using smartphone app to support visually impaired ...docs.trb.org/prp/13-0231.pdf · 1 using...
TRANSCRIPT
Using Smartphone App to Support Visually Impaired Pedestrians at Signalized 1
Intersection Crossings 2
3
4
5
6
Chen-Fu Liao* 7
Minnesota Traffic Observatory (MTO) 8
Department of Civil Engineering 9
University of Minnesota 10
500 Pillsbury Drive Southeast 11
Minneapolis, MN 55455 12
E-mail: [email protected] 13
Ph: 612-626-1697 14
Fax: 612-626-7750 15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Word Count: Body Text (6,249) + 5 Figures (5 x 250 = 1,250) = 7,499 36
*Corresponding Author37
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 2
ABSTRACT 1
People with vision impairment rely heavily on walking and public transit for their transportation 2
needs. A major challenge for this population is safe crossing of intersections. As a result of the 3
American with Disabilities Act (ADA), Accessible Pedestrian Signal (APS) systems at 4
signalized intersections have improved significantly since 2000. However, these systems still 5
have shortcomings for both users and municipalities, and new approaches are needed to 6
adequately serve pedestrians with low vision. As part of our ongoing effort to develop a 7
prototype Mobile Accessible Pedestrian Signal (MAPS) application for the blind and visually 8
impaired, we interviewed ten blind and low-vision people to better understand what types of 9
information they use at intersection crossings and to identify information types that could assist 10
them. With these survey results, a MAPS system was developed that provides signal and 11
intersection geometry information to smartphone users at signalized intersections. User 12
interaction is via simple tactile input (single or double-tap) and Text-To-Speech (TTS) 13
technology. Field experiment at two signalized intersections was conducted among 18 visually 14
impaired participants to validate the use and functioning of the MAPS system in real-world 15
application. The purpose is to identify if the smartphone app could effectively provide geometry 16
and signal timing information and thus provide decision support for the visually impaired 17
pedestrians. Objective and subjective measures were developed to evaluate the performance of 18
visually impaired users while using the MAPS. Participants reported that the MAPS system 19
provides helpful geometry information (82%) and helpful signal information (59%). 20
21
22
KEYWORDS 23
Pedestrian, Blind, Mobility, Safety, Smartphone, Traffic Signal, Wireless Communication24
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 3
INTRODUCTION 1
The blind and visually impaired usually travel on foot or use public transit as their primary mode 2
of transportation to attend their daily activities. Due to differences in spatial perception as 3
compared to sighted people, they usually encounter physical and information barriers that limit 4
their transportation accessibility and mobility. 5
6
Individuals with vision impairment often have difficulty crossing intersections due to insufficient 7
information available to them about traffic, signal timing and intersection geometry. Among the 8
intersection crossing sub-tasks, locating the crosswalk, determining when to cross and 9
maintaining alignment with the crosswalk while crossing are the most difficult tasks for the blind 10
and visually impaired to execute. A Federal Highway Administration (FHWA) Accessible 11
Pedestrian Signal (APS) settlement [1] stated that “The ADA regulations at 28 Code of Federal 12
Regulations Part 35.130(b)(1)(iii) requires that the aids, benefits, or services provided to 13
individuals with disabilities must be as effective in affording equal opportunity to obtain the 14
same result, to gain the same benefit, or to reach the same level of achievement as those 15
provided to others. The FHWA finds that the lack of accessibility for blind pedestrians is a 16
violation of the ADA.” 17
18
Current actuated intersection requires the blind to search for a pushbutton, if one even exists. It 19
usually requires the pedestrians to move away from their path of travel, which is often used as an 20
alignment cue for crossing. Because of the additional cost of the APS installation, most agencies 21
usually do not deploy them at all signalized intersections. In addition to the installation and 22
maintenance costs which accrue to the local traffic agency, current APS systems contribute 23
“noise” to the local neighborhood. Furthermore, the auditory guiding cues provided by the APS 24
are often inaudible because of the ambient traffic noise associated with rush hour. There is room 25
for improvement in terms of the design and accessibility of both APS and non-APS crosswalk 26
signals for blind and low-vision pedestrians. 27
28
Challenges at Intersection Crossings 29
People with vision impairment primarily use auditory, white cane and limited visual information 30
that they can gather to make safe crossing decisions at a signalized intersection. They generally 31
have difficulty crossing intersections due to the lack of information available to them about the 32
traffic and geometry at intersections [2]. A study of blind pedestrians’ intersection crossing 33
behavior in three cities found that only 49% of crossings started during the walk interval without 34
APS [3]. A follow-up study, after newer types of APS were added, had different results [4]. The 35
follow-up study found that the rate of crossings ended after the onset of the perpendicular traffic 36
decreases from 27% to 7% [4]. 37
38
At signalized intersection crossings where using a pushbutton is required, Barlow et al. [3] found 39
that few (0~16%) looked for and found the button; they also began walking only 20% of the time 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 4
during the walk signal as compared to 72% of the time when the pedestrian phase was on 1
“recall”. The reason may be because searching for the button often requires pedestrian to move 2
away from their path of travel, which is often used as an alignment cue for crossing. In addition, 3
Barlow et al. [3] found that although 72% of blind participants started with an appropriate 4
alignment, location, or both, 42% ended their crossing maneuver outside the crosswalk. 5
6
Therefore, enhancing blind pedestrians’ ability to perceive useful cues at signalized intersections 7
may be an effective method of reducing potential conflicts. There is room for improvement in 8
terms of the design and accessibility of both APS and non-APS crosswalk signals for blind and 9
low-vision pedestrians. 10
