The RCW treatment reduced average pedestrian delay from 17.0 s to 8.0 s. The PHB reduced delay from 16.0 s to 5.8 s on average for crossing two lanes of traffic. The team concludes that without treatment in place, pedestrians who are blind and who cross the street are exposed to an unacceptable level of risk. It is further concluded that the risk level appeared acceptable after either of the tested treatments was installed at this site and under prevailing traffic conditions. It is unclear if and how crossing performance would change with higher traffic volumes or at a site with different geometry. The team was surprised that the intervention rate for both treatments was zero, as it was anticipated that the raised cross- walk would not yield as great a risk reduction as the PHB. Field notes from a team observer indicate that there were eight risky multiple threat crossings at the raised crosswalk that did not actually result in interventions. No such events were recorded at the PHB. Additional research at other locations and with other individuals is necessary to determine whether there is in fact no difference in risk between these two treatments. To summarize, the key observations from the field studies were: ⢠That it is possible to apply the analysis framework to field studies and to describe pedestrian crossing performance in measurable terms consistent with the traffic engineerâs 84
understanding of general traffic performance. The analysis framework is sensitive to differences between sites and the impact of pedestrian crossing treatments. ⢠That a measurable concept of accessibility can be quanti- fied in terms of the availability and utilization of crossing opportunities in the form of gaps and yields and that these measures can be used to describe the performance measure of delay, although delay is not a measure that can, by itself, be used to measure accessibility to blind pedestrians. ⢠That significant variability across study participants was observed for all sites, highlighting individual differences in terms of travel skills, decision-making, and level of crossing comfort. Significant differences also exist across different test sites, including the three tested single-lane roundabouts. ⢠That some CTL crossings are more hazardous than single- lane or two-lane roundabouts for pedestrians who are blind, and that the tested CTL treatmentsâsound strips and pedestrian-actuated flashing beaconsâdid not effec- tively alleviate all accessibility concerns. The sound-strip treatment did show potential, and modified materials and installations should be explored further, in addition to other potential modifications at CTLs that reduce vehicle speed and increase yielding. A major concern is the ability of pedestrians to detect yielding vehicles in the high ambi- ent noise environment of most CTLs. Treatments such as sound strips, lane delineators, and pedestrian-actuated flashing beacons cannot be expected to provide adequate accessibility at two-lane CTLs. ⢠That the three single-lane roundabouts varied considerably in the availability and usability of crossing opportunities, delay, and risk for pedestrians. At least one tested single-lane roundabout (PS-RAL) showed a relatively high incidence of interventions (3.9% mean, 15.6% for one individual), and another (DAV-CLT) had relatively high pedestrian delays (25.3 s mean, 74.0 s maximum individual average). Therefore, while for one single-lane roundabout crossing and some blind pedestrians no treatments appear neces- sary under tested conditions, at least one high-risk and one high-delay crossing warrant further treatment evaluations. ⢠That both tested two-lane roundabout treatments, raised crosswalk and pedestrian hybrid beacon, appeared to sig- nificantly enhance the accessibility at the tested site. The improvements were evident in the form of significantly reduced pedestrian delays and no posttest interventions with either of the treatments. However, field observations by a trained observer noted more potentially risky events at the raised crosswalk and none at the PHB, suggesting that the latter treatment represents the safest (tested) cross- ing condition for blind pedestrians at this site. The present studies suggest that the accessibility of a site is a function of the conflicting traffic volume, the speed of traffic, the ambient noise, the level of driver courtesy toward pedes- trians, and probably other variables yet to be identified and studied. A range of treatments is available and discussed in Chapter 2 that have the potential of influencing one of more of the dimensions of accessibility defined in the framework outlined in Chapter 4. Policy Implications While the U.S. Access Board draft PROWAG specifies a pedestrian-actuated signal at two-lane roundabout crosswalks with pedestrian facilities, the ADA allows equivalent facilita- tion in all implementations of requirements. Consequently, other treatments that provide equivalent accessibility are acceptable. This is to allow for improvements in technology, developments in materials or research, or the implementa- tion of new ideas and information. It is up to the designer and/or constructing jurisdiction to provide justification for installation decisions in the case of an ADA complaint. The team believes that there is some confusion in the interpreta- tion of these standards in that some may fail to recognize the inherent difference in civil rights laws and engineering stan- dards. While the current draft requirements focus on two- lane crossings at roundabouts and CTLs, as well as treatments that provide information about the crosswalk location such as landscaping or barriers, there is still a responsibility to design and build all facilities to be âaccessible to and usable byâ pedes- trians with disabilities (DOJ 1990). An increasing national debate in this area is very positive. However, a narrow focus in that debate on the signalization of two-lane crossings is associated with two major concerns: (1) by focusing on two-lane roundabouts, the accessibility of single-lane approaches is being largely ignored, and (2) the emphasis on signalization gives a perceived blanket obliga- tion for a one-size-fits-all treatment at all two-lane locations, regardless of the site specific geometry, traffic volumes, and driver behavioral patterns. There is no real debate over the fact that well-designed modern roundabouts are generally safer for vehicular traffic than many traditional intersections they replace. There is fur- ther no real debate that many sighted pedestrians can safely negotiate single-lane roundabouts and most two-lane round- abouts. However, research to date is unclear about the ability of other pedestrian populations, including elderly pedestri- ans, children, and those with mobility impairments, to cross safely at these locations. Further, it is unclear what challenges are posed by newly emerging three-lane roundabout designs with potentially high design speeds and high traffic volumes. The data presented in the present work strongly argues against the belief that all single-lane roundabouts are created equally. While one of the studied sites showed generally low delay and risk, a second site had high pedestrian risk, while 85
the third exhibited high delays. There was high inter-participant variability that emphasizes differences among individual blind travelers. Another important finding was the difficulty and level of risk encountered by blind pedestrians when attempting to cross a relatively high-speed, high-volume CTL. A sound- strip treatment did not prove to be effective in reducing risk, partly due to the high noise levels and difficulty of auditory detection in noise, and partly due to the lower sound output of the sound strips when vehicles were moving slowly over them. A pedestrian-actuated beacon with an audible message improved yielding behavior somewhat, but the posttest cross- ing performance was still associated with a high rate of O&M interventions. Consequently, the accessibility of this CTL site was not established with the treatments tested. Attention needs to be given to this access issue in addition to the atten- tion to accessibility of roundabout crossings. In general, these studies have shown that the tested treat- ments can, in fact, change the behavior of drivers as well as pedestrians and that these changes can be measured and quan- tified. The treatments differed in their effect on drivers and on pedestrians who are blind, and represent various degrees of installation cost and impact to the driver and pedestrian populations. The results and conclusions previously discussed should not be construed as absolute, and readers should remain cautious