International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

DOI: [Link] UDC: 656.13.078

ASSESSING AND MITIGATING THE IMPACTS OF DYNAMIC

MESSAGE SIGNS ON HIGHWAY TRAFFIC
Miao Song1, Jyh-Hone Wang2, Sam Cheung3, Merve Keceli4
1
Virginia Tech Transportation Institute, Blacksburg VA, USA
2
Department of Mechanical, Industrial and Systems Engineering, University of Rhode Island, Kingston RI, USA
3
U.S. Coast Guard Research and Development Center, New London CT, USA
4
Google, Mountain View, CA, USA
Received 5 January 2016; accepted 29 February 2016

Abstract: As a critical component in Intelligent Transportation Systems (ITS) of modern

traffic management, dynamic message signs (DMS) are widely used in many countries as
an effective means to provide motorists with up-to-date information regarding accidents,
congestion, road conditions and travel time, etc. However, it was observed in traffic data that
motorists tended to slow down when approaching active DMSs and speed up after passing
these signs. The speed variations could pose safety hazards to other motorists on highways.
To gain insights into this issue and understand associated causes and risks, a human factors
study was conducted to help ease the speed variation when approaching DMSs. The effect of
DMS messages on traffic approaching and passing the signs were assessed by analyzing highway
traffic data near DMSs. A questionnaire survey was conducted to identify associated causes
and risks. Participants were further surveyed about their preferences on DMS characteristics
such as message category, message type, number of frames, message details, and the use of
graphics. Individual drivers’ responses to various DMS designs were evaluated through a
simulation test. This study has provided updated knowledge regarding the effects of DMSs
on highway traffic and has identified measures to improve the design and display of DMS
messages to help ease the speed variation of approaching traffic.

Keywords: dynamic message sign, variable message sign, ITS, speed variation, driving
simulation, graphic-aided message.

1. Introduction read and comprehend presented information

on the sign with limited time. As a result,
Dynamic message signs (DMS) or variable drivers are likely to slow dow n when
message signs (VMS), have been widely approaching active DMSs and speed up after
used as a critical component in Intelligent passing them. Speed variations in a short
Transportation Systems (ITS) of modern period of time could introduce safety hazards
traffic management. In Rhode Island alone, on highways. This study intends to assess the
13 overhead dynamic message signs are impact of active DMSs on highway traffic and
currently in service on major highways. explore potential countermeasures.
W h i le DMSs are prov id ing rea l-ti me
information and assistance to drivers, they I n for mat ion presented on DMSs ca n
demand drivers’ attention off the road. genera l ly be c ategor i z ed i nto t h ree
Moreover, drivers often felt compelled to types (CEDR, 2003): Danger/Warning
1
Corresponding author: msong@[Link]
1
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