11
At intersection crossings, pedestrians typically need to understand traffic patterns for crossing 12
safety and then select a strategy to cross with lower risk. Giudice and Legge [5] reviewed various 13
technologies developed for blind navigation. They concluded that no single technology can 14
provide both indoor and outdoor navigation and guidance for the blind. It is critical to gain more 15
insight from perception study in order to have a clear understanding of the cognitive demands on 16
the blind when they interpret information received by the sensory system. 17
18
As noted earlier, blind people’s transportation choices are mostly limited to walking, taxi and 19
transit. In order to improve mobility, accessibility and their level of confidence in using the 20
system, it is important to remove not only the physical barriers but also the information barriers 21
that potentially impede their mobility. 22
23
Blind pedestrians face higher risk of being involved in a crash when they veer toward the center 24
of an intersection [6]. Blind people tend to veer when crossing quiet streets and the spatial 25
characteristics of the veering tendency differ between and within individuals [6]. Kallie et al. [7] 26
conducted a study of the veering of blind and blindfolded sighted participants and a study of the 27
same participants’ thresholds for detecting the curvature of paths they were guided along. They 28
found humans tend to veer without guideline (or visual feedback) to walk along or a target to 29
walk toward [7]. 30
31
Although there are many aids (such as electronic, Braille map, etc.) to assist wayfinding, blind 32
people tend to use their cognitive map and spatial knowledge as primary guidance [8]. People 33
with low vision, when taught to pay more attention to auditory cues for determining when to 34
cross intersection, often increase their street crossing ability [9]. Street crossing is an important 35
yet challenging task for many vision impaired individuals. However, training and technology can 36
complement each other to improve blind pedestrians’ mobility, safety and accessibility at 37
intersections. 38
39
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 5
Though not always reliable, many environmental cues are available, to support their decision 1
making on various components of the street crossing task. It is important to understand the 2
challenges, identify what information is needed for the blind pedestrian and what is available. 3
Decision making using auditory feedback for the visually impaired usually requires longer time 4
than that based on visual information received by sighted people [9]. 5
6
Pedestrian Navigation and Guidance 7
In order to provide signal information to blind users, Bohonos et al. [10] demonstrated a 8
Universal Real Time Navigational Assistance (URTNA) system using Bluetooth beacons 9
incorporated into a traffic controller to transmit signal timing to a user’s cell phone. URTNA has 10
proven that appropriate software can be developed, but a review of the literature yielded no 11
further research in this area. Barbeau et al. [11] developed a Travel Assistance Device (TAD) 12
using a GPS-enabled smartphone to assist transit riders navigating the public transpiration 13
system. The TAD prompts the rider in real-time with a recorded audio message, visual images, 14
and vibration alerts when the rider should pull the stop request cord to exit the bus. 15
16
Some regions in Europe have begun to develop pedestrian navigation technology to assist the 17
visually impaired or disabled. For example, the city of Stockholm’s e-Adept project [12] has 18
developed a digital pedestrian network, consisting of pedestrian paths, sidewalks, signs, stairs 19
and many detail features based on an open platform that integrates pedestrian navigation 20
technology in order to assist the visually impaired or disabled. The pedestrian navigation system 21
includes digital map, GPS receiver, mobile phone, and inertia navigation module. The digital 22
network integrates municipal data such as road geometry, facility, and traffic information, to 23
provide personal navigation services to elderly and people with disability in both outdoor and 24
indoor environment [13, 14 & 15]. 25
26
Finland’s NOPPA [16] project is designed to provide public transport passenger information and 27
pedestrian guidance through speech interface. The NOPPA system uses GPS, mobile phone and 28
information server to provide door-to-door guidance for visually impaired or sighted users taking 29
public transportation [17]. 30
31
The ASK-IT project [18], partly funded by the European Commission under the 6th Framework 32
Programme, uses personal profiling and web services to provide users with navigation, 33
transportation and accessibility information. The ASK-IT architecture is designed to allow 34
mobility impaired people to live more independently. Users will have access to relevant and real 35
time information through their mobile device primarily for travelling but also for home, work 36
and leisure services. The emphasis is on a seamless service provision and a device that is 37
intelligent enough to address the personal needs and preferences of the user [19 & 20]. 38
39
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 6
The Mobiville project [21] aims to develop a real time multimodal transportation information 1
service and provide location based navigation service for pedestrian using GPS mobile phones. 2
In addition, GeoVector® (http://www.geovector.com) developed an application called World 3
Surfer™ to allow compass-enabled GPS smartphone users to point their phone in a particular 4
direction and search for information about points of interest. This service allows travelers to 5
utilize their smartphone device as a personal travel guide [22]. 6
7
Accessible GPS devices, such as Sendero GPS for the BrailleNote, StreetTalk, Mobile Geo, 8
Trekker Breeze, Loadstone, and others (www.accessibleGPS.com), are commercially available to 9
provide routing information and basic intersection descriptions for the visually impaired. In 10
addition, there are smartphone based applications, such as WalkyTalky and Intersection Explorer 11
on Android platform, and Ariadne GPS and Sendero LookAround GPS for iPhones, that can 12
provide navigation and wayfinding information to assist the visually impaired. But, these 13
systems do not provide traffic signal information. 14
15
To overcome the lack of accessibility of touch-screen-based mobile platforms, Apple Inc. 16