about basing policy decisions on these limited data. This report provides a firm conceptual measurement- driven approach to the study of the effect of such treatments, but it is clear that more field research is needed to explore and substantiate treatment effects. Discussion In the following, the team offers some additional discus- sion to highlight various aspects of the field study results. The section will highlight specific aspects of the study results that are important in guiding the decision-maker in evaluating the level of accessibility of a given crosswalk to pedestrians who are blind and who are relatively experienced travelers. Single-Lane Versus Two-Lane Approaches A pedestrian crossing with two conflicting lanes is gener- ally more challenging than one with a single lane because the vehicle state in both lanes affects the decision-making. Visual obstruction and auditory masking of vehicles in the near lane may block activity in the far lane, which may result in multiple-threat situations (Zegeer et al. 2002). The experiments at the Golden two-lane roundabout showed that pedestrians who are blind are often capable of utilizing dual crossing opportunities at this location, characterized by a yield or crossable gap in both lanes. However, the analysis showed that the rate of occurrence of these types of opportuni- ties can be low, resulting in high pedestrian delays. The occur- rence of dual crossing opportunities is expected to be even less at approaches with higher conflicting volumes and speeds. In addition to delay, participants experienced a significant amount of risk, as indicated by the frequency of O&M inter- ventions. Both delay and risk were reduced with the installation of each of the two tested treatments (PHB and RCW), thereby improving the accessibility and usability of the site. Before treat- ment installation, the same participants experienced higher delay and greater risk at the two-lane crossings compared to the single-lane roundabout. After treatment installation, the crossing performance at the two-lane approaches improved to levels comparable to the tested single-lane roundabout crossings without treatments. However, some safety concerns remained at the RCW, evident by other risky events (that did not result in interventions) that were noted by a trained observer recording data during the studies. While some of the tested single-lane approaches seemed more easily crossable than an untreated two-lane approach, some exceptions need to be highlighted. In particular, the high- speed, high-volume, single-lane CTL crossings resulted in very high delays and the highest risk for any of the test sites. Also, the Raleigh roundabout had high intervention rates despite lower speeds and generally courteous driver behavior. Con- sequently, even a single-lane crossing can be challenging and potentially dangerous to cross if traffic volumes, vehicle speeds, and driver behavior are not conducive to crossings by pedes- trians who are blind. The Impact of Vehicle Speed High vehicle speeds have been linked to a decreased likeli- hood of driver yielding (Geruschat and Hassan 2005) and are further associated with an increased pedestrian injury rate when collisions occur. In the context of pedestrian crossings at roundabouts and CTLs, a âhigh speedâ is categorized as a design speed or average observed speed at the crosswalk greater than 20 mph. In this research, the high-speed channelized right turn lanes resulted in the greatest pedestrian delay and risk, sup- porting the hypothesis that pedestrian accessibility is tied to vehicle speeds. The treatment effect of the RCW at the two-lane roundabout provides further evidence for this. The RCW sig- nificantly reduced pedestrian delay and reduced risk while not being associated with any form of red signal display like the pedestrian hybrid beacon. After RCW treatment installa- tion, only a small percentage of drivers passed in front of the waiting pedestrian without yielding at the tested location. The findings on vehicle speed have two primary implications for practitioners: 86
87 First, while the tested single-lane roundabouts were found to be manageable by most pedestrians who are blind, it is expected that a higher-speed design will have a severe adverse effect on accessibility. Following the same argument, larger two-lane roundabouts with higher design speeds and two or more lanes can be expected to be less accessible than the tested two-lane roundabout in Golden. Second, if an existing site features high speeds through a combination of roadway geometry and driver behavior, it is hypothesized that a higher-impact treatment would be needed to make the site accessible. Low-impact treatments like the ones tested at the high-speed CTL location are apparently not sufficient to make a high-speed crossing accessible. It is hypoth- esized that a traffic calming treatment like the raised crosswalk or a red signal display would be more appropriate at high- speed locations. However, more research on these treatments is necessary to solidify this claim, mainly due to the observed multiple threat incidents at the RCW in the posttest. Crossing Geometry In all crossings observed in this research, vehicles approached from only one direction at a time. Both roundabouts and channelized turn lanes feature splitter islands that provide the benefit of refuge for pedestrians before they complete the second part of the crossing. It is important to emphasize that this form of traffic pattern is a potential benefit of both types of intersection compared to intersections with one-stage crossings across two-way traffic. Despite the advantages of one-way approaching traffic, identifying gaps and yields with curved vehicle trajectories and elevated levels of ambient and vehicular noise contribute to crossing challenges at these types of crosswalks for pedestri- ans who are blind. In particular, prior research (e.g., Ashmead et al. 2005, Guth et al. 2005) has noted that the exit portion of a roundabout crossing is more difficult to cross due to diffi- culties discerning exiting traffic from circulating traffic and also due to less frequent yielding behavior. From a driver behavioral perspective, pedestrians may be considered more of a nuisance when exiting the roundabout. Upon entering a roundabout, vehicle trajectories are characterized by deceleration (to be able to safely navigate the circle) and the potential to have to come to a stop (to yield to traffic in the circulating lane). On the con- trary, the exiting driver accelerates and psychologically has cleared the delay-causing intersection. In this research, the hypothesized difference between enter- ing and exiting traffic was evident in a lower likelihood of driver yielding at the exit. Interestingly, this difference did not result in a consistent and notable increase in delay or safety at the exit portion of the crosswalk across sites. Another aspect of crossing geometry is the relative place- ment of the crosswalk in relation to the roundabout or the channelized turn lane. At both types of facilities, crosswalks are typically located at least one vehicle length away from the circulating (downstream) lane to ensure (1) storage for one vehicle downstream of the crosswalk and (2) a separation between the driverâs cognitive tasks of interacting with pedes- trians and the circulating traffic (or downstream CTL traffic). However, this placement of the crosswalk often results in vehicle queues across the crosswalk. Some interventions (espe- cially at the CTL) occurred in response to vehicles in queues that began moving as the blind pedestrian approached the side of the vehicle, not realizing it was there. While this research did not perform any field studies on alternative crosswalk locations, a simulation-based analysis of roundabout signals demonstrated advantages in vehicular delay of the zigzag or distal crosswalk location. The greater exit-lane separation enhanced these two objectives (more storage and better sepa- ration of cognitive tasks for the driver), but is associated with two tradeoffs that may negatively affect pedestrians: (1) with increasing separation, the speed-reducing effect of the round- about is reduced, and (2) the pedestrian travel distance is increased if measured from the main intersection. However, it is unclear at this point whether an alternate crosswalk ge- ometry has any (positive or negative) impact on the crossing ability of pedestrians who are blind. Inter-Participant Variability The analysis results showed a very large variability between different blind participants. It needs to be emphasized that at all of the crossings, there were some individuals who were able to cross with