Content uploaded by Miao Song
Author content
All content in this area was uploaded by Miao Song on Feb 23, 2017
Content may be subject to copyright.
Messina, Song, Ortiz-Varela, Wang 1
ASSESSING THE MESSAGE DESIGN ON VARIABLE MESSAGE SIGNS IN MITIGATING
THE BOTTLENECK ISSUE AT WORK ZONES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
by
J. Messina
Department of Mechanical, Industrial and Systems Engineering
University of Rhode Island, Kingston, RI 02881
Phone: (401) 626-8384
E-mail: jmessina777@gmail.com
M. Song
Department of Mechanical, Industrial and Systems Engineering
University of Rhode Island, Kingston, RI 02881
Phone: (401) 525-1232
E-mail: songm@egr.uri.edu
J.D. Ortiz-Varela
Department of Civil Engineering and Surveying
University of Puerto Rico, Mayagüez, Puerto Rico 00680-9000
Phone: (787) 310-1375
E-mail: josue.ortiz3@upr.edu
J. H. Wang [Corresponding Author]
Department of Mechanical, Industrial and Systems Engineering
University of Rhode Island, Kingston, RI 02881
Phone: (401) 874-5195, Fax: (401) 874-5540
E-mail: jhwang@egr.uri.edu
Prepared for the 91st Annual Meeting of the
Transportation Research Board
Washington D.C.
January 2012
Submitted November 15, 2011
Word Count
Abstract: 250
Body: 5449
Tables: 3 × 250 = 750
Figures: 4 × 250 = 1000
Total: 7449
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 2
ABSTRACT 47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
In this study, drivers’ preferences of and responses to text and graphic road sign messages at
work zones were analyzed in an attempt to reduce the bottleneck conditions at lane-reduced work
zones. Advisory messages associated with three driving advisory conditions (DACs), “Merge to
the Right Lane,” “Zip Merge” (vehicles take turns), and “Continue Travel Normally,” were
assessed through a questionnaire survey and driving simulation to seek the best messages in
advising drivers in different traffic conditions when approaching work zones.
A questionnaire survey was first deployed to identify participants’ preferences towards a
series of messages posted on variable message signs (VMSs). Participants rated each message
from one to five as to their effectiveness in advising drivers in different conditions. Between the
highest rated text and graphic messages under each DAC, participants gave their preferences
toward either text or graphic messages. A total of 81 subjects participated in the survey. Survey
results indicated that text messages were strongly preferred over graphic messages in all DACs.
The effectiveness of several top rated messages identified in the survey was further
assessed through a driving simulation. Various text and graphic messages were posted on
portable VMSs along a straight freeway in a fix-based driving simulator. Subjects were asked to
verbally respond with a number when they identified a message, denoting the DAC associated
with that message. It was found that graphic messages were most effective in all three DACs in
terms of response time and accuracy. Recommendations for messages in each DAC were made.
Keywords: work zone, variable message sign, dynamic lane merging system
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 3
INTRODUCTION 68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
With the increasing motorization, urbanization and population growth, traffic flow management
has become an important task for transportation authorities throughout the nation. As the
roadway transportation system of the United States has matured, roadway construction under
varied traffic conditions has become the rule rather than the exception. Traffic flow management
at work zones has become a top priority among all traffic management issues. At work zones,
natural traffic flow is disrupted when lane closures occur. Due to the capacity diminution, heavy
congestion or “bottlenecks” resulting from lane merging occur in the midst of high traffic
demand.
Motivated by the bottlenecks observed in lane-reducing work zones, this study seeks a
better way to display messages on portable dynamic lane merging system (PDLMS) to advise
drivers approaching work zones. It is aimed to better manage traffic flow and eliminate
bottlenecks at work zones to improve the safety and efficiency in freeway travel.
It has been observed that dangerous vehicle maneuvers exist at freeway work zones
where traffic merging occurs. Such maneuvers include quick braking and speeding up, as well
as vehicles in the open lane not allowing vehicles in the closed lane to merge sufficiently because
of a perceived right of way. Messaging at lane-reduced work zones, through the use of variable
message signs (VMSs), is in high regard for the state of Rhode Island’s Intelligent
Transportation System.
To understand the effectiveness of certain advisory messages in promoting desired
driving behavior, the effect of such messages on vehicle behavior, and to improve the message
design and displays at work zones, the study was carried out in two parts. First, a questionnaire
survey was conducted to examine drivers’ preferences towards the design and display of VMS
messages as they would apply to certain work zone driving advisory conditions (DACs). The
survey was developed as a series of PowerPoint slides, where the messages were presented under
three DACs. A driving simulation experiment was next conducted to determine the effectiveness
of these messages, specifically how accurately and quickly drivers responded to their intended
meanings. The findings of this study could help transportation authorities improve their
management of freeway work zone traffic and eliminate dangerous driving maneuvers in such
condition.
BACKGROUND
Bottleneck Issue at Work Zones
Traffic congestion is often observed at work zones with temporary capacity reduction (1).