messages (accidents, disabled vehicles, larger deviation in speed, implying that

unplanned lane/tunnel/bridge closures drivers tend to pay more attention to new
e t c .), I n f o r m a t i v e/C o m m o n R o a d information posted on the sign, and as a
Conditions messages (roadwork closures, result, more time is needed to process the
delays, congestions, travel time etc.), and information. Moreover, when a DMS is
Regulatory/Non-Traffic Related messages displaying the same message for a long period
(work zone speeds, seatbelt use, cell phone of time, drivers become familiar with it and
regulations, motorcycle awareness, etc.). In thus less time is needed to read it (Boyle
order to broadcast important/critical traffic and Mannering, 2004). However, extending
information and advisories to drivers in a the display time of a DMS message, when it
timely manner, DMSs are designed to attract is no longer relevant, can cause confusion
drivers’ attention from driving by using to drivers as well as a loss of public trust in
overhead/roadside display, bright color and/ the intelligent transportation system. In
or dynamic messages. It has been found that order for DMSs to be used effectively by
the majority of drivers worldwide notice the transportation authorities, DMS messages
existence of DMSs, identify that DMSs are should only provide updated information to
providing very useful information, and act drivers (Dudek, 2004).
accordingly (Benson, 1996; Bonsall, 1992;
Peng et al., 2004; Tarry and Graham, 1995). Considering their impacts on highway
traffic, DMS messages should provide
S i nc e d r i v e r s a r e e x p e c t i n g u s e f u l meaningful and straightforward messages
information from active DMSs and time that can be read and responded to quickly,
available for them to read the messages is especially during emergencies (Ullman et
limited, they usually slow down to gain extra al., 2005). To mitigate the impacts of active
time to read and comprehend the messages DMSs on highway traffic speed variation, it
(Harder et al., 2003; Wang et al., 2009; Spell is important to improve DMS message design
et al., 2014). To compensate for their speed to result in faster reading time and quicker
reduction for reading messages, drivers speed response. It has been shown in several
up immediately after passing DMSs (Boyle studies that using graphics to help convey
and Mannering, 2004; Wang et al., 2009). traffic information and advisory messages
The consequences of speed variations in a provides many advantages over text-only
short time window could be severe and pose messages (Bruce et al., 2000; Hanowski
a threat to other vehicles because crashes and Kantowitz, 1997; Wang et al., 2007).
are highly correlated to driving speed Graphic-aided messages could be more
(Hauer, 1971; Lave, 1985; Rodriguez, 1990; easily and quickly identified compared to
Solomon, 1964; TRB, 1998). Let alone the text-only messages from a further distance.
distraction itself caused by some DMSs is The use of graphics or symbols on traffic
safety critical (Harder et al., 2003; Harder signs has been widely employed in many
and Bloomfield, 2008). Europea n countr ies to overcome t he
language barriers (CEDR, 2003). Adding
It is arguable that the impacts of DMSs graphics to DMS messages could enhance
on traffic vary by situation. For example, drivers’ understanding of and responses to
when drivers encountered a new DMS those messages and potentially help ease
message, they were more likely to have a aforementioned impacts.

2
International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

2. Study Approaches were displayed. T he actual messages

displayed during this period were all text-
This study was conducted to assess the only messages covering accidents, disabled
impacts of DMSs on highway traffic, identify vehicles, unplanned lane closure, roadwork
the causes of these impacts, obtain drivers’ closures, delays, and reduced speeds. The
preferences regarding DMS messages, and study first compared “pre-display” and “first
ultimately provide improved DMS message 5-min” in a before-and-during analysis.
designs as a solution to mitigate the impacts Speed reduction was reported in ten of the
of DMSs. The study consisted of a three- sixteen cases with four being statistically
prong approach, a traffic data analysis, a significant. Supplementary analysis of the
driver questionnaire survey, and a driving second pair of 5-minute intervals (“last
simulation experiment. A description of each 5-min” and “post-display” ) demonstrated
part of the approach is given below. that speed increases occurred in seven of
the sixteen cases, with two of them being
2.1. Traffic Data Analysis significant. To determine the extent of speed
variation caused by active DMSs, same traffic
To gain an understanding of the impact data were re-analyzed using 30-minute
of DMSs, traffic data were collected and increments in this study. In addition to
analyzed. A traffic study was conducted to average volume (vehicles per minute) and
collect traffic data such as mean volumes average speed (km/h), average vehicle range
and speeds in 5-minute increments. Those (m) was also analyzed in each of the four time
traffic data were gathered by several mobility periods (Fig. 1) for each message display. It
technolog y units (MTUs) near DMSs is hypothesized that when DMS is active,
along I-95 in Rhode Island between June both speed and vehicle range would reduce
1st and June 14th in 2007 when messages while volume would increase.