a reasonable amount of delay and without expe- riencing any interventions. Others experienced multiple inter- ventions at the same crossing, and others had some trials that were terminated because a 2-min time-out limit was reached. Appendix A provides more detail on the inter-participant vari- ability and shows the distribution of key performance measures across participants. This variability causes challenges for the analysis, the interpretation of the results, and ultimately the decision to consider a site accessible with or without treatment. The results in this report emphasize average performance and associated variability (i.e., standard deviation) when eval- uating pretestâposttest treatment effects. Consequently, any differences that are statistically significant are to be considered noteworthy in light of high standard deviations. For example, the two-lane roundabout treatments seemed to have improved crossing effectiveness and safety for all participants to the point of reducing delay and eliminating interventions (at the given sample size). On the other hand, the CTL treatments did not help all participants and left many with significant challenges in crossing at the location. The inter-participant variability also means that sites that may be accessible for some travelers pose severe challenges for others. For example, some participants
readily crossed at the DAV-CLT single-lane roundabout while others experienced over 80 s of average delay. Also coming into play was the situation behind the lead vehicle. If there was no vehicle following the lead vehicle so closely that there was a high chance of a collision, the O&M specialist often did not intervene. Participants apparently were often unaware that a vehicle was approaching. Also, many participants crossed when a vehicle had yielded for them but they did not realize that a vehicle had yielded until it accelerated across the crosswalk after they had crossed. In short, while the frequency of interventions was relatively low, conversations with participants after trials and after the study led the team to believe that the participants often were unaware of the situation when they crossed the street. Some variability in performance may have been due to par- ticipant behavior. The team presumes that some behaviors result in lower risk crossings than others, but the team did not manipulate participant behavior and did not explicitly study it. Some participants stood very upright and still while listening and waiting to cross, with their canes held vertically in front of their bodies, while others leaned forward as they anticipated crossing or took a step forward and extended their canes. Some turned their heads toward the traffic (and may have been per- ceived as looking at traffic) and others looked down while listening. These various postures appeared to the research team to affect yielding behavior and crossing behavior. A lack of training on roundabout crossings also may have allowed participants to use less than optimal techniques. The same general points made above for roundabouts appear to be true for CTLs. Many of the blind participants did not understand the layout of the CTL before the familiariza- tion provided as part of the study. For example, they were unaware of the potential crosswalk locations, the shape of the island, and the addition of a deceleration lane at the intersec- tion, and many were confused about the fact that the signal did not also control the right turn lane. Pedestrians who are blind may also benefit from instruction that helps them better understand the geometry and traffic flow at CTLs and various crossing strategies. It is hypothesized that even with intense and repeated O&M training, many of the crossing challenges at busy and high-speed crossings cannot be alleviated. Accord- ingly, additional infrastructure-based treatments as tested in this research are needed to satisfy concerns of pedestrian accessibility. Learning Effect and O&M Training The inter-participant variability raises the question of whether individual travel skills can be improved through train- ing and education, to the point where they represent a viable treatment to enhancing accessibility. This form of training is the basis of the field of orientation and mobility, which goes well beyond the scope of this report. All participants in this research were familiarized with the test location by walking the perimeter of the site, by exploring the geometry through use of a tactile map, and by performing some (assisted) practice crossings prior to starting the actual experiment. During orientation to the study, participants did not receive any instruction in making crossing decisions. This lack of instruction from the O&M specialist was part of the research protocol, and participants were told that instruction or feed- back would not be provided. However, it may have resulted in some participants assuming that they were making good decisions about crossings if interventions did not occur, when in fact the O&M specialist recognized that the situation was potentially risky but did not intervene because she knew that the approaching vehicle could (and was) taking evasive action (usually braking or accelerating to pass the crosswalk before the pedestrian arrived in the case of the two-lane roundabout). Also, some participants stepped out at times that forced a yield, apparently without realizing they were doing so. In this situation the O&M specialist did not intervene if she was con- fident that the vehicle was yielding and would stop in time, even in situations of relatively fast vehicle deceleration. In this research, the single-lane roundabout in Golden pro- vides insights into the effect of added exposure or experience on crossing. Since no treatments were installed at the site, the pretestâposttest comparison represents a control for any learn- ing effect that may have affected the concurrent Golden two- lane roundabout study. Since no such effect was measured, two conclusions can be drawn. First, the pretestâposttest improve- ments in crossing performance at the two-lane roundabout are likely attributable to a treatment effect and not a learning effect. Second, the repeated exposure (twice) of participants to the same single-lane roundabout did not significantly improve their crossing performance. While anecdotal evidence suggests that participants were more comfortable in the second round of exposure (the post- test), their earlier experience at the roundabout was not O&M training or instruction, which might be beneficial, particularly at single-lane roundabouts. Itâs possible that O&M instruction may have a more signifi- cant and permanent effect. There is clearly a need for education of pedestrians who are blind or visually impaired about round- abouts, including how to make safe crossings. Just as little is known about the most appropriate engineering treatments for promoting the accessibility of roundabouts, little is known about the most appropriate crossing strategies for blind pedes- trians. As illustrated by this project, the engineering profes- sion and the O&M profession must work together to promote accessibility. It is therefore essential that engineering profes- sionals have a general understanding of the O&M profession, and vice versa. What follows are some general principles of O&M instruction to help promote this mutual understanding. 88