Increased travel time, queue length, aggressive behaviors, and roadway accidents are commonly
seen (2). Between 1982 and 2005, the percentage of the major road system that is congested
grew from 29% to 48% in the United States (3). Approximately 10% of travel time delays occur
at roadway work zones (4). Work zones on freeways are estimated to account for nearly 24% of
non-recurring delay (5). In 2000, Federal Highway Administration conducted a survey in which
32% of people were dissatisfied with the areas of construction, placing work zone dissatisfaction
as the second highest rate in user dissatisfaction on major highways (6).
Despite all conventional efforts, work zones remain hazardous places (7). Research has
shown that drivers are slow to recognize they have entered a work zone, causing crashes and
subsequent decreases in roadway capacity. According to the Fatality Analysis Reporting System
(FARS), there were 720 work zone fatalities in the United States in 2008; this figure represents
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 4
2% of all roadway fatalities for that year. There was one work zone fatality every 10 hours and
one work zone injury every 13 minutes (8). Undoubtedly, it has become a critical challenge for
traffic management and safety engineers to maintain a satisfactory level of efficiency and safety
at work zones without sacrificing roadway functions.
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
Dynamic Lane Merging System
To increase the efficiency and safety of traffic flow in the lane-reduced transition areas,
engineers around the world have been exploring innovative techniques to facilitate traffic flow
through these bottleneck areas. Among them, traffic control devices and merging strategies play
an important role in managing the flow of vehicles.
Based on the conventional Manual of Uniform Traffic Control Devices (MUTCD) lane
merge control strategy, alternative strategies have been found to enhance the safety and
efficiency of transition areas and deal with the traffic flow control issues. Among them are the
“Early Merge (EM)”, “Late Merge (LM)”, “Dynamic Early Merge (DEM), and “Dynamic Late
Merge (DLM)” (9,10,11,12). Normally, EM and DEM work well as long as congestion does not
develop. When the traffic demand exceeds the capacity of the work zone, queues may extend
back beyond the advance warning signs, often surprising approaching traffic and increasing the
accident potential. LM addresses many of these problems, which maximize the traffic capacity
of the work zone. When there is no congestion and speeds are high, potential confusion among
drivers at the merge point becomes a concern, and the DLM concept is proposed in the interest of
providing the safest and most efficient merging operations at all times.
The Minnesota Department of Transportation developed a dynamic traffic control
strategy, the DLM system, and deployed it on a session of US 10 in 2003. In addition to the
standard signs, this system consists of three VMSs and a Remote Traffic Microwave Sensor
(RTMS) detector. When congestion begins to form, the signs are activated to provide lane use
instruction to drivers. It was found that the percentage of drivers utilizing the discontinuous lane
increased dramatically (almost 60% during the heaviest demand) when the VMSs were activated,
which indicated that the queue length decreased and traffic capacity increased (13).
The Michigan Department of Transportation has deployed an early merge strategy known
as the Dynamic Early Lane Merge Traffic Control System (DELMTCS), in an attempt to
increase vehicle throughput and overall safety near construction lane closures. This strategy
employed EM by setting up a dynamic no passing zone. The DELMTCS helped Michigan DOT
achieved its goals of reducing aggressive driving behavior, improved overall safety, and reduced
lane closure related delay (12).
Zip merge is a strategy that encourages drivers to take turns when merging into reduced
lanes at work zones, and can be applied in the strategies previously mentioned. Since both lanes
are used, people take turns, and stress and road rage are correspondingly reduced. This operation
requires motorists to follow a “zipper rule,” in which drivers in a continuing lane permit adjacent
vehicles to merge in an alternating pattern. In this instance, right-of-way assignment is
suspended until the congested period ends (14). It is considered an effective tactic for merging
traffic from several to fewer lanes with the least road rage. Zip merge operations provide an easy
and efficient solution for traffic flow management when lane closures occur. However,
operational difficulties are often experienced by transportation authorities. Most motorists in
open lanes would not give up their right-of-way at the merge point. They commonly try to
prevent drivers in the closed lane from passing them by straddling the centerline or traveling
slowly in tandem with another vehicle in the closed lane. As a result, heavy congestions are
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 5
formed in closed lanes, and orderly merging operations are lost when impatient drivers
remaining in the closed lane attempt to squeeze into the open lane. These maneuvers tend to
reduce the capacity of the merging operation and increase the accident potential and road rage
among drivers (15).
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
To make sure that zip merge is enforced and followed by all motorists at work zones, a
research team in the Czech Republic has developed a portable Dynamic Lane Merging System
(DLMS). It is consisted of two mutually communicating telematic units - a mobile display
system with an evaluation unit (MDSE) and a mobile telematic station (MTS). The MDSE unit
is equipped with traffic information display, portable variable message sign, and radar traffic
detector. It could display both text messages and pictograms to assist motorists with lane
merging at work zones. The modular MTS is equipped with a surveillance camera, a weather
station, and a radar traffic detector. Traffic data collected by the MTS are subsequently
interpreted and used to display short messages or pictograms on the MDSE to help manage
traffic flow at the work zone by facilitating zip merge behavior (16).