Fig. 1.
An Illustration of Initiating Stage and Ending Stage Analysis

Two approaches were employed in this the first approach, data were divided into
analysis, one based on the length of DMS two groups according to the length of DMS
display and the other based on the time of display. The two groups were: DMS message
day in which the DMS was displayed. In displays that were active for less than 30

3
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

minutes at a time and DMS message displays drivers’ opinions and preferences regarding
that were active for greater than 30 minutes various DMS message design and display
at a time. In the second approach, data were features. Each question, presented with text
divided into two groups according to the and graphic DMS messages, was designed to
time of display during the day. There were assess a single feature with multiple DMS
message displays that were active during displays as possible choices. These features
“rush hours” (6AM - 9AM and 4PM - 7PM) included message category (danger warning,
and message displays that were active during informative, and regulatory messages),
“non-rush hours” (other time periods during message type (text-only, graphic-aided with
the day). In each approach, two hypotheses full text, and graphic-aided with partial text
using paired t-test were employed. They message), and number of frames used to
were called respectively the “initiating stage display the same message (single frame vs.
analysis” and “ending stage analysis”. two-frame). The questionnaire survey was
designed using Microsoft PowerPoint with
In the initiating stage analysis, data collected Visual Basic macros. Either a single answer
in “pre-display period” (30 minutes before or multiple answers were expected from the
the start of display) and “first 30-min” (30 participants when they were answering each
minutes after the start of display) were question.
compared. In the ending stage analysis,
data collected in “last 30-min” (30 minutes 2.2.2. Participants
prior to the end of display) and “post-display
period” (30 minutes after the end of display) The survey was conducted at multiple public
were considered (see Fig. 1). locations in Rhode Island in order to obtain
a representative sample of the Rhode Island
2.2. Driver Questionnaire Survey driving population. DM V office, mall,
and university campus were among the
To gain an understanding of the impact locations where the survey took place. A
of DMSs from drivers’ perspectives, a total of 150 licensed drivers participated
questionnaire survey was designed and in the survey voluntarily. Among them, 75
conducted. It intended to help identify causes were between 18 and 40 years old, 39 were
of the impacts of active DMSs on highway between 41 and 60, and 36 were older than
traffic and to gain insights about drivers’ 60. There were 71 females and 79 males.
preferences on various DMS message design Age and gender percentages of the survey
and display features. participants followed those of the Rhode
Island population.
2.2.1. Survey Design
2.3. Driving Simulation Experiment
The survey contained 24 multiple-choice
questions designed to collect drivers’ To help mitigate the impact of DMSs by
opinions on slow-downs and DMS. The improving reading and comprehension time
first six questions were employed to identify of DMS messages, a video-based driving
the causes of slow-downs in general and simulation experiment was developed and
those associated with DMS designs. The conducted to evaluate various DMS message
remaining questions were used to assess designs in a simulated driving environment.

4
International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

2.3.1. Driving Simulation Video and superimposed to fit onto the blank DMS
board in the baseline video frame by frame,
A 30-second baseline video was recorded while and then rendered back into test videos (one
the subject vehicle was driving southbound video per message). In a completed test video,
on RI Rte. 4 at approximately 80 km/h (50 a DMS message first appeared as a small dot
mph). In this video, a blank overhead DMS and gradually increased in size as the vehicle
board (Fig. 2) was approached from about approached the “active” DMS. The message
700 meters away and passed. During post would be visible from around 800 meters and
process, all selected test messages were resized become legible within 200m (MUTCD, 2009).

Fig. 2.
The DMS on RI Rte. 4 Recorded in the Baseline Video
Source: Google Map

2.3.2. Experiment Design factor, governs how the message was displayed.
There were three settings, a graphic-aided
Two main factors, message category and message with full text (GFT), a graphic-aided
message type, were considered in a factorial message with partial text (GPT), and a text
experiment with three repetitions. Message message with no graphic (T). Each of three
category contained three groups according to message categories was included with three
their contents: danger warning, informative, message types (GFT, GPT, and T), creating a
and regulatory. Message type, the other main total of nine test messages (see Fig. 3).