First, it is important to understand that people who are blind rarely receive training for every route that they travel or at every intersection they cross. O&M instruction is similar to driver education; people may receive individualized on-the-street training, but they typically receive such extensive training just once and then generalize to new areas and update their skills as they travel and encounter new situations. For individuals who have been blind for many years, formal O&M instruction may have been received decades ago. Depending on an individualâs travel experience and capacity for self-instruction, he or she will be more or less knowledgeable about developments in traf- fic engineering and their implications for non-visual travel. Just as there is often a need for an educational campaign for drivers about how to manage roundabouts when roundabouts are new in an area, there may be a need for an educational cam- paign for pedestrians who are blind or visually impaired when roundabouts are installed. Second, O&M instruction is usually not amenable to class- room instruction, written material, or website content. Most O&M instruction is provided âon the streetâ to ensure that a blind pedestrian can experience and practice a new set of skills. At roundabouts, this would involve actual experience with roundabout layout and crossings, along with instruction by a certified O&M specialist. In addition to providing the initial instruction and structured experience, the O&M spe- cialist is present to reduce risk (including intervening) and to provide feedback during the initial stages of learning. The initial instruction may involve the use of tactile (raised line) maps of the roundabout. But it is essential to (eventually) experience roundaboutsâfor example, by walking around and through them and listening to traffic. Ideally, as with signalized intersections, this instruction would begin with relatively simple roundabouts and progress to more com- plex ones. Most of the general concepts to be mastered by the blind pedestrians are the same as those for sighted pedes- trians, including, for example, intersection and crosswalk geometry, traffic movements, and crossing strategies. Other concepts have greater relevance to blind pedestrians, such as the 3 to 4 s of sound masking created by a vehicle that has just passed the crosswalk in front of (or in back of) a pedestrian. Knowledge of these concepts would then be followed up with practice, feedback, and more practice. As noted above, the most appropriate treatments and strategies for non-visual roundabout crossing remain an open question, and this proj- ect has been a start toward developing such treatments and strategies. As a final caveat on this topic, it is important to note that this project has focused on pedestrians who are blind, not pedestrians with low vision. Related research (e.g., Geruschat et al. 2006; Hassan, Geruschat, and Turano 2005) has identi- fied challenges experienced by pedestrians with low vision that are not experienced by pedestrians with normal vision. Confidence in Decision-Making In Chapter 4, the term âutilizationâ was used to describe a pedestrian crossing in a yield or gap. This term objectively describes observable behavior rather than making presump- tions about the pedestrianâs ability to detect a yield or crossable gap. In fact, through the experimental trials in this research it became evident that some blind travelers may well be aware of the presence of, for example, a yielding vehicle (i.e., detec- tion), but may be reluctant for one reason or another to utilize the crossing opportunity. The individualâs rationale for non- utilization may be tied to past experience (a driver who yielded but then changed his or her mind), uncertainty (about whether the vehicle is yielding to the pedestrian or for another reason), or ultimately a lack of confidence in the viability and accuracy of the detection. Sighted pedestrians have the advantage over blind pedestri- ans in being able to judge gaps and detect yielding vehicles using both visual and auditory information. Lack of access to visual information places the blind pedestrian at a distinct disadvantage. The distance at which single vehicles can be detected is shorter, perception of direction from which single vehicles are coming is less precise, perception of vehicle trajec- tory is very difficult on curving paths, perception of rate of approach (hence time-to-contact) is less precise, the sound of a vehicle that has just passed the crosswalk can mask the sound of an approaching vehicle, and it is often difficult to detect the presence of a yielding vehicle. The presence of multiple vehi- cles exacerbates all these difficulties. As a consequence of these difficulties, pedestrians who are blind typically delay crossing longer than pedestrians who are not. This may result in vehic- ular queuing. It may also result in a yielding driver accelerat- ing at precisely the time the pedestrian who is blind decides to begin to cross. This type of behavior was observed during this research. Sighted pedestrians also have the advantage in that they can attempt to force yields given the ability to estab- lish non-verbal (visual) communication with the approach- ing driver(s). Blind pedestrians lack that advantage. From observations, the confidence of pedestrians to make crossing decisions is related to their personal travel skills, expe- rience, and willingness to accept risk. It may further be (par- tially) influenced by the opportunities presented, where the crossing strategy is a function of traffic patterns during the study. Under low-volume conditions, participants can reason- ably expect that a large gap is likely to occur and may therefore be more inclined to reject shorter gaps or yields. It is likely that varying degrees of confidence contributed to the observed inter-participant variability described above. In other words, the difference between participants may be a combination of skill (e.g., hearing ability, cognitive ability) and confidence. It is presumed that confidence in decision-making may be improved with repeated exposure and training. 89
However, the analysis also suggests that pedestrian confi- dence was increased with some of the crossing treatments, even where an effect was not hypothesized. Most notably, the installation of the raised crosswalk and flashing beacon treat- ments resulted in an increase in the rates of opportunity uti- lization, P(GO|Y) and P(GO|CG). At the same time, the two-lane roundabout RCW treatment also contributed to an increase in the number of forced yields, defined as the pedes- trian stepping into the roadway before a yield was initiated by the driver. A similar increase in the willingness of sighted pedestrians to accept risk was observed in a field study of two different pedestrian crossing treatments (Schroeder 2008). In observational before-and-after studies with in-road pedestrian warning signs and in-pavement pedestrian-actuated flashing beacons, the author noted a decrease in pedestrian critical gap after treatment installation. It was hypothesized that pedes- trians felt âempoweredâ or at least more noticeable with the treatment in place, which effectively lowered their perceived risk threshold. These trends raise the question of whether the apparent increase in confidence is desirable. Clearly, more forced yields would be a concern if the driver had not paid attention. On the other hand, even with higher opportunity utilization and more forced yields, the rate of interventions decreased. Self-Assessment of Risk and Travel Skills The apparent increase in confidence is also reflected in the post-experiment participant questionnaires, where participants responded positively to a question regarding their confidence in crossing safely, both in the pretest and posttest, and to the treatments. Across all settings, participantsâ ratings of confi- dence in crossing safely, both before and after treatments, were quite high (a range of 3.66 to 4.83 on a scale of 1, low confi- dence to 5, high confidence). However, the range of interven- tions across all sites and with or without treatment was also large, from 0.5% to 9.4%. This raises questions about the rela- tionship between (stated) confidence in decision-making and the quality of these decisions. It is quite likely that participants often failed to recognize when they made risky crossings. On many crossings, partici- pants may have been unaware of the extraordinary avoidance maneuvers of drivers, or of how close they came to a crash. The O&M specialist gave participants no feedback regarding crossings, many of which were quite risky but nonetheless did not result in interventions because the O&M specialist could see that drivers were taking avoidance action. There is a large body of research indicating that people tend to underestimate the likelihood that they will be involved in an adverse event, reviewed in Greening and Chandler (1997). In the typical par- adigm for this research, participants are told the average rate of occurrence of a given adverse event, such as a crash, and are then asked to estimate the likelihood of their being involved in a crash. Especially when participants have control over behavior that could result in an adverse event (as they did in the present research), they are likely to underestimate their risk of being involved in the adverse event. They believe that they are better than average at the skills necessary to avoid the adverse event. The high confidence ratings in the present research, despite the relatively high risk, could be an example of this phenomenon that has been documented for individu- als having unimpaired vision. Perhaps even more to the point, estimates of cumulative risk have been found to be especially skewed toward over- estimation of safety (Knauper et al. 2005). Thus an individual who actually required experimenter intervention on only one of 20 crossings may nonetheless have decided during