Portable Variable Message Sign
As a critical component in the Dynamic Lane Merging System, a Portable Variable Message
Sign (PVMS) is a useful tool for managing traffic in real time at work zones. By giving drivers
clear and direct instructions, a well-designed PVMS could help effectively reduce congestion,
while maintaining safety at lane-reduced work zones.
PVMS messages should not only be recognizable to drivers, but also be coherent and
legible from a distance. Therefore, installation and message operations considering drivers’
legibility performance are very important. Drivers’ legibility distances are dependent on various
factors, such as geometric conditions, travel speed and driver characteristics, etc. (17). There
have been a variety of relevant research studies and experiments on legibility performance for
VMS. In 1994, Armstrong and Upchurch emphasized reflecting ergonomic factors into
designing a VMS, and suggested legibility distance models through experiments that compare
legibility of fiber-optic and Light-Emitting Diode (LED) characters (17). In 2005, Wang, J.H.
and Cao, Y. (18) developed a VMS information legibility model with number of lanes and
number of lines of messages as the main variables, and age and gender as the other variable,
using a driving simulator (19).
Efficiency of Graphic and Text Sign Messages
The content of a road sign, in terms of level of detail, could significantly affect the reaction time
of motorist, as well as their willingness to make a maneuver, and therefore affect the safety of
the roadways. Thus, the more specific and clearer a message is, the more persuasive and
influencing it becomes (20). Plummer et al investigated the effectiveness of text versus graphics
in conveying a desired message. Graphics were found more effective in conveying the intended
message, whereas the difference in comprehension speed between text and graphic messages
could not be determined (21). Wang et. al. (22) conducted a study on the use of graphics on
VMSs and found that most test drivers both preferred and responded faster to graphic-aided text
messages than text-alone messages. It is also recommended by the Conference of European
Directors of Roads (C.E.D.R.) report and Lucas et. al.’s study that graphics and symbols should
be used as much as possible to avoid the problem of disseminating information to drivers who
speak and use different languages (23,24).
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 6
Symbols and graphics displays on PVMS offer potential advantages because drivers can
read and understand symbols and graphics quicker and farther upstream of the sign in
comparison to word messages (25). Field and laboratory results have indicated that efficient
graphic message sign have several advantages over word message signs such as legibility for a
given size and at shorter exposure durations, recognition when the information is degraded due
to poor environmental legibility, quicker extraction of information and drivers who have
difficulty understanding text sign messages are able to comprehend pictographs (26).
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
DESCRIPTION OF STUDY
To gain insight into drivers’ preference and reactions towards work zone VMS messages, three
DACs, “Merge to the Right Lane,” “Zip Merge” and “Continue Travel Normally,” were
employed in the study, selected by transportation engineers and researchers at the Rhode Island
Department of Transportation and Federal Highway Administration, with influence from signs
already deployed in the Czech Republic for zip merging. Key points of interest in the
collaborative meetings were the differences between text and graphic sign messages as well as
the differences of these DACs.
Two approaches were employed in the study, a questionnaire survey and a driving
simulation experiment. The survey was developed to assess participants’ preferences towards
either text or graphic sign messages, and also their ratings of individual messages in determining
their effectiveness in advising drivers at different conditions. The driving simulation experiment
aimed to gain insight to drivers’ responses to these messages as a resemblance of their responses
in actual driving through freeway work zones. The experiment design allowed a complete
analysis of the participants’ accuracy and speed in identifying the messages with different
message type (text or graphic) and under different DACs (Figure 1). A detailed description of
both the questionnaire survey and driving simulation experiment is given below.
Merge to the Right Lane (DAC 1) Zip Merge (DAC 2) Continue Travel Normally (DAC 3)
Text Graphic Text Graphic Text Graphic
231
232
233
234
235
236
237
238
239
240
* Last row messages not used in driving simulation experiment
FIGURE 1 Variable messages by message type (M) and driving advisory condition (DAC).
Questionnaire Survey
Design a Computer-based Questionnaire Survey
The survey, created using Microsoft PowerPoint with Visual Basic macros, consisted of nine
questions, with three questions each pertaining to one of the three DACs. Prior to the
presentation of each question, the DAC was explained in a way to avoid possible bias towards
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 7
one message or another. The participants were asked to evaluate several road sign messages as to
their ability to promote the desired driving behavior in that specific DAC. The participants were
asked to give a 1-5 rating to both text and graphic sign messages associated with a DAC. A total
of 18 messages were used in the survey, with three text and three graphic sign messages under
each DAC. Ratings for each of the 18 sign messages used were gathered through the first two
questions in the survey. The participants were next prompted to choose between their top rated
text and graphic sign messages for each DAC. This was made possible by using Visual Basic
macros, where participants were automatically directed to a specific question based on their
ratings given in previous questions.
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
The “text versus graphic” questions were included to reveal possible participants’
preferences towards either message type. Figure 2 shows the set of three questions in a series
used for the “Merge to the right” DAC.
Conducting the Survey
The survey was conducted at the Warwick Mall and at the University of Rhode Island. A total of
81 subjects, 37 females and 44 males, participated in the survey. Among them, 35 were between
18 and 25 years old, 20 between 26 and 40 years old, and 26 were 41 years and older.