(http://www.apple.com) has included assistive technology as standard features for people with 17
disabilities. The VoiceOver screen reader feature on Apple products allows the device to speak 18
the names of the onscreen menu. It also allows people with vision impairments to interact with 19
the device via gesture combinations. 20
21
ACCESSIBLE PEDESTRIAN SIGNALS (APS) 22
APS systems that indicate the onset of the pedestrian phase are usually deployed at selected 23
intersections to assist blind people with intersection crossing. A typical APS system generates 24
auditory cues continuously to help the blind pedestrian locate the pushbutton. After an APS 25
pushbutton is activated, the APS system announces an auditory message such as “Broadway. 26
Walk sign is on to cross Broadway” [27] when the visual pedestrian signal display is in the 27
“WALK” phase in the corresponding direction. 28
29
Traditional Systems 30
Audible pedestrian signals first appeared in 1920 in the United States. However, they were not 31
included in the U.S. standard, Manual on Uniform Traffic Control Devices (MUTCD) until 2000 32
[23]. In the mid 1970’s, audible signals were mounted on top of the pedestrian signal display 33
(also called pedhead-mounted APS). Two different auditory tones were used to distinguish the 34
north/south (Cuckoo) and east/west (Chirping) directions in the U.S. In Canada, the high-pitched 35
"chirp, chirp" of a small bird, inaudible at busy intersection, is being replaced recently by a four-36
note tune known as the Canadian Melody. 37
38
The audible signal mounted above the pedestrian walk signal has several shortcomings. The 39
“Walk” signal indication is ambiguous and it requires blind pedestrians to know their direction of 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 7
travel at all times. The audible sound is active only when the “Walk” signal is on and there is no 1
indication of pushbutton location if a pushbutton even exists. 2
3
Furthermore, there are several common problems with traditional APS, including the volume of 4
announced messages, not knowing which street has the “WALK” signal on and confusion of 5
alerting tones with traffic noises [24]. Respondents to a survey [25] indicated that “direction 6
taking at the starting position” and “keeping direction while walking in the crosswalk” were 7
problems, even with an APS. Other problems cited in the literature include: confusing acoustic 8
signals from the APS system [26], pushbuttons that are difficult to locate [3]. 9
10
Newer Systems 11
The 2009 MUTCD (section 4E.09 to 4E.13) includes specific ADA standards, guidance and 12
options for APS devices [27]. The newer generation of APS system addresses many of the 13
shortcomings of the earlier system. It provides audible and vibrotactile indication of the “Walk” 14
signal. A pushbutton locator tone that repeats continuously every second is added to provide 15
information about the presence and location of a pushbutton. The pushbutton includes a tactile 16
arrow that points in the direction of travel on the crosswalk. Some of the latest APS system can 17
automatically adjust the volume of audible cue with respect to the ambient noise level. 18
19
There are concerns about noise of APS from residents near the installations. The repeating tone 20
adds 5 decibels of noise within 6 to 12 feet of pushbutton. Ongoing maintenance and Braille 21
verification require additional effort for the public agency. An additional stub is often required 22
for installing pushbutton station poles. In 2009 MUTCD section 4E.08, it requires the pushbutton 23
to be installed “between 1.5 and 6 feet (or no farther than 10 feet) from the edge of curve, 24
shoulder, or pavement”. However, there is no standardized pushbutton pole location. 25
26
In France, E.O. GUIDAGE (http://eo-guidage.com/eng/) uses APS device that is working “on 27
demand” to reduce the noise pollution. The user, using a special remote control, a smartphone or 28
a pedestrian GPS (with Bluetooth), can activate the APS device and get audio messages 29
regarding the status of the pedestrian signal at a fixed-time intersection. The Bluetooth is not 30
used to transmit the status of the pedestrian signal to the user but as a protocol to activate or 31
program APS devices wirelessly. 32
33
New Obstacles 34
In addition to aforementioned challenges for the visually impaired at intersection crossings, two 35
new obstacles are becoming more frequent on US roadways, namely the modern roundabout 36
intersection and quieter Electric Vehicles (EV). Both of which make it even harder for 37
pedestrians with low vision to determine walking direction, select a safe gap between vehicles, 38
and maintain alignment [9]. A study from the National Insurance Institute for Highway Safety 39
[28] reported that intersections converted to roundabouts show a 39% decrease in all crashes and 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 8
an 89% decrease in fatal crashes. In addition to reduction of vehicular injury crashes when 1
intersections were converted to roundabouts, NCHRP report 672 found a reduction in pedestrian 2
injury crashes in British and Dutch [29]. However, the impact on pedestrian accessibility and 3
safety at roundabout crossings, particularly for the elderly, disabled, blind and visually impaired, 4
is still debating. Modern roundabout is aimed to reduce potential conflicts between vehicles but 5
it may not necessarily reduce the conflicts between vehicles and pedestrians. 6
7
OBJECTIVE 8
The objective of this study is to provide intersection geometry and signal timing information via 9
a smartphone for the blind and visually impaired pedestrians. Needs analysis was first assessed 10
to better understand current challenges and what information may be needed to improve 11
mobility. Secondly, a smartphone based accessible pedestrian signal system was developed by 12
taking advantages of sensors on smartphones and incorporating surveyed user needs to 13
effectively provide intersection information to users. Finally, subjective and objective measures 14
were collected through a field experiment to evaluate system performance, usefulness and user 15
satisfaction. 16
17
APPROACH 18
Blind and low-vision pedestrians often report being dissatisfied with their general lack of 19
situation awareness while crossing intersections. A survey was conducted prior to designing the 20
user-interface to better understand the challenges experienced by blind pedestrians and the types 21
of information they use at intersection crossings. Various information modalities that can assist 22