subjec- tive questions that he or she would feel comfortable making the crossing in the future because there may have been no realization that that was a very high rateâan average likeli- hood of one risky crossing every 2 weeks if the individual commuted back and forth across the crossing daily. Research by Ayers et al. (1998) provides further insight into why con- fidence in crossing safety was much higher than warranted by the number of interventions. These researchers found that risk perception may not play as dominant a role in common behavioral choice as perception of whether an action can be completed successfully. Having completed a number of appar- ently successful crossings, participants may simply have ignored interventions or other indications of their risky decisions and based their ratings of crossing safety simply on the fact that they had made a number of safe crossings. To understand the implications of the present research, in which most participants felt quite confident that they could safely cross at roundabouts and channelized turn lanes despite average performance data to the contrary, one must recognize that overestimation of safety is a common occur- rence. Therefore the objective measure of safetyâthat is, interventionsâis the most reliable (available) indication of crossing safety at roundabouts and channelized turn lanes. One should not be misled into thinking that crossings at these locations are safe because participants perceived them to be. Viability of Interventions as a Measure of Risk The accessibility framework applied in this research ulti- mately focuses on the concepts of pedestrian delay and risk. Pedestrian delay is reliably and objectively measured. The mea- surement of pedestrian risk is more challenging. The tradition- ally used objective measure of risk in the transportation field is the frequency and rate of collisions. Clearly, crash statistics are not applicable given the experimental design used, and conse- quently other surrogate safety measures need to be identified. 90
The selected safety measure of O&M interventions had been used in several prior studies involving crossings of blind pedestrians. The measure is related to the concept of traffic conflicts, which is an increasingly common safety perfor- mance measure. A traffic conflict is typically defined as an interaction between two vehicles (or a vehicle and a pedes- trian) for which a collision is imminent pending an evasive action by any of the involved parties. The categorization for what constitutes a risky time-to-collision is typically defined by a 2-s threshold. The approach of using O&M interventions as a surrogate safety measure uses a similar underlying principle. The O&M specialist intervenes when he or she feels that the crossing deci- sion by the pedestrian would result in undue risk. For many of the interventions, it is likely that an actual collision could have been avoided if the driver had performed an emergency brak- ing maneuver. However, this level of risk was deemed un- acceptable to the pedestrian as well as to the approaching driver and any following vehicles. Clearly, the intervention measure is subject to variability among different O&M specialists. It is human nature to have different perceptions of risk, and consequently it is expected that the threshold for an intervention may differ somewhat between two O&M experts. In this research, the team attempted to minimize this variability by having the same O&M specialist perform all the experimental trials. The O&M intervention is further a coarse measure in that it stratifies pedestrian risk on a binary (yes/no) scale. In real- ity, the O&M interventions included a range of situations that represent varying degrees of risk that are difficult to define. For example, an intervention may include a multiple-threat situation at a two-lane approach where a pedestrian utilizes a near-lane yield by a heavy vehicle that masks visual and audi- tory information about a fast-moving vehicle in the far lane. On the other hand, an intervention may have occurred if the driver of a previously stopped vehicle accelerated at the same time that the pedestrian stepped out to utilize the yield. These events occurred in particular at the CTL, where traffic queues frequently spilled back across the crosswalk from the down- stream merge point. The intervention measure does not give a comprehensive safety assessment of the crossing attempts. Similar to collisions, interventions and conflicts are rare events that depend on a suf- ficiently large sample size. The fact that all of the sites exhibited some level of interventions (despite the relatively low sample size) speaks to their risk to the blind travelers. While no sighted pedestrians were involved in these experiments, it is hypothe- sized that a sighted participant would have been very unlikely to experience an O&M intervention rate at, for example, 16 crossing attempts at a single-lane roundabout, which is sup- ported by research comparing blind and sighted pedestrians at roundabouts (NIH 2010). The team attempted to develop a secondary measure of pedestrian safety by using expert ratings of video clips recorded during the trials. The approach aimed to extract potentially risky video clips that included, but were not limited to, all inter- ventions recorded at a site. These short video clips were then shown in randomized order to a panel of expert reviewers, who rated pedestrian safety in the clip on a scale of 1 to 5. This approach was tested for the pretest condition at the CTL, but did not result in a reliable alternative to the intervention measure. The exercise did establish that there is agreement within most of the reviewers that events that resulted in O&M interventions were in fact risky. It further established that most of the control clips that were added to the sample to establish a baseline for safe behavior resulted in mostly low- risk ratings. However, the exercise proved inconclusive as far as any borderline risky events are concerned. Overall, the activity gave the team confidence in the via- bility of the O&M intervention measure for pedestrian risk but did not result in any further information about safety performance. Due to project constraints, this approach was not applied to the remaining test sites. The team does believe, however, that there is merit in continuing research in this area. In particular, it would be beneficial to develop an objec- tive performance of safety that is based on the field-measured time-to-collision based on vehicle trajectories (rather than on expert judgment). The results of the safety rating exercise are presented in Appendix H to be a resource for further research activities. Driver Yielding and Enforcement The treatments tested as part of NCHRP Project 3-78A all represented infrastructure-based treatments, which was a principal requirement of the project objectives and in keeping with Access Board guidelines. The analysis therefore did not include any treatments that would be carried by the pedestrian in the form of wayfinding or traffic detection technology. The scope further did not include any policy-based treatments such as increased law enforcement of yielding laws. Nonetheless, the results do seem to suggest that vehicle yields represent valid crossing opportunities that may be utilized by pedestrians who are blind, although at utilization rates less than 100%. Infrastructure treatments that resulted in increased yielding behavior and reduced vehicle speeds were shown to improve accessibility. The motor vehicle codes in most U.S. states require driv- ers to yield to pedestrians in the crosswalk and in some cases even to pedestrians waiting at the crosswalk. However, field- observed yielding rates vary widely and are inconsistent across locations even with the same crossing treatment installed (Fitzpatrick et al. 2006). Most states offer additional protection to blind travelers to what is commonly referred to as âwhite cane laws.â These are intended to promote driver awareness, 91