Participants were informed that the survey was completely voluntary and that they were not
obligated to finish. This was also included in the introductory PowerPoint slides prior to the
questions, as was an acceptance of consent in the form of an electronic signature (approved by
the University’s Institutional Review Board.) Participants were told the survey should take about
five minutes to complete, but no time constraints were put on participants. After completing the
nine questions of the survey, participants were asked to provide certain demographic information
including age group, gender, and native language.
Driving Simulation Experiment
Design of Experiment
The driving simulation experiment was designed to assess the text and graphic sign messages
used in the survey, under the same three DACs. Due to time considerations, the lowest rated text
and graphic sign messages in each DAC were not included in the driving simulation. Thus, only
two text and two graphic sign messages were tested under each DAC (the first two rows of
messages in Figure 1). With the additional three “dummy” messages, a total of 15 messages
were presented in the simulation. Each of the 15 messages would be viewed twice by each
participant in a random order during the experiment.
Driving Simulator
The TranSim VS IV Simulator, used in the driving simulation experiment, is a fixed-base
simulator which consists of a regular driving module and three channel plasma monitors in an
immersive driving environment that combines the look and feel of a real vehicle. Participants
interacted with the simulator using the sedan’s steering wheel and pedals that provide real-time
feedback. A separate program called “ScenarioBuilder” was used to create desired conditions of
the experimenter and delivers sharp visuals and crisp images.
A total of six modules with five messages each were created for this experiment.
Messages chosen in a module and the order of the modules presented to a subject were
completely randomized. In addition, no more than three text or three graphic messages were
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 8
allowed in a module to preserve a balance of text and graphic messages throughout the
experiment runs.
287
288
289
(a) 290
(b) 291
(c) 292
293
294
295
296
297
298
299
300
301
FIGURE 2 A sequence of PowerPoint slides used in the questionnaire survey for the
“merge to the right” driving advisory condition.
To minimize possible sources of error in the experiment runs, a straight section of
freeway was chosen to display the five consecutive VMS messages in each module. Markers
were placed upstream from each VMS position. The first marker was placed at the visual sight
distance of the VMS, or where the VMS first came into view, at a distance 2,124 virtual feet.
The second marker was placed 1,600 virtual feet before the VMS. The markers were placed to
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 9
allow the recording of participants’ response times during video playback as the duration from
reaching the marker to giving a response (see Figure 3).
302
303
304
305
306
307
308
309
310
311
A motorcycle was placed in the right lane in the scenario to pace the speed of the
participants’ vehicle. Participants were advised to stay in the right lane and travel at a constant
speed, not passing the motorcycle. The experiment setup did not call for actual merging
maneuvers, so the locations of the passenger car and roadside messages were chosen simply for
subject message identification and ease of understanding. The motorcycle was chosen because it
was short enough to not obstruct the participants’ view of the VMS messages. Route vehicles
were placed in both lanes as well as in the opposite lanes to mimic real driving on a highway.
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
FIGURE 3 Snapshots of the driving simulator during the experiment.
Conducting the experiment
All the experiments were conducted in the Driving Simulation Lab at the University of Rhode
Island. A total of 36 participants, balanced in age and gender, took part in the experiment. A
script was followed to explain the experiment to the participants in a consistent, detailed manner.
Participants were informed that they would be making verbal responses, in the form of a number
(1-3 for DAC 1-3, respectively, and 4 for dummy messages), while approaching roadside
portable VMSs in the scenario. Participants were provided with a study sheet showing the three
DACs and messages associated with each DAC and were allowed to study the sheet prior to the
experiment. The participants were then briefed about the driving simulator and given a warm-up
run, followed by the experiment. A participant’s driving was recorded by a video camera which
captured both the driving scene and the participant’s responses.
RESULTS AND DISCUSSION
Questionnaire Survey
Nine questions were presented to participants with six of the questions designed to gather ratings
for the individual messages. For each DAC in the questionnaire survey, there were a total of
three questions. The first and second questions asked participants to give ratings for text-versus-
text and graphic-versus-graphic messages, respectively. The third prompted participants to
choose between their highest rated text and graphic signs.
Ratings of Message Signs
Participants rated messages on a 1-5 scale in how well they communicated the associated DAC,
with 5 being the best. The ratings were grouped by several factors, including driving advisory
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 10
conditions (D), message type (M), age group (A), and gender (G). The nature of the 1-5
ascending scale of message rating called for ordinal logistic regression, with p(π
339
340
341
342
j) is the
probability of being at or below rating value j (1-5).
01 2 3 4
()
ln( )
1()
jj
j
pDMAG
p
(1) 343
344
345
346
347
348
349
350
351
352
353
354
355
The goodness-of-fit tests show adequately high p-values for both of the tested methods
(Table 1). The reference levels for this regression are cited in the table, where the coefficients
and p-values show comparisons between these levels for each factor. From the regression output,
we found that both DAC (D) and message type (M) are significant predictors for message sign
rating. The positive coefficients for “Merge to the Right” (MR) and “Continue Travel Normally”
(CTN) indicated that “Zip Merge” (ZM) messages obtained significantly lower message ratings
when compared with MR and CTN. Conversely, the negative coefficient under message type
suggests text messages were rated significantly higher than graphic messages overall.