them were also identified. A small sample of visually impaired pedestrians (10 people) was 23
interviewed with several questions regarding their vision, navigation and orientation experience, 24
and use of assistive technologies in wayfinding [30] 25
26
Criteria Identified for Accessible Mobile Decision Support 27
Key recommendations of a mobile decision support system from the user survey are listed as 28
follows. 29
30
1. Need additional information about the intersection and signal timing 31
2. Auditory message should be brief and clear 32
3. Use tactile feedback for warning 33
4. Activate pushbutton automatically or from a mobile device 34
5. Decision support system should not interfere with user’s wayfinding ability 35
36
System Design 37
A Mobile Accessible Pedestrian Signal (MAPS) system for the blind and visually impaired was 38
developed. System communication diagram was illustrated in Figure1. The objective of the 39
system design is to integrate information from sensors that are commonly available on a 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 9
smartphone in determining a user’s location and orientation. The mobile system wirelessly 1
communicates with the traffic signal controller to obtain real time Signal Phasing and Timing 2
(SPaT) information, which together can then inform the blind pedestrian as to when to cross and 3
how to remain aligned with the crosswalk. A roadside device is installed inside a traffic 4
controller cabinet to obtain SPaT and activate a “pedestrian call”. 5
6
For field experiment, a Data Collection Unit (DCU) was installed in a NEMA TS2 traffic 7
controller cabinet to obtain real-time signal timing and phasing information through the 8
Synchronous Data Link Control (SDLC) interface, i.e., Bus Interface Unit (BIU) [31]. The SPaT 9
information is wirelessly transmitted to a signal database server located in the Minnesota Traffic 10
Observatory (MTO) in the Civil Engineering Department at University of Minnesota. When 11
users perform a single-tap on the smartphone screen while pointing to a desired direction, the 12
smartphone app will provide available intersection geometry information by referencing to a 13
digital map loaded on the phone. After determining which direction to cross, the pedestrians can 14
simply perform a double-tap on the screen to submit pedestrian crossing request. The double-tap 15
action then immediately sends the request to the traffic signal controller through the signal 16
database server. 17
18
The pedestrian call interface, residing in a Virtual Private Network (VPN), handles the requests 19
from smartphones to active the pushbutton inputs in the controller cabinet. The VPN is to ensure 20
system security that no unauthorized users or systems can trigger the pushbutton request 21
remotely. Ideally, the wireless router/modem will not be necessary in the future when the signal 22
data can be accessed through the MnDOT's firewall as illustrated in Figure 1. 23
24
A digital map, containing intersection geometry, street name, crossing width in number of lanes, 25
and direction information is stored in a spatial database on the smartphone as a navigational 26
reference. The geospatial database is structured to identify each corner of an intersection and its 27
neighboring nodes (intersection corners) in the vicinity. In order to handle the GPS positioning 28
uncertainty, a Bluetooth device can be included to help determine a pedestrian’s location with 29
respect to an intersection. 30
31
Knowing a crosswalk exists and finding the beginning of a crosswalk are the two key 32
challenging tasks for blind pedestrians approaching an intersection. The ADA ramp and the 33
tactile dome surface often provide useful cues for blind pedestrians to identify the beginning of a 34
crosswalk. However, they do not necessarily provide directional cues for crosswalk alignment. 35
Blind pedestrians usually listen to traffic and use the sidewalk boundaries as a natural alignment 36
for determining the direction of crosswalk. A geo-reference ID can be used to help inform (when 37
geometric clues are not available) and confirm blind pedestrians that they are at a particular 38
corner of an intersection. Blind pedestrians will still need to use their wayfinding skills they 39
already learned from orientation and mobility training to find the beginning of the crosswalk. 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 10
1
User Interface 2
We have developed a Bluetooth geo-ID using commercially available Bluetooth modules. Each 3
Bluetooth module, operating in non-pairing discoverable mode, has a unique Media Access 4
Control (MAC) address. The MAC address of each module is referenced by the smartphone app 5
to identify the corner location of an intersection. The relationship of a geo-ID and corresponding 6
coordinates of an intersection corner are prerecorded and stored in a geospatial database as part 7
of the digital map. The geo-ID can be installed at intersections where GPS reception is weak or 8
unreliable such as an urban canyon environment. 9
10
The smartphone app continuously monitors the location of the phone through GPS after 11
initialization. Initial position and direction of travel are used to search for the initial destination 12
node in the geospatial database. When a user is within the range of a geo-ID tag, the application 13
transmits a Bluetooth scan to identify a nearby tag. 14
15
While standing at a corner of an intersection, a user points the smartphone to a desired direction 16
and taps on the phone once to obtain intersection information announced via Text-To-Speech 17
(TTS) technology. The application then responds with an auditory message, such as “Heading 18
north to Washington Ave, 4 lanes” as illustrated in Figure 2 or “No information in this direction” 19
if the pointed direction is not aligned with a crosswalk within 15 degree. 20
21
After determining which direction to cross, the user points the phone toward the desired direction 22
and taps twice on the smartphone screen to confirm direction of crossing. Request for a walk 23
signal (“pedestrian call”) is then sent to the traffic signal controller through wireless network 24
after the user’s confirmation. The MAPS system then requests and monitors current signal timing 25
from the signal controller, and updates latest signal status with “Wait for signal” or informs the 26
user with 1-sec vibration alert followed by an auditory message such as “Harvard. Walk sign is 27
on to cross Harvard, 20 seconds to cross” as illustrated in Figure 3. 28
29
FIELD EXPERIMENT 30
Locations 31