courtesy, and yielding to pedestrians who identify themselves as having vision impairment, by either carrying a long white cane or by walking with a guide dog. The net benefit of increased driver yielding behavior is hypothesized to be greatest when traffic volumes are high since the occurrence of crossable gaps and all-quiet periods decreases. The fact that high yield com- pliance can result in low delays even under high volumes is made evident through the data from the PS-RAL single-lane roundabout, although it may also be associated with a safety trade-off. Despite a high AADT and heavy peak-hour flows, pedestrian delays were less than at the low-volume DAV-CLT roundabout. The difference is most likely attributable to a higher willingness of drivers to yield at the PS-RAL site. In this case, however, the PS-RAL site remained inaccessible to pedestrians due to a high rate of interventions. In evaluating yielding behavior, two other considerations are important. First, the length of time a driver is willing to yield (i.e., driver patience) likely affects the ability of the pedes- trian to utilize that yield. The detailed study results shown in Appendix A show that most participants required several seconds before crossing in front of a yielding vehicle, and some waited 10 s or more. Therefore, a yield that only lasts for a few seconds is unlikely to result in a utilized crossing opportunity for blind pedestrians. Second, the physical location of yields relative to the crosswalk is believed to affect the pedestriansâ ability to detect these events. It is believed that the observed posttest increase in yield utilization at the two-lane round- about for both treatments is at least partly attributed to the fact that drivers tended to yield closer to the crosswalk and there- fore the rate of deceleration was more rapid. In other words, it seemed easier to detect a vehicle that quickly decelerated to a stop close to the crosswalk than one that slowly coasted to a stop farther from the crosswalk. Interestingly, two data collection sites with different traffic conditions were located in Charlotte. At the tested single-lane roundabout, volumes were low and participants were delayed despite ample gap crossing opportunities as many waited for all-quiet periods. At the two CTL crossings, traffic volumes were much higher and the delay was exacerbated by a very high level of ambient noise from the main intersection. All three crossing locations were characterized by a low propen- sity of drivers to yield, which may be characteristic of the local driving culture or may be coincidence. Implications for Facility Design This section on implications for facility design is intended to advance the discussion in this area and to direct future research. The points below are based on field results from this and other studies as well as anecdotal evidence that evolved from observ- ing many crossing attempts by blind travelers at different loca- tions. Taken together, they comprise a toolbox for increasing the accessibility of crossings at CTLs, single-lane roundabouts, and two-lane roundabouts for pedestrians who are blind. Channelized Turn Lanes There is anecdotal evidence that a crosswalk located in the middle of the turn lane is preferable to a crosswalk at the upstream or downstream portion of the turn lane. The middle crosswalk establishes a short crossing path roughly perpendi- cular to the trajectory of turning vehicles (and therefore is useful for establishing alignment), and it physically separates the conflict of turning drivers and pedestrians with the down- stream merge point. Based on turning radii and associated design speeds, this is also likely to be the location where speeds of right-turning vehicles are lowest. Since no alternate cross- walk locations were tested at the CTL (i.e., upstream or down- stream), it is unclear from this research whether a different crosswalk location would have any (positive or negative) impact on the crossing ability of blind pedestrians. Field tests suggested that high vehicle speeds contributed to the high incidence of unsafe crossings at the tested CTL; therefore, geometric designs and treatments intended to reduce vehicular speed, such as traffic-calming designs, raised cross- walks, pork-chop island design, narrow CTL width, small curve radii, and the absence of an acceleration lane may decrease the likelihood of unsafe crossing judgment by pedestrians who are blind. However, none of these treatments were tested in this research. Support for low turning-vehicle design speeds for CTLs is also given in the AASHTO Guide for the Planning, Design, and Operation of Pedestrian Facilities (2004). CTL designs with raised crosswalk have been observed in several cities across the United States. No CTLs with acceleration lanes were studied in this research, but based on anecdotal evidence it is expected that they may further increase speeds and may result in low compliance with crosswalk laws and installed treatments. Dual-lane CTLs are included in the draft PROWAG as facilities where a pedestrian-actuated and APS-equipped signal meets the accessibility requirements. While no dual-lane CTLs were part of this research, the crossing challenges observed at the tested single-lane CTL suggest that dual-lane CTLs may in fact be very challenging places to cross for a pedestrian who is blind. With two lanes, the expected challenges are related to higher volumes, higher speeds, and a risk of multiple-threat situations, compared to single-lane CTLs. Detectable warnings complying with the draft PROWAG are required at both curb and island ends of crosswalks to warn pedestrians who are blind that they are leaving the pedestrian way and entering the vehicular way. A potential treatment that facilitates the auditory discrim- ination of right-turning (conflicting) and through traffic should be considered and studied further. The sound strips 92
tested in this research generally served their intended pur- pose; however, the selected type of material evaluated proved to be too quiet given the high ambient noise at the test site. The biggest problems with the treatment arose when vehicles were traveling very slowly and the audible cues from the sound strips were not noticeable. The use of a pedestrian signal at the CTL is a possible treat- ment that can be tied in with the existing signal control at the main intersection. There are some challenges to tying the CTL signal into the existing controller, especially when an intersec- tion has multiple CTLs that are to be signalized. But an exist- ing signal phasing strategy is to use vehicle overlap phasing with a pedestrian signal across the crosswalk. If adding the CTL signal to an intersection with a long cycle, the expected pedes- trian delay may be high and needs to be assessed in the context of the total pedestrian crossing. At a busy intersection such as the test site, a diagonal pedestrian crossing (e.g., from the southeast to northwest corner) would entail the use of four sig- nalized crossings (CTL, main road, side road, CTL). Signalized CTLs have been observed in several cities across the United States and are rather common in other countries. In addition to a standard pedestrian-actuated signal, a PHB may be another alternative for CTLs. The advantages of the allowable vehicle movements during the âFlashing Redâ phase are reduced with shorter crossings and associated shorter âFlashing Donât Walkâ phases. However, some vehicle delay savings are expected to remain. The allowable provision in the MUTCD to let the pedestrian display at the PHB rest in a dark mode may be considered for CTLs since many (sighted) pedestrians may not require the added assistance provided by the signal. Single-Lane Roundabouts The design of single-lane roundabouts should encourage low vehicle speeds in the vicinity of the crosswalk. Low speeds are shown to correlate with increased yielding behavior and reduced injury in case of a collision. There is some concern that lower speeds (and associated lower vehicle noise) may reduce the ability of a blind pedestrian to detect crossing opportunities (yields), which is a question that deserves the attention of researchers. The design of a single-lane roundabout should encourage narrow (or standard) lane widths in the vicinity of the cross- walk. Lanes that flare out too early unnecessarily increase the crossing distance for pedestrians and further may allow vehi- cle passing in the vicinity of the crosswalk, thereby creating a potential multiple-threat situation. This was observed at one of the three tested single-lane roundabouts and resulted in some near-interventions. Detectable warnings complying with the draft PROWAG are required at both curb and island ends of crosswalks to warn pedestrians who are blind that they are leaving the pedestrian way and entering the vehicular way. Also, planting strips along the sidewalk serve as a barrier that discourages pedestrian access to the roadway at places other than the crosswalk and make it less likely that a blind pedestrian will inadvertently step from the paved walkway into the paved roadway at any point other than the crosswalk or begin crossing from the wrong point without realizing the intersection is a roundabout. They also provide a trailing surface that long cane users can use to locate the crosswalk. The splitter island