TABLE 1 Ordinal Logistic Regression Output for Message Ratings in Questionnaire Survey
Factor Level Coeff P-value Link Function: Logit
MR 0.265 0.021 Variable: Sign Ratings (1-5 Scale)
CTN 0.408 0.000 Value Count
DAC
ZM Reference Level Worst 1 276
2 159
Text Reference Level 3 318
Message Graphic -0.358 0.000 4 272
Best 5 433
18-25 0.162 0.138
Responses
Total 1458
26-40 0.151 0.227
Age
41+ Reference Level Goodness-of-Fit Tests
Method Chi-Sq DF P-value
Female Reference Level Pearson 25.9309 29 0.629
Gender Male 0.082 0.382
Deviance 30.2912 29 0.400
356
357
358
359
360
361
362
363
364
365
* Significance level of α = 0.05
Preferences towards Text and Graphic Signs
Three questions, one for each DAC, were designed to determine whether text or graphic
messages were preferred. Overall, there was a 79.4% preference towards text over graphic
messages among all participant responses.
A binary logistic regression was performed to analyze the text versus graphic preference
responses in the survey, with p(γ) representing the probability of choosing a text sign message
and driving advisory condition (D), age (A) and gender (G) as the factors in the following model.
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 11
01 2 3
()
ln( )
1()
pDAG
p
(2) 366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
The goodness-of-fit tests shows that the model fits very well with the data set (Table 2),
which is more predictable in the binary case when compared to the ordinal logistic regression fit.
The output shows that age was the only significant factor among tested factors to influence text
preference, where the 41+ age group showed a significantly stronger text preference compared to
the other groups. However, age did not correlate with text preference, as the 18-25 age group had
a slightly greater preference towards text messages than the 26-40 age group. There seemed to be
no effect on message type preference as a result of gender. Similarly, the different DACs did not
seem to impact participants’ preferences of either text or graphic messages. A small sample of
six non-native English speakers showed a less significant difference in preference of 61.1% text
and 38.9% graphic messages.
TABLE 2 Binary Logistic Regression Output for Survey Text versus Graphic Preference
Factor Level Coeff P-value Link Function: Logit
MR -0.151 0.697 Variable: Text Preference
CTN 0.252 0.539 Value Count
DAC
ZM Reference Level Text 1 193
Graphic 0 50
18-25 -1.055 0.022
Responses
Total 243
26-40 -1.624 0.001 Goodness-of-Fit Tests
Age
41+ Reference Level Method Chi-Sq DF P-value
Pearson 0.9570 12 1.000
Female Reference Level Deviance 0.9458 12 1.000
Gender Male -0.115 0.727
* Significance level of α = 0.05
Driving Simulation Experiment
Six modules with five messages each were created for the driving simulation experiment.
Response time for each message was recorded as the duration from subject’s vehicle reached the
second marker till a verbal response was given.
As noted, a visual sight distance for the signs was selected so that responses could not be made
prior to the starting point, producing absent data. A large portion of unusual observations was
due to unusually quick response times for merge to the right lane (row 2) graphic and zip merge
(rows 1 and 2) graphics (Figure 1). Another contributor to unusual observations occurred when
subjects drove much slower than the average subject, producing many of their observations to be
noted unusual. Any subject that produced unusual observations due to consistently slow times
(>20 seconds) had their data removed and were not included in the final 36 subjects.
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 12
Response Time Analysis 395
396
397
398
399
400
*
Analyzing mean response time allowed for an analysis of variance (ANOVA). The same factors
used in the questionnaire survey analyses were used in the ANOVA. In addition to observing
factor significance, the interactions of the main factors (D and M) and the blocking factors (A
and G) were observed using the following model.
*
y
DMDM AGAG
(3) 401
402
403
404
405
Through the ANOVA, all factors and the interactions between main factors (D and M)
and blocking factors (A and G) were found significant with p-values near zero. Figure 4 shows
the factors’ effects and their interactions on response time.
16
15
14
13
12
11
10
Graphic Text
Merge Right
Zip Merge
Continued Travel
Conditon
Driving Advisory
Message Type
P = 0.000
16
15
14
13
12
11
10
Female Male
16-25
26-40
41+
Age
Gender
P = 0.000
Continued TravelZip MergeMerge Right
16
15
14
13
12
11
10
TextGraphic
41+26-4018-25
16
15
14
13
12
11
10
MaleFemale
Driving Adv isory Conditon
Mean Response Time (sec)
Message Type
Age Gender
P = 0.000 P = 0.000
P = 0.000 P = 0.000
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
FIGURE 4 Main effects, blocking effects, and interactions for mean response time.