Fig. 3.
Nine Test Messages in the Driving Simulation Experiment

5
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

The message contents were directly adopted response time was defined as the duration
from the actual messages presented during between the start of the video and the
the traffic data collection period and the moment when a response button was pressed.
message signs were automatically generated
using the VanGuard control software package 2.3.3. Participant and Apparatus
by Daktronics, a DMS manufacturer. It
should be noted that text which appeared A total of 36 drivers with valid driver’s
in GFT and GPT messages was single stroke l icen ses pa r t ic ipated i n t he d r iv i ng
but appeared as double stroke in T messages. simulation test. They are evenly distributed
in age and gender groups with six males
To control the lear ning ef fect, three and six females in each of the three age
additional messages (irrelevant, one in groups. Fig. 4 shows a schematic view of
each message ty pe) were added to the the laboratory setup for the test. Participants
message collection. Each message had three sat in a fix-base vehicle (Ford Taurus) and
repetitions totaling 36 separate trials (9 x 3 + were asked to respond by pressing one of the
3 x 3 = 36) which were presented in a random four response buttons which were mounted
order. There are two blocking factors, age on the steering wheel. Participants were
(three groups: 18-40, 41-60, and above 60) instructed to press associated buttons as
and gender (male, female) considered in the soon as they comprehended the messages
experiment. (“1” if a roadwork message was observed,
“2” for a crash message, “3” for a slippery
A few practice trials were given to ensure road message, and “4” if other messages were
participants’ readiness prior to the actual observed). Simulation videos were projected
simulation. Both response time and accuracy onto the screen via a digital computer
were collected in the experiment where projector (see Fig. 4).

Fig. 4.
Setup of the Simulation Test (Left - Schematic View, Right - Actual Picture)

6
International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

3. Results observed in all the cases of the “Ending stage

analysis” evidenced by a significant increase
3.1. Traffic Data Analysis in the average speed. As for DMS message
displays that were active for greater than 30
Table 1 shows the results of the traffic data minutes, drivers decreased their speed and
analysis in both approaches in terms of speed, headway distance as they approached active
volume, and headway distance. No significant DMSs in the “initiating stage analysis” and
difference was observed in average volume increased their speeds and headway distances
and headway distance when comparing an in the “ending stage analysis.” This was also
active DMS with an inactive one. But average observed when DMS messages were active
speed was significantly lower when DMS was during non-rush hours. The decrease in
active. It also found that speed variations both speeds and headway distances in the
in traffic were not noticeable when the “initiating stage analysis” and the increase
DMS messages were displayed less than 30 in both speeds and headway distances in the
minutes as well as when DMS messages were “ending stage analysis” provided evidence
displayed during rush hours in the “initiating of the impacts of active DMSs on highway
stage analysis”. The speed variations were traffic.

Table 1
Summary of Paired t-tests for Initiating Stage and Ending Stage Analyses
Approach Length of display Time of day
Group <30 Min ≥30 Min Rush Hours Non-rush Hours
Traffic Data Spd Vol Hdwy Spd Vol Hdwy Spd Vol Hdwy Spd Vol Hdwy
Sample size N = 22 N = 23 N = 10 N = 32
Period 1 58.52 64.4 295.1 60.14 48.4 563.0 55.89 74.4 212.0 60.77 52.2 487.0
Initiating
Period 2 60.67 63.0 301.8 59.46 45.4 539.0 61.04 72.8 248.9 60.19 50.6 469.0
Stage
P-value 0.906 0.781 0.709 0.078* 0.987 0.313 0.924 0.952 0.979 0.060* 0.910 0.296

Period 3 61.13 63.0 303.9 59.23 47.6 596.0 62.06 69.0 253.4 59.84 52.0 509.0
Ending
Period 4 61.91 62.2 307.6 60.24 48.4 658.0 63.11 67.8 266.9 60.78 53.0 544.0
Stage
P-value 0.020* 0.358 0.386 0.039* 0.767 0.135 0.065* 0.286 0.131 0.016* 0.829 0.193
* Significant @ α = 0.1