should be wide enough for pedestrian refuge and to enable a two-stage crossing. Note that splitter islands that are not raised islands but are simply painted on the pavement are not detectable to blind pedestrians. Several blind pedestrians commented that landscaping and trees on the splitter island (at the two-lane roundabout) blocked some of the sound from the lane behind them when they were crossing from the island to the curb. This helped with sound separation and discrimination of the traffic com- ing toward them from the traffic going away from them, and may therefore be beneficial for single-lane roundabouts as well. Landscaping on the splitter islands should not block the view of the crosswalk for drivers. Two-Lane Roundabouts The two-lane roundabout design should promote low speeds at the crosswalk through geometric design, where possible, or through supplemental traffic-calming treatments (bulb-outs or raised crosswalks). The raised crosswalk design showed potential at the tested location and resulted in significantly reduced pedestrian delay and interventions; however, there is concern related to observed multiple threat conflicts, and more research is need to clarify risk. The impact on traffic operations is believed to be directly related to the design of the raised cross- walk (vertical elevation and transition slope). In the tested installations, vehicle impacts were reasonable. More testing is necessary to ensure that this treatment has broader applica- tion to other geometries and traffic patterns. The use of a PHB showed promise at the tested location in terms of reducing pedestrian delay and interventions, but was associated with some misunderstanding and/or noncompli- ance on the side of drivers and blind study participants. A simulation-based sensitivity analysis showed that the use of the PHB phasing, a two-stage crossing, and an offset exit portion of the crosswalk all result in improvements to vehicular oper- ation compared to a standard one-stage pedestrian-actuated signal. If signalization is considered at a two-lane roundabout, these alternate signalization strategies should be considered. Detectable warnings complying with the draft PROWAG are required at both curb and island ends of crosswalks to warn pedestrians who are blind that they are leaving the pedestrian 93
way and entering the vehicular way. Also, planting strips along the sidewalk serve as a barrier that discourages pedestrian access to the roadway at places other than the crosswalk and make it less likely that a blind pedestrian will inadvertently step from the paved walkway into the paved roadway at any point other than the crosswalk or begin crossing from the wrong point without realizing the intersection is a roundabout. They also provide a trailing surface that long cane users can use to locate the crosswalk. The splitter island should be wide enough for pedestrian refuge and to enable a true two-stage crossing. Several blind pedestrians commented that the landscaping and trees on the splitter island blocked some of the sound from the lane behind them when they were crossing from the island to the curb. This helped with sound separation and discrimination of the traffic coming toward them from the traffic going away from them. Landscaping on the splitter islands should not block the view of the crosswalk for drivers. Additional physical separation of the crosswalk from the circulating lane may be considered in the design of the two- lane roundabout to separate driver decision points and to provide added queue storage at the exit leg for yielding drivers. However, a crosswalk too far from the circle may lose the roundaboutâs traffic-calming effect, which reduces speeds and encourages yielding. Low design speeds and traffic-calming treatments may mitigate that tradeoff. Wayfinding and Alignment Treatments This research was primarily focused on the aspect of acces- sibility that is related to the actual decision of when to initialize a crossing. As discussed earlier in this report, the full accessi- bility of a crossing involves three other critical tasks: (1) locat- ing the crosswalk, (2) aligning to cross, and (3) maintaining alignment during crossing. Several treatments are available that can assist in these important accessibility tasks. Ongoing research for the NIH (2010) is currently comparing the effec- tiveness of various wayfinding and alignment treatments. However, even today anecdotal evidence suggests that certain facility design elements and supplementary treatments can be valuable assets to blind travelers. The task of locating crosswalks at roundabouts and CTLs is challenging because crosswalks in these situations are not located at corners. Unless pedestrians who are blind are aware that they are approaching a roundabout or CTL crossing, it is common for them to continue around the bend without real- izing for some time, if at all, that they have gone past the crosswalk. This is true both for those who travel using a long cane and those who use guide dogs. Design elements that help pedestrians locate the crosswalk are landscaping along the curb except at the crosswalk, and the presence of a curb ramp at the crosswalk. This landscaping also may provide a clue to blind pedestrians that the intersection is a roundabout. When the sidewalk is paved to the curb, blind pedestrians may assume they are at a rounded corner and cross at the point where they detect a curb roughly in front of them, which would result in them crossing the circulatory roadway. However, many pedes- trians who are blind prefer to travel near the edge of a sidewalk that is furthest from the street. In this case, unless they are specif- ically looking for a non-corner crossing, they may not be aware of either a break in a landscaping strip or a curb ramp. For those who travel using guide dogs, detecting the break in a landscap- ing strip or a curb ramp requires the use of special strategies. Two treatments are currently suggested for providing effec- tive cues to the location of crosswalks that are not where they are expected. At a crossing with an accessible pedestrian sig- nal, placement of the pushbutton with its pushbutton locator tone immediately beside the curb ramp leading to the cross- walk provides an audible cue to the presence and location of a crosswalk. This feature was noted by pedestrians during the Golden two-lane roundabout posttest with the PHB. A 24-in.- wide strip of a linear texture, sometimes referred to as a âbar tile,â running perpendicular to the sidewalk and across the entire width of the sloped ramp, provides a tactile cue that can be detected underfoot and by use of the long cane. Use of linear textures is uncommon in the United States at pres- ent but has been required or is commonly used to guide pedes- trians who are blind to crosswalks in other countries (Bentzen, Barlow, and Franck 2000). The task of properly aligning to cross can be assisted by a âsquareâ geometry. At a conventional four-legged intersection, blind pedestrians can often align using the cues of adjacent and perpendicular traffic. Since traffic patterns at roundabouts and channelized turn lanes are on curved trajectories, addi- tional alignment cues may be needed. Directional tactile sur- face lines that are accessed by foot and installed concurrent with detectable warnings have some potential, as do returned curb installations that provide a hard edge on each side of the crosswalk ramp that is in line with the direction of travel on the crosswalk. Another potentially effective treatment option involves presenting an auditory signal from the far side of the crosswalk (for example, through an audible device). A far-side audible signal could also be expected to assist with the task of maintaining alignment during crossing. Other alignment treatments are raised markings that delineate the crosswalk or guidance strips that can be raised parallel to the crosswalk. The Access Board draft PROWAG and other U.S. Access Board resources provide additional detail on these and other wayfinding and alignment treatments. The reader is encour- aged to refer to these references for further information. Future Research Needs This research provided a proposed framework for pedes- trian accessibility and presented field study results and other 94