Participants responded the quickest to the “Merge to the Right Lane” messages among all
the DACs at an average of 12.74 seconds. “Zip Merge” and “Continue Travel Normally”
messages had average response times of 13.18 and 13.97 seconds. Among all tested factors,
message type showed the most significance. Graphic messages were responded to much quicker
than text messages (11.74 versus 14.85 seconds), and this was true for each individual DAC. For
the “Merge to the Right Lane” DAC, graphic sign messages yielded an average response time of
11.10 seconds, whereas text sign messages yielded responses of 14.38 seconds. In the “Zip
Merge” DAC, graphic and text sign messages showed response times of 10.80 and 15.56,
respectively, the greatest split between graphic and text sign message response times for any
DAC. In the “Continue Travel Normally” DAC, graphic and text sign messages response times
showed the least significant difference at 13.34 and 14.61 seconds.
As participant age increased, so did participant response times. The 18-25, 26-40 and
41+ age ranges had average response times of 11.97, 12.69 and 15.23 seconds, respectively.
Males responded quicker at an average of 12.7 seconds versus females at 13.9 seconds.
Considering the age and gender interaction, 18-25 year old male and females showed similar
response times of 11.9 and 12.0 seconds, respectively. As participants’ ages increased, the
response time gap widened between males and females. Males in the 26-40 age range showed
response times averaging 12.33 seconds, whereas females in the same age range yielded
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 13
responses of 13.06 seconds. In the 41+ age range, males had an average response time of 13.94
seconds, and females showed a significantly longer average response time of 16.52 seconds.
427
428
429
430
431
432
433
434
Response Accuracy Analysis
The accuracy of responses was analyzed using binary logistic regression with the following
model with p(γ) representing the probability of a correct response and the factors of driving
advisory condition (D), message type (M), age (A) and gender (G).
01 2 3 4
()
ln( )
1()
pDMAG
p
(4) 435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
Table 3 shows that the goodness-of-fit tests show that the binary regression model for
response value is adequate, with sufficiently high p-values. Table 3 also shows the significance
of the levels of factors given the chosen reference levels. Age was the only factor not found to be
a significant predictor of response accuracy. DAC, message type and gender were considered
significant with message type being the greatest single predictor of response accuracy by
coefficient. Graphic message signs resulted in a significantly greater accuracy than text
messages. For DACs, zip merge signs elicited much lower accuracy than either of the other two
DACs. Upon further inspection, much of this lower accuracy was due to the response made to
the zip merge-text message, whereas zip merge-graphic messages produced a similar accuracy as
graphic messages in the other two DACs. Females responded more accurately than males in the
driving simulation, despite males responding more quickly in the response time analysis.
TABLE 3 Binary Logistic Regression Output for Response Accuracy in Driving Simulation
Factor Level Coeff P-value
MR 0.985 0.005 Link Function: Logit
CTN 1.072 0.003 Variable: Correct Responses
DAC
ZM Reference Level Value Count
Correct 1 808
Text Reference Level Incorrect 0 56
Message Graphic -1.424 0.000
Responses
Total 864
18-25 0.195 0.59 Goodness-of-Fit Tests
26-40 -0.115 0.734 Method Chi-Sq DF P-value
Age
41+ Reference Level Pearson 25.9309 29 0.629
Deviance 30.2912 29 0.400
Female Reference Level
Gender Male -1.117 0.000
451
452
453
454
455
* Significance level of α = 0.05
Discussion
Participants showed an almost 4-to-1 preference towards text as opposed to graphic sign
messages in the survey, but responded much faster and more accurately to graphic sign messages
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 14
in the simulation experiment. These results may reflect the unlimited time available to
participants when conducting the survey, whereas time was limited in the driving simulation. In
addition, participants’ preferences towards text sign messages may be influenced by a built-in
bias due to the majority of current signage being text-based.
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
In the survey, the “Zip Merge” DAC messages received the lowest ratings. In the driving
simulation experiment, “Zip Merge” text sign messages were responded to the slowest and with
the lowest accuracy, but the “Zip Merge” graphic sign messages were responded to the quickest
among all the combinations of DACs and message type, while having a comparable accuracy
level to other graphic DAC messages.
In the simulation alone, participants responded much quicker and far more accurate to
graphic sign messages than text sign messages. The effect of message type was most evident in
the “Zip Merge” DAC, where graphic sign messages were responded to much quicker and much
more accurately than text sign messages.
While males responded quicker than females, females responded more accurately. Age
was found to be significant for response time, but was not significant in participants’ accuracy in
identifying messages.
A possible source of error for comparing the results between survey and simulation lays
in the visual message displays. Survey messages on a computer screen being viewed over an
unconstrained time period were compared to simulation messages that were viewed while
driving. While the limited time aspect is embedded into the driving simulation experiment
design, there can be investigations into driving simulator screens as to whether it may be easier
to identify graphic over text messages due to the complex nature of many lines for text messages.
Identification may also be explained by the need for more of a spread in dimensions typically
with text messages, which has defined a one-pixel thickness in standardized lettering, whereas
graphic messages typically have the ability to be created with a greater thickness.