Two intersections in Golden Valley, west of the City of Minneapolis were identified and selected 32
for the field experiment. The Winnetka Avenue, 5 lanes, goes in north-south direction and the 33
Golden Valley Road, 4 lanes, is in the east-west direction. This intersection is equipped with 34
APS system manufactured by Polara Engineering, Inc. (http://www.polara.com). The second 35
intersection, two blocks away from the first intersection, is located at Rhode Island Avenue and 36
Highway 55. The Rhode Island Avenue, 4 lanes, goes in north-south direction and Highway 55, 37
7 lanes, is in the east-west direction. This intersection does not have APS system installed. 38
Pedestrians are required to use existing pushbutton to cross the intersection. 39
40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 11
Participants 1
The research team randomly selected 20 potential participants from 31 visually impaired adults 2
who contacted us after the advertisement. Among the 20 names selected, 18 (11 male and 7 3
female) individuals responded and participated in the experiment, with mean age of 44.2 years 4
and standard deviation of 15.2 years. Among the 18 participants, two are veterans. 10 5
participants have total peripheral blindness and the others have low vision. 13 of the participants 6
use a long cane and 5 use a guide dog. All participants have completed orientation and mobility 7
training during their lives. Regarding intersection crossing frequency, 9 of the participants cross 8
more than 4 intersections a day, 5 of the participants cross 1 to 2 intersections a day, 2 of the 9
participants cross 1 to 4 intersections a week, and the other 2 participants cross less than 1 10
intersection a week. 9 (50%) of the participants own and use a smartphone, 8 of the participants 11
have regular cell phone, and 1 person does not have a mobile phone. 12
13
Procedures 14
At each location, participants were asked to perform intersection crossings and then interviewed 15
by a research assistant before and after each crossing task. The crossing tasks and interviews 16
focus on participants’ experiences while crossing signalized intersections, using audible 17
pedestrian signals, or a smartphone based accessible pedestrian signal device provided by the 18
research team. 19
20
For each crossing task, a certified orientation and mobility specialist brought each participant to 21
a starting point located about 100-200 feet (north) away from the northeast corner of the 22
intersection. The participants were asked to travel along the sidewalk using their own 23
navigational skills to reach the corner of the intersection. While at the intersection, the visually 24
impaired participants need to find and use the pushbutton to request a pedestrian walk signal or 25
use the smartphone based pedestrian signal device to determine when it is possible to cross. 26
Participants then cross the street that is perpendicular to the sidewalk they just travelled and 27
arrive at the other side of street. 28
29
Each experiment session is about 30 minutes for each participant. The field experiment and 30
interviews were recorded using video and audio equipments to effectively capture participants’ 31
responses and the crossing performance, such as walking speed, in-position time and time to step 32
into crosswalk after the walk sign is on. 33
34
RESULTS 35
Self Assessment 36
Self-reported travel skills and preference of navigation assistance from the participants were 37
reported in the self-assessment questionnaires (multiple choices) prior to the field experiment. 38
The results indicated that white cane was selected as the most preferred method of assistance 39
(44%) followed by asking other people (26%) and using a guide dog (22%). 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 12
1
With regard to the travel skills of the visually impaired participants, 5 participants (28%) 2
responded that their general sense of orientation is about average, while 12 people (66%) 3
responded that their general sense of orientation is above (4 participants, 22%) or well above 4
average (8 participants, 44%). For travel independence, 4 participants (22%) responded as 5
average independent traveler, and 14 participants (78%) considered their independent travel skill 6
is above (5 participants, 28%) or well above average (9 participants, 50%). The survey result 7
shows that 4 participants (22%) considered their skills in crossing signalized intersection as 8
average, and 14 individuals (78%) considered their street crossing skills above (7 participants, 9
39%) or well above average (7 participants, 39%). 10
11
Objective Measures 12
APS pushbutton is required at intersection #1, operated under actuated signal control plan, to 13
request walk signal. At intersection #2, participants were asked to use existing pushbutton for 14
crossing in the first task and then to use the smartphone device to request signal information for 15
the second crossing. Participants’ travel speed as observed from video data at sidewalk (2.5 mph) 16
is slower (statistically significant) than the speed at crosswalk (2.9 mph) using paired t-test (p-17
value = 0.01). However, the travel speed difference on sidewalk and crosswalk at intersection #2 18
is not significant using paired t-test (p-value = 0.06 and 0.77, respectively). 19
20
In average, the visually impaired participants spent 7.8 seconds (SD 6.1 sec) in searching for 21
APS pushbutton at intersection #1. At intersection #2, where no pushbutton-locating tone is 22
available to pedestrians, the participants spent 26.6 seconds (SD 56.5 sec), in average, to locate 23
the regular pushbutton. The pushbutton station pole at intersection #2 is located behind the ADA 24
ramp on grass; it took several participants more than 2 minutes in locating the pushbutton. Four 25
participants veered outside crosswalk path at both intersections. The research team also observed 26
that guide dog may easily be distracted by people close by and thus guided the visually impaired 27
participant toward a person nearby. For example, two guide dogs led the participants toward the 28
camera man at the last quarter of crosswalk instead of staying on the crosswalk path. 29
30
At intersection #1, the average time to step into crosswalk is about 3.1 seconds (SD 1.5 sec). At 31
intersection #2 without the APS device, participants wait about 7.1 seconds (SD 4.7 sec) to step 32
into crosswalk. Without audio notification of when the walk sign is on, the visually impaired 33
pedestrians usually have to listen to the parallel traffic surges. In the experiment, the participants 34