3.2. Driver Questionnaire Survey slow down when approaching an active DMS,
33% indicated that they always slowed down,
When participants were asked about the 57% slowed down sometimes, and 10% never
general causes of slow-downs on highways, slowed down. W hen cross-examined by
t he top c hoice s were weat her, poor age and gender groups, it found that elder
visibility, roadwork, crashes, police vehicles, drivers slowed down most of the time while
congestion, and DMSs. When asked about the majority of young drivers slowed down
which traffic signs they were likely to ignore occasionally. For those who slowed down,
while driving on highways, temporary road 65% indicated that they slowed down since
signs were picked over active DMSs. When they need more time to read the messages.
participants were asked whether they would For those who did not slow down, 77%

7
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

indicated that they did not read the DMS and more quickly from a further distance.
messages because they were distracting. The survey also indicated that participants
Answers obtained from the above questions preferred a single-frame message (58%)
indicated that most drivers paid attention to over a two-frame message (42%) if the
DMSs while driving and active DMSs did same contents were displayed. While most
cause speed variations. participants preferred a two-frame message
(59%) over a single-frame (41%) if the two-
When asked about the issues associated with frame message gives more information and
DMS designs that could cause difficulty uses no abbreviations whereas the single-
in their reading and understanding of the frame message contains less information
messages, the top choices were lengthy and uses abbreviations.
messages, complexly worded messages,
abbrev iated messages, and unfamiliar To summarize, the majority of drivers
me s s age s . T he y a l so i nd ic ated t h at i nd ic a t e d t h a t t h e y w o u l d r e d u c e
poor visibility, poor weather conditions, speed when approaching active DMSs.
distractions on the road, and driving at high Furthermore, lengthy, complexly worded,
speed could make it difficult to read and and abbreviated messages could cause
understand DMS messages. The majority drivers to slow down. It was also found
of participants (46%) indicated that danger that drivers preferred text-only messages
warning messages received most of their over graphic-aided messages and single
attention, followed by informative (30%) frame over two-frame messages if they
and regulatory messages (24%). displayed the same content. Finally, two-
frame messages with more information
When participants were asked about their were preferred over single-frame messages
preferences on message types, text-only with less information and abbreviations.
messages were preferred (44%) over graphic-
aided messages with full text (34%) and 3.3. Driving Simulation Experiment
graphic-aided messages with partial text
(22%). Text-only messages were particularly Among the 972 trails taken by 36 participants
preferred by the majority of participants in where each has made responses to 27
the middle age group (41-60) and old age messages (9 test messages x 3 repetitions),
group (above 60). 84 trials yielded inaccurate responses (91.4%
accuracy rate). The accuracy data was broken
Young participants (18- 4 0), however, down by message category and message type
preferred graphic-aided messages (42%) and showed in Fig. 5. With graphic-aided
over text-only messages (32%) and graphic- messages (GFTs and GPTs), participants’
aided messages with partial text (26%). response accuracy rate remained consistently
This finding is at odds with the common above 90%. Response accuracy rate of text
understanding that graphics are better than only messages (Ts) showed greater variation
text because they can be identified easier among different message categories.

8
International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

Fig. 5.
Response Accuracy in the Simulation Test

Participants’ response time to DMS messages compared with text-only messages but its
in the simulation test were collected and response time was much shorter than that of
analyzed. Through the analysis of variance graphic-aided message with full text (GFT).
(ANOVA), it found that both main factors, This might be due to the fact that GFTs had
message type (T) and message category (C); more information on them and thus required
and blocking factors, age (A) and gender (G), longer response time.
were significant (see Table 2) in affecting
drivers’ response time. The interaction Overall, young participants (age 18-40)
between age and gender was significant but responded the fastest among their peers;
not the interaction between message type fema le pa r ticipants responded faster
and message category. than males except for the old age group;
participants responded faster to danger
Graphic-aided messages with partial text warning messages than to other message
(GPT) had similar response time when categories. Fig. 6 shows these results.