The higher text message preference in the survey and the faster graphic message response
in the simulation are worth much attention. This, with the added result of zip merge-text
combinations being so unidentifiable in simulation, leads belief of a comfort zone existing with a
text bias. This would be further supported by the notion that text sign messages contribute to the
greater majority of road advisory signs in Rhode Island today. Zip merging concepts is widely
unknown and has not been formally introduced to the public. With the graphic messages having
superior results to text messages, it can be speculated that over time, with introduction and
implementation, the proposed zip merge messages and other graphic messages may alter the
public’s preferences towards sign messages. A possible shift in preference is supported by results
in the survey showing the greatest text bias observed in the oldest age group, where age showed
to be the only significant predictor of preference.
CONCLUSIONS AND FUTURE WORK
This study identifies effective sign messages that could be displayed on portable VMSs to help
mitigate bottleneck issues observed at work zones. The survey results showed that most of
participants preferred text sign messages. Driving simulation experiments revealed, however,
drivers responded far more quickly and accurately to graphic sign messages. As the driving
simulation experiment was designed to mimic real life driving, without the added safety risks
associated, it should be held at a higher weight than the survey findings.
The results showing graphic sign messages’ effectiveness were most supported by the
proposed “Zip Merge” DAC, where graphic sign messages outperformed text sign messages.
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 15
This is an indication that graphic sign messages are more effective in painting the picture in
drivers’ minds as to how they should behave in unfamiliar driving conditions, and future study
may investigate the effectiveness of text versus graphic sign messages for newly introduced
driving conditions and the support of the assumption that many people may be visual learners.
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
526
527
528
530
531
533
535
536
537
539
540
542
543
545
546
The results lead to recommend the use of graphic sign messages in work zones, to
mitigate the bottleneck issue, under the three tested work zone DACs of “Merge to the Right
Lane,” “Zip Merge” and “Continue Travel Normally.” The optimal configuration at lane-
reduced work zones may include multiple or multi-frame messages including a simple, non-
confusing text sign message to aid the graphic message to convey the intended behavior and
reduce traffic congestion.
The next stage of this research will involve field tests and deployment of the Portable
Dynamic Lane Merge System, developed by the Czech Republic researchers, at RI work zones.
This field study will test the recommended text and graphic sign messages from the above-
mentioned study and provide a comparison between the DLMS and the traditional MUTCD
setup in traffic management in lane-reduced work zones.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the Rhode Island Department of Transportation, the
Federal Highway Administration, the Czech Republic Transportation Research Center, the
University of Rhode Island Transportation Center and the University of Puerto Rico (Mayagüez)
for their support and guidance on this research project.
REFERENCES
1. Carlson, P.J, M.D. Fontaine and H.G. Hawkins Jr. Assessment of non-recurrent traffic 525
congestion caused by freeway work zones and its statistical analysis with unobserved
heterogeneity. In Transport Policy, Vol. 18, No. 1879-1, Transportation Research Board of
the National Academies, Washington, D.C., 2000, pp. 587-594.
2. Al-Kaisy, A.F. and F. Hall. Guidelines for Estimating Freeway Capacity at Reconstruction 529
Zones. In Journal of Transportation Engineering (ASCE), Vol. 129, No. 5,Transportation
Research Board of the National Academies, Washington, D.C., 2002, pp. 572-577.
3. Schrank, D.L. and Lomax, T.J. The 2007 Urban Mobility Report. Publication DTRT06-G-532
0044. US Department of Transportation, 2007.
4. Chitturi, M.V., and R.F. Benekohal. Effect of Work zone length and speed difference 534
between vehicle types on delay-based passenger car equivalents in work zones. Proceedings
of the 87th Annual Meeting of the Transportation Research Board, Washington, D.C., 2008,
No. 08-3021, pp. 1-13.
5. Chin, S.M., O. Franzese, D.L. Greene and H.L. Hwang. Temporary Losses of Highway 538
Capacity and Impacts on Performance. Report Number ORNL/TM-2002/3, US Department
of Energy, 2002.
6. U.S. Department of Transportation, Federal Highway Administration. Moving Ahead: The 541
American Public Speaks on Roadways and Transportation Communities, February 2001.
http://www.fhwa.dot.gov/reports/movingahead.pdf. Accessed June 19, 2011.
7. Carlson, P.J., M.D Fontaine and H.G. Hawkins Jr. Evaluation of Traffic Control devices for 544
rural high-speed maintenance work zones. Report Number FHWA/TX-00/1879-1, Texas
Department of Transportation, 2000.
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 16
8. U.S. Department of Transportation, Federal Highway Administration Safety Program. Work 547
Zone Safety Fact Sheet, http://safety.fhwa.dot.gov/wz/facts_stats/, Accessed June 17th, 2011. 548
550
551
552
554
555
557
558
559
561
562
564
565
566
568
569
570
572
573
575
577
578
579
581
582
583
585
586
587
589
590
9. Beacher, A.G., M.D. Fontaine, and N.J. Garber. (2005) Field Evaluation of Late Merge 549
Traffic Control in Work Zones. In Transportation Research Record: Journal of the
Transportation Research Board, Issue 1911, Transportation Research Board of the National
Academies, Washington, D.C., 2005, pp. 33-41.