were asked to cross in parallel with Highway 55, which is a busy state highway with Annual 35
Average Daily Traffic (AADT) of 33,500 vehicles. The average time for the visually impaired 36
participants to step into crosswalk at non-APS equipped intersection will vary depending on 37
parallel traffic volume because the visually impaired travelers were taught to judge possible time 38
to cross by listening to the traffic. 39
40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 13
When using MAPS system, the participants averagely waited about 5.5 seconds (SD 3.2 sec) to 1
step into the crosswalk. This is about 2.5 seconds longer than the time observed at APS 2
intersection (#1). The extra two seconds is probably incurred by, (a) the data communication 3
between the smartphone app and the signal controller (1 sec), and (b) the announcement duration 4
of the audible message from the smartphone app when users were trying to listen to and 5
understand what the message meant before stepping into the crosswalk. In addition, the visually 6
impaired pedestrians are more familiar with existing APS system and messages. Shorter APS 7
message may also contribute to the shorter step-into-crosswalk responding time. We expect the 8
average step-into-crosswalk time will decrease when users are more familiar with the MAPS 9
system. 10
11
Subjective Measures 12
Survey results from both intersections were analyzed. Most of the participants (89% or 16 out of 13
18) do not use any accessible GPS device for navigation. 16 participants experienced intersection 14
crossing using APS device previously. At the APS equipped intersection, participants (94% or 17 15
out of 18) preferred pushbutton to activate walk signal request. 14 participants did not have 16
difficulty in locating the pushbutton. However, 3 participants had difficulty locating the button 17
during busy traffic. Most of the participants (11) felt that the APS system provide sufficient 18
information to support their street crossing task. Two participants commented that the audible 19
message was announced in all directions when walk sign is on and the APS system does not 20
provide street width (e.g., number of lanes) information. 11 participants (61%) felt the length of 21
walk phase is sufficient, but the other 7 participants (39%) would prefer more time for crossing. 22
14 participants felt they were aligned at the beginning of the crosswalk. 4 individuals veered 23
outside crosswalk path during the crossing task. 24
25
At second intersection, most of the participants (11 or 65%) preferred not to locate the 26
pushbutton a s compared to using the smartphone to activate walk signal request automatically. 8 27
participants (47%) had difficulty in locating the pushbutton because of the non-standardized 28
pushbutton location. Most participants (15 or 88%) reported that the existing crossing time is 29
sufficient. 15 participants aligned themselves well at beginning of crosswalk; however, 4 people 30
wavered off the crosswalk. 31
32
During the second crossing task while using the MAPS device, a few participants experienced 33
unreliable GPS signal that led to the result of the smartphone app being confused and considered 34
the participant being already at the other side (NW corner) of the intersection. The inconsistent 35
GPS positioning solution caused the MAPS to provide inconsistent information. 36
37
As compared to the 2 crossing tasks at intersection #2, 11 participants felt (65%) the MAPS 38
system provide sufficient information, 14 (82%) people responded that MAPS provides helpful 39
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 14
geometry information, but only 10 participants (59%) felt the MAPS provides helpful signal 1
information due to the GPS positioning difficulty. 2
3
Usefulness and Satisfaction 4
Van der Laan et al. [32] developed a simple procedure to measure the perceived satisfaction and 5
usefulness of a new system. The technique consists of nine 5-point rating items with bipolar 6
adjective scales. These scales are summed to generate separate scores for the perceived 7
satisfaction and usefulness measures. Score of each individual question ranges from a bipolar 8
scale -2 to 2. Survey results from each participant on usability and acceptance were analyzed. 9
The scores for the perceived satisfaction and usefulness were 1.04 and 0.95, respectively. Results 10
of the usability questionnaire indicated that the visually impaired participants considered the 11
MAPS system to be moderately useful and satisfying as shown in Figure 4. 12
13
System Trust and Confidence 14
The trust questionnaire was used to measure different dimension of trust: purpose, process, 15
performance, and overall trust [33]. The performance dimension of trust measures the 16
expectation of consistent, stable and desirable performance of the system. The process dimension 17
of trust represents the understanding of underlying qualities that govern the system behavior. The 18
purpose dimension of trust is the underlying motives or intent of the system. Survey results on 19
system trust and confidence are displayed in Figure 5. 20
21
In average, the participants reported 65% of trust in system performance and confidence. 22
Participants reported 79% and 71% in understanding the purpose of the system and its 23
underlying behavior, respectively. Overall, the visually impaired participants reported 57% of 24
trust in using the system. The system trust and confidence scores from the participants are not as 25
high as we hope for. We believe improving the positioning reliability together with additional 26
training and instruction on using MAPS will help improve users’ trust and confidence in the 27
future. 28
29
SUMMARY AND CONCLUSION 30
We have interviewed 10 blind and low-vision people to understand what types of information 31
they use at intersection crossings and identified information types that could assist them. 6 high-32
level recommendations were included for the design of MAPS. 33
34
The MAPS system is a smartphone based decision support system that is intended to provide 35
geometry and signal timing information to the visually impaired at intersections. The MAPS 36
system uses wireless communications, GPS, Bluetooth, and motion sensors embedded in the 37
smartphone to determine orientation and location of a user at an intersection. Intersection 38
geometry information is provided to users when tapping once on the screen of a smartphone 39
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 15
while pointing the phone to a desired direction. Pedestrians can use the information contained in 1