material to inform the ongoing nationwide discussion of the accessibility of roundabouts and CTLs. Clearly there are lim- itations to this research, which are most notably tied to the number of sites that could be captured in the field studies. The number of roundabouts in the United States has grown over the course of this project. When the project team was looking for sites, we had difficulty finding roundabouts with appropriate features, in areas where there were adequate num- bers of blind pedestrians, and in municipalities that were inter- ested in testing various treatments. Now there are examples of roundabouts in various locations with some features that may address accessibility for pedestrians who are blind, for example, yellow pedestrian-actuated beacons installed at the Bird Rock corridor roundabouts in the San Diego area, and other municipalities that are considering the installation of PHBs and raised crosswalks. An ADA-complaint/legal action in Oakland County, Michigan, resulted in court-ordered testing of treatments at two three-lane roundabouts. This study and other activities will provide additional information, which in combination with results of NCHRP Project 3-78A, can give more guidance and comparisons for evaluation of treatments in other locales. But there is much more that needs to be explored in developing crossing solutions for pedestrians with vision disabilities at roundabouts and channelized turn lanes. This research showed serious accessibility problems at the selected CTL site and that the low-cost/low-impact treat- ments were not sufficient to establish accessibility at that site. Since the number of CTLs currently far outnumbers round- abouts, additional research is needed that field tests treat- ments at CTLs and further investigates issues at different CTL designs. Specifically, two additional CTL studies are proposed: (1) testing of a traffic calming treatment such as the raised crosswalk or a pedestrian hybrid beacon at high-volume and high-speed locations, and (2) testing of low-cost treatments (sound strips and flashing beacons) at low-volume and low- speed locations. This additional research would greatly aid the understanding of crossing challenges at CTLs and would lead to the development of a more extensive treatment catalogue for these locations. As discussed earlier in this chapter, little is known about the impact of education and training on the behavior of blind pedestrians, and little is known about how different move- ments or crossing strategies of pedestrians who are blind might affect driver behavior. Both could be fruitful areas for more exploration and research. First, the team observed a number of different techniques used by blind pedestrians in crossing, getting the attention of drivers, and detecting yields, but at this point it is unknown what strategies and techniques work best. Second, there is a need for training programs for pedestrians who are blind, although there is currently no consensus on the form and extent of this training. Consequently, there is a need for determining effective strategies and techniques for deter- mining that an intersection is a roundabout, locating the cross- walks, aligning to cross, determining a safe time to cross, and maintaining alignment while crossing. Developing a training program that could be used by O&M specialists and supported by state DOTs may be a direction to explore further. Future research should perform field testing of additional two-lane roundabout treatments. In particular, more testing is needed to determine under what conditions (geometry, traf- fic) a particular treatment is most appropriate. This research identified two treatments (raised crosswalk and pedestrian hybrid beacon) that showed good potential at the tested two- lane roundabout. These findings should be validated with addi- tional testing, including some at higher-volume roundabouts. Since the potential signalization of two-lane roundabouts is a politically sensitive topic, additional research would give engineers, policy makers, and the U.S. Access Board a more complete understanding of the crossing challenges. In partic- ular, research may get closer to developing actual require- ments and thresholds for the installation of pedestrian signals at two-lane roundabouts with the objective of enhancing the accessibility and usability to pedestrians who are blind. In addi- tion, more research should be done to test low-impact treat- ments at low-volume two-lane roundabouts. Future studies should evaluate the use of, for example, rectangular rapid- flashing beacons (RRFBs) and other flashing beacons, as well as traffic-calming treatments that are more suitable for cold climates or other regions where raised crosswalks may not be allowed in public rights of way. While some of the tested single-lane roundabout crossings may be deemed accessible under prevailing traffic conditions, the analysis did point to some areas of concern, including unexpectedly high delays at one low-volume site (DAV-CLT) and high interventions at a high-volume location (PS-RAL). The team believes that low-cost treatments could enhance blind pedestrian access at these locations. Future tests should explore the impacts of traffic-calming treatments (e.g., raised cross- walks) and auditory treatments such as sound strips. Flashing beacons may further improve driver awareness of pedestrians and promote yielding in the right environments, and more research could clarify where treatments are most useful. In addition, there is a need for new and improved field- based risk performance measures to provide a more objective and consistent assessment of risk and safety. The NCHRP Project 3-78A analysis was limited to the use of O&M interven- tions as a measure of pedestrian risk, which from an analysis perspective has the drawback of being a relatively rare event. The availability of a more readily observed and continuous measure of risk could facilitate the development of safety pre- diction models (similar to the mixed-priority delay models) and could further guide the process of surrogate safety assess- ment in simulation. Potential new risk measures include the time-to-collision measured at the time a pedestrian steps into 95
the crosswalk or the necessary deceleration rate of vehicles to come to a stop prior to the crosswalk. Chapter 6 discussed some initial efforts and demonstrations showing that it is feasi- ble to obtain these data from field and/or video-based measures (Schroeder 2008), and work is currently underway in applying these concepts to multi-lane roundabouts (NIH 2010). A field-based validation of the mixed-priority delay models and work toward extending these models to data from addi- tional sites would provide greater confidence in the modeling. Due to the limited data available in this research, all observa- tions were used for model development. If the models were validated against other sites, analysts would likely have much increased confidence in their viability. Finally, future research should include an increased focus on the auditory environment in the vicinity of the crosswalk to gain a better understanding of the relationship of traffic vol- umes, associated noise patterns, and ultimately the challenges to blind travelers to identify crossing opportunities based on auditory cues. Anecdotal evidence from this research suggests that pedestrians are able to readily cross during all-quiet peri- ods while hesitating or waiting during times of high ambient noise, although this strategy raises concerns in light of more frequent occurrence of quiet (hybrid) vehicles (Wall Emerson and Sauerburger 2008). However, most crossing situations occur somewhere on the continuum between these very quiet and very loud conditions, and it is unclear what types of sound environments and cues are most difficult or most helpful to the blind pedestrian. In closing, significant work remains to be done in this area that focuses on additional treatments, and most importantly, on extending these findings to more locations that are differ- ent geometrically and/or from a traffic operational perspective. This report has established a framework and analysis method- ology that is readily extended and applied to other sites and that will allow future research efforts to be tied directly to findings from this report. The ongoing national debate on the accessibility of modern roundabouts and CTLs has spurred municipalities to take the initiative and tackle some of these accessibility issues prior to the completion of this report and prior to the final adoption of PROWAG. Additional research should take advantage of this momentum and perform con- trolled pretestâposttest evaluations at these locations as treat- ments are being installed. A significant amount of time in this project was devoted to selecting treatments and identifying municipalities willing to install them for study. With already- planned installations, additional research can make very effi- cient use of project resources by focusing on the field studies and data analysis of these locations. 96