10. McCoy, P.T., and G. Pesti. Dynamic Late Merge Control Concept for Work Zones on Rural 553
Interstate Highways. Transportation Research Record, No. 1745, Transportation Research
Board of the National Academies, Washington, D.C., 2001, pp. 20-26.
11. Grillo, L.F., T.K. Datta, and C. Hartner. Dynamic Late Merge System at Freeway 556
Construction Work Zones. Transportation Research Record: Journal of the Transportation
Research Board, Issue 2055,Transportation Research Board of the National Academies,
Washington, D.C., 2008, pp. 3-10.
12. Datta, T.K., K.L. Schattler, P. Kar and A. Guha. Development and Evaluation of an 560
Advanced Dynamic Lane Merge Traffic Control System for 3 to 2 Lane Transition Areas in
Work Zones. Report No. RC-1451. MDOT, U.S. Department of Transportation, 2004.
13. URS. Evaluation of 2004 Dynamic Late Merge System for the Minnesota Department of 563
Transportation. Minnesota Department of Transportation,
http://www.dot.state.mn.us/trafficeng/workzone/doc/2004DLMS-Evaluation.pdf, Accessed
July 7, 2011.
14. Idewu, W.A., and B. Wolshon. The Joint Merge and its Impact on Merging Speeds in 567
Construction Zone Lane Reduction Areas. In Transportation Research Record: Journal of
the Transportation Research Board, Issue 2169, Transportation Research Board of National
Academies, Washington, D.C., 2010, pp. 31-39.
15. McCoy, P.T., and G. Pesti. Dynamic Late Merge Control Concept for Work Zones on Rural 571
Freeways. Proceedings of the 80th Annual Meeting of the Transportation Research Board,
Washington, D.C., 2001, pp. 20-26.
16. Hit Hofman S.R.O. User Manual and Technical Specifications. Warning Trailer ZSP-1.1 574
LED, http://www.hit-hofman.cz/main_ram.htm, Accessed July 8, 2011.
17. Armstrong, J.D. and Upchurch, J.E. Human Factors design considerations for Variable 576
Message Freeway Signs. In Journal of Transportation Engineering, Vol. 120, Issue 2,
Transportation Research Board of the National Academies, Washington, D.C., 1994, pp. 264-
282.
18. Wang, J.H, and Cao, Y. Assessing Message Display Formats of Portable Variable Message 580
Signs. In Transportation Research Record: Journal of Transportation Research Board ,
Issue 1937, Transportation Research Board of National Academies, Washington, D.C.,
2005, pp. 113-119.
19. Song, T.J., C. Oh, T. Kim and J.Y. Yeon. Estimation of Legibility Distance for Portable 584
Variable Message Signs. In Journal of the Eastern Asia Society for Transportation Studies,
Vol. 8, Transportation Research Board of the National Academies, Washington, D.C., 2010,
pp. 1609-1620.
20. Wang, J.H., C.M. Yang, D. Waters, and C.E. Collyer. Effects of Message Display on 588
Motorist Response to Dynamic Message Signs. Proceedings of the 85th Annual Meeting of
the Transportation Research Board, Washington, D.C., 2006, No. 06-0902, pp.1-25.
TRB 2012 Annual Meeting Paper revised from original submittal.
Messina, Song, Ortiz-Varela, Wang 17
21. Plummer, R.W., J.J., Minarch, and E.L. King. Evaluation of Driver Comprehension of Word 591
versus Symbol Highway Signs. Proceedings of the 18th Annual Meeting of the Human
Factors Society, Santa Monica, October, 1974, pp. 1-7.
592
593
595
596
597
599
600
601
603
604
605
607
608
610
611
22. Wang, J.H., S. Hesar, and C. Collyer. Adding Graphics to Dynamic Message Sign Messages. 594
In Transportation Research Record: Journal of the Transportation Research Board, Vol.
2018, Transportation Research Board of the National Academies, Washington, D.C., 2007,
pp. 63-71.
23. Conference of European Directors of Roads, Action FIVE. Framework for harmonized 598
implementation of Variable message signs in Europe. Draft amended version 3.5. 2003.
http://www.esafetysupport.org/download/documents/Harmonisation_VMS_Systems.pdf,
Accessed on July 8, 2011.
24. Lucas, A., A. Arbaiza, and M.T. Blanch. New Steps towards an International Symbolic 602
Language: Some Results from Spain. 13-15 Sept. 2005. Proceedings of the 8th International
IEEE Conference on Intelligent Transportation Systems, http://ieeexplore.ieee.org, Accessed
July 7, 2011.
25. Ullman, B.R., N.D. Trout and C.L. Dudek. Use of Graphics and Symbols on Dynamic 606
Message Sign: Technical Report. Publication FHWA/TX-08/0-5256-1. FHWA, TxDOT,
U.S. Department of Transportation, 2009.
26. Wang, J.H., M. Keceli, and V. Maier-Speredelozzi. Effect of Dynamic Message Sign 609
Messages on Traffic Slowdowns. Proceedings of the 88th Annual Meeting of the
Transportation Research Board, Washington, D.C., 2009, No. 09-1964, pp. 1-16.
TRB 2012 Annual Meeting Paper revised from original submittal.