the audible messages to make more informed decisions in determining which street to cross. 2
After determining which direction to cross, the visually impaired pedestrians can perform a 3
double-tap on the smartphone screen to request pedestrian walk phase and receive signal timing 4
updates. Pedestrians can use the audible messages to better determine when walk sign is on. 5
6
The MAPS system has employed human-centered design by incorporating survey feedbacks 7
from previous study on user needs and preferred user interfaces. The simple and easy-to-use 8
inputs and short auditory messages are intended to minimize additional task load for the visually 9
impaired at intersection crossings. 10
11
One of the goals is to validate the use and functioning of the MAPS system in real-world. The 12
purpose is to identify if it could provide geometry and signal timing information and thus 13
provide decision support for the visually impaired pedestrians. This was accomplished by 14
recruiting 18 visually impaired participants for a field experiment. Pedestrians were given brief 15
tutorial on how to use smartphone app including pointing to different direction for geometry and 16
signal timing information. Participants were asked to perform 1 crossing task at first intersection 17
already equipped with APS system and 2 crossing tasks at second intersection with and without 18
the use of MAPS system. Objective measures were compared to evaluate users’ performance 19
using different type of assistances. In fact, current work sought to evaluate the usability of the 20
MAPS system to better understand users’ perceptions of workload, satisfaction, usefulness, and 21
willingness to use the information presented by the smartphone app. 22
23
Although, the input interface of using single and double-tap may seem relatively simple and easy 24
for the visually impaired participants to operate. However, 15-minute of tutorial on how to 25
properly use the device is not sufficient. There is still a learning curve for the users to understand 26
how the system works. For example, a common issue is that the participants usually point the 27
phone to a desired direction and immediately bring the phone to their ears in order to better listen 28
to the auditory message at a busy intersection. The MAPS system was confused by the desired 29
direction of information when users brought the phone to their ears before the orientation 30
measurements from the digital compass were finalized. 31
32
The system was tested at experiment site (relatively open sky) by the research team for several 33
times without GPS signal reception difficulty. However, during the 10-hour of field experiment, 34
a few participants experienced incorrect information due to the fact that the GPS receiver on the 35
phone was not able to indentify user’s correct location. A Bluetooth geo-ID, as previously 36
proposed for urban area, is definitely needed to reliably determine user’s location at 37
intersections. Enhancements, such as reducing the latency of signal data communication and 38
adaptively learning available pedestrian walking time from the signal controller, were made to 39
the existing system after field experiment. This will allow the MAPS to provide walk time 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 16
countdown information to the visually impaired pedestrians if they prefer to hear the countdown 1
announcement from user configuration in the future. 2
3
In the future, we would like to explore and investigate alternatives to make the system hands-free 4
and include user configurable settings for selectable vibration pattern, message frequency, and 5
speech rate as suggested by the participants. We also would like to use this smartphone based 6
approach to investigate solutions for navigating visually impaired pedestrians around work zones 7
or at bus stops by incorporating Bluetooth Geo-ID or other positioning technologies. 8
9
ACKNOWLEDGMENTS 10
This project was sponsored by the Intelligent Transportation Systems (ITS) Institute, University 11
of Minnesota, with funding from the US Department of Transportation through the University 12
Transportation Centers program. The author thanks the participants for completing challenging 13
street crossing tasks. Many thanks to Linda Spaulding and Nancy Prussing (orientation and 14
mobility specialists), Michael Rakauskas (human factors scientist), Steve Misgen and Timothy 15
Bangsund (MnDOT traffic engineers), and Avanish Rayankula, Sowmya Ramesh and Niveditha 16
Baskar (graduate students). The author also would like to thank Professor Gordon Legge 17
(Department of Psychology) and Dr. Michael Manser (HumanFIRST program) for numerous 18
insightful discussions. 19
20
REFERENCES 21
[1] FHWA APS Settlement (2005), http://www.acb.org/pedestrian/fhwa-letter-concerning-22
aps11.html, accessed May 2011. 23
[2] Ponchillia, P., Rak, E., Freeland, A., & LaGrow, S. (2007). Accessible GPS: 24
Reorientation and Target Location Among Users with Visual Impairments. Journal of 25
Visual Impairment & Blindness, 101(7), 389-401. 26
[3] Barlow, J.M., Bentzen, B.L., & Bond, T. (2005). Blind Pedestrians and the Changing 27
Technology and Geometry of Signalized Intersections: Safety, Orientation, and 28
Independence. J. of Visual Impairment & Blindness, 99(10), 587-598. 29
[4] Scott, A.C., Barlow, J. M., Bentzen, B.L., Bond, T.L.Y. & Gubbe, D. (2008). Accessible 30
pedestrian signals at complex intersections: Effects on blind pedestrians. Transportation 31
Research Record, No. 2073, pp. 94-103. 32
[5] Giudice, N. A. & Legge, G. E., (2008). Blind Navigation and the Role of Technology, in 33
The Engineering Handbook of Smart Technology for Aging, Disability, and Independence 34
(eds A. Helal, M. Mokhtari and B. Abdulrazak), John Wiley & Sons, Inc., Hoboken, NJ. 35
[6] Guth, D., & LaDuke, R., (1995). Veering by blind pedestrians: Individual differences and 36
their implications for instruction. Journal of Visual Impairment & Blindness, 89, 28–37. 37
[7] Kallie, C. S., Schrater, P. R., & Legge, G. E. (2007). Variability in stepping direction 38
explains the veering behavior of blind walkers. Journal of Experimental Psychology: 39
Human Perception and Performance, 33(1), 183-200. 40
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 17
[8] Golledge, R. G. & Gärling, T., (2004). Cognitive maps and urban travel, in: D. A. 1
Hensher, K. J. Button, K. E. Haynes and P. R. Stopher (Eds) Handbook of Transport 2
Geography and Spatial Systems, Ch. 28. Amsterdam: Elsevier. 3
[9] Long, R. (2007). Crossing Streets Without Vision: Access to Information, Strategies for 4
Traveling, and the Impact of Technology, Training, and Environmental Design. In J. J. 5
Rieser, D. H. Ashmead, F. Ebner & A. L. Corn (Eds.), Blindness and Brain Plasticity in 6
Navigation and Object Perception (pp. 335-351): Psychology Press. 7
[10] Bohonos, S., Lee, A., Malik, A., Thai, C., & Manduchi, R. (2008). Cellphone Accessible 8
Information via Bluetooth Beaconing for the Visually Impaired. Santa Cruz: University 9
of California, Santa Cruz. 10
[11] Barbeau, S.J., Georggi, N.L., & Winters, P.L. (2010). Integration of GPS-Enabled Mobile 11
Phones and AVL: Personalized Real-Time Transit Navigation Information on Your 12
Phone, Proceedings of the National Academy of Sciences, Transportation Research 13
Board 89th Annual Meeting, Washington, D.C. http://www.locationaware.usf.edu/wp-14
content/uploads/2010/04/TRB2010-Paper.pdf, accessed June, 2012 15
[12] The e-Adept accessibility project (2009). http://www.eadept.se, accessed October 2011. 16
[13] Jonsson, A., Sandberg, S., & Uppgård, T., (2007). Urban guidance for the blind and 17
visually impaired – pedestrian route planning and navigation using a mobile handset, 18
Expanding the Knowledge Economy: Issues, Applications. Case Studies, P. Cunningham 19
and M. Cunningham (Eds), IOS Press, 2007 Amsterdam, ISBN: 978-1-58603-801-4. 20
[14] Dawidson, E., (2009). Pedestrian Navigation in Stockholm, How Local Data Together 21
with Advanced Positioning Techniques Can be Used for Detailed Routing. Proceedings 22
of the 16th ITS World Congress, Stockholm, Sweden. 23
[15] Johnni, P., (2009). Stockholm – A City for Everyone How New Technology Can Make 24
Every Day Life Easier for Elderly and People with Disabilities. Proceedings of the 16th 25
ITS World Congress, Stockholm, Sweden. 26
[16] The NOPPA project (2009). http://virtual.vtt.fi/virtual/noppa/noppaeng.htm, accessed 27
October, 2011. 28
[17] Virtanen A. & S. Koshinen (2004). NOPPA – Navigation and Guidance System for the 29
Blind. Proceedings of the 11th ITS World Congress, Nagoya, Japan. 30
[18] The ASK-IT project (2009). http://www.ask-it.org/, accessed October 2011. 31
[19] Bekiaris, E. , K. Kalogirou, & M. Panou, (2007). Personalised infomobility services 32
within ASK-IT project. Proceedings of the 14th ITS World Congress, Beijing, China. 33
[20] Edwards, S., Bryan, H., & Wagner, J., (2007). Indoor Localization for Disabled People: 34
ASK-IT Project, Proceedings of the 14th ITS World Congress, Beijing, China. 35
[21] Coldefy, J., (2009). The Mobiville Project: Large Scale Experimentation of Real-Time 36
Public Transport Information Service Providing Satellite Based Pedestrian Guidance on 37
Smart Phones. Proceedings of the 16th ITS World Congress, Stockholm, Sweden. 38
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 18
[22] Markoff, J., & Fackler, M., (2006). With a Cellphone as My Guide, The New York 1
Times. http://www.nytimes.com/2006/06/28/technology/28locate.html, accessed June 2
2012. 3
[23] Accessible Pedestrian Signals, http://www.apsguide.org/index.cfm, accessed June, 2012 4
[24] Bentzen, B. L., Barlow, J. M., & Franck, L. (2000). Addressing barriers to blind 5
pedestrians at signalized intersections. J. of Institute of Transportation Engineers (ITE). 6
70(9), 32-35. 7
[25] Harkey, D.L., Carter, D.L., Barlow, J.M., Bentzen, B.L., (2007). Accessible Pedestrian 8
Signals: A Guide to Best Practices, NCHRP 117A, Transportation Research Board. 9
[26] Tauchi, M., H. Sawai, J. Takato, T. Yoshiura, & K. Takeuchi., (1998). Development and 10
Evaluation of a Novel Type of Audible Traffic Signal for the Blind Pedestrians. The 9th 11
International Mobility Conference Proceedings. Rehabilitation Research and 12
Development Center, Veterans Administration Medical Center, Decatur, GA, pp 108–13
109. 14
[27] FHWA, 2009 Manual on Uniform Traffic Control Devices (MUTCD), 15
http://mutcd.fhwa.dot.gov/kno_2009.htm, accessed July, 2012 16
[28] Bergh, C., Retting, R.A., & Myers, E., (2005). Continued reliance on traffic signals: the 17
cost of missed opportunities to improve traffic flow and safety at urban intersections, 18
Insurance Institute for Highway Safety. Arlington, VA. 19
http://www.iihs.org/research/paper_pdfs/mf_1848.pdf, accessed June, 2012 20
[29] NCHRP Report 672, (2010). Roundabouts: An Informational Guide – Second Edition, 21
Section 5.3.3 Pedestrians, Transportation Research Board, p.5-19~20. 22
http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_672.pdf, accessed October, 2012 23
[30] Liao, C.F., Rakauskas, M., & Rayankula, A., (2011). Development of Mobile Accessible 24
Pedestrian Signals (MAPS) for Blind Pedestrians at Signalized Intersections, Final 25
Report, Intelligent Transportation Systems Institute, Center for Transportation Studies 26
(CTS 11-11), University of Minnesota. 27
http://www.cts.umn.edu/Publications/ResearchReports/pdfdownload.pl?id=1552, 28
accessed July, 2012. 29
[31] Liu, H., Wu, X., Ma, W., & Hu, H. (2009). Real-Time Queue Length Estimation for 30
Congested Signalized Intersections, Transportation Research Part C, 17(4), 412-427. 31
[32] Van der Laan, J.D., Heino, A. & de Waard, D. (1997). “A Simple Procedure for the 32
Assessment of Acceptance of Advanced Transport Telematics.” Transportation Research 33
Part C: Emerging Technologies. Vol.5, No.1, Elsevier, P.1-10 34
[33] Lee, J. & Moray, N., (1992). Trust, control strategies, and allocation of function in 35
human-machine systems, Ergonomics, 35, 10, pp.1243-1270. 36
37
38
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 19
LIST OF FIGURES 1
2
Figure 1 System Diagram of MAPS 3
Figure 2 Single-Tap for Obtaining Intersection Geometry Information 4
Figure 3 Double-Tap for Confirming Crossing and Obtain Signal Information 5
Figure 4 System Usefulness and Satisfaction 6
Figure 5 System Trust and Confidence 7
8
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 20
1
2 Figure 1 System Diagram of MAPS 3
4
5
6
7
Heading North to
Washington Ave 4‐Lane
Heading West to
Oak Street 4‐Lane
8 Figure 2 Single-Tap for Obtaining Intersection Geometry Information 9
10
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 21
1 Figure 3 Double-Tap for Confirming Crossing and Obtain Signal Information 2
3
4
5
6 Figure 4 System Usefulness and Satisfaction 7
8
9
10
11
12
TRB 2013 Annual Meeting Paper revised from original submittal.
Liao (#13-0231) 22
65%
71%
79%
57%
65%
0% 25% 50% 75% 100%
Performance (1,4)
Process (2,5)
Purpose (7,8)
Overall (3,9)
Confidence (6)
1 Figure 5 System Trust and Confidence 2
3
4
TRB 2013 Annual Meeting Paper revised from original submittal.