Table 2
ANOVA Table of Response Time in the Simulation Test
Source DF Seq SS Adj SS Adj MS F P
T 2 65.084 66.543 33.271 15.78 0.000*
C 2 35.656 35.265 17.633 8.36 0.000*
T*C 4 4.154 4.154 1.038 0.49 0.741
A 2 683.232 682.895 341.447 161.97 0.000*
G 1 27.936 25.126 25.126 11.92 0.001*
A*G 2 56.607 58.596 29.298 13.9 0.000*
Error 855 1802.41 1802.41 2.108
Total 868 2675.07
* Significant @ α = 0.05

9
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

22.5 22.5

Response Time (sec)

Response Time (sec)
22 22
21.5 21.5
21 21
20.5 20.5
20 20
19.5 19.5
GFT GPT T Danger Informative Regulatory
Message Type Warning
Message Category
22.5

22
Response Time (sec)

21.5 Female

21 Male

20.5

20

19.5
18-40 41-60 60+
Age

Fig. 6.
Main Effect Plots (Top) and Interaction Plot (Bottom) of the Simulation Test

4. Discussion findings of the survey first indicated that the

majority of drivers would reduce their speed
A traffic data analysis, a driver questionnaire when approaching active DMSs, especially
survey, and a driving simulation experiment elder drivers. Lengthy, complexly worded,
were employed in this study to assess the and abbreviated messages could cause drivers
impacts of active DMSs on highway traffic to slow down.
and identify potential means to mitigate
the impacts. In order to mitigate the impacts of active
DMSs on traffic, DMS message design
The speed variation caused by active DMSs needs to be improved to produce faster
was confirmed by the traffic data analysis. reading and quicker response. Through
Significant speed changes of traffic were survey, it found that drivers preferred text-
observed regardless of the length of DMS only messages (Ts) over graphic-aided
display or time of day at which the DMS was messages (GFTs and GPTs). Single frame
displayed. This indicated that drivers did messages were preferred over two-frame
slow down when approaching active DMSs messages when their contents were the
and sped up when they passed the signs. same. Drivers’ actual responses to various
Through the driver questionnaire survey, it DMS messages were tested through the
identified the causes of the speed changes, driving simulation experiment. The results
and obtained participants’ preferences indicated that participants responded faster
on and responses to different DMSs. The to “danger warning” messages, followed by

10
International Journal for Traffic and Transport Engineering, 2016, 6(1): 1 - 12

“informative,” and “regulatory” messages. from the simulation could more accurately
This agreed with the questionnaire survey reflect a driver’s response in real driving. It
where “danger warning” messages attracted is recommended to adopt graphic-aided DMS
most participants’ attention, followed by messages with partial text to mitigate the
“informative,” and “regulatory” messages. impacts of active DMSs on highway traffic.
The results of the simulation, however,
found that participants responded faster Although participants’ response time and
to graphic-aided messages with partial text accuracy were assessed through the driving
(GPT), followed by full text (T) messages, simulation experiment, the fact that they
and graphic-aided messages with full text were not actually driv ing limited our
(GFT). This differs from the conclusions ability to eventually verify the effects of
drawn from the survey when text-only (T) different DMS messages designs on driver’s
messages were the most preferred, followed speed maintenance and distraction. Future
by graphic-aided messages with full text studies are necessary to examine individual
(GFT), and graphic-aided messages with drivers’ driving performance and eye glance
partial text (GPT). when approaching various active DMSs in
naturalistic driving environment.
Although graphics may not be well accepted
by drivers to be used in DMS messages as Follow-up field studies could be conducted to
indicated by the survey, graphic-aided assess the influence of various DMS messages
messages w it h pa r t ia l te x t d id show on overall traffic dynamic. Findings from this
promising results by producing quicker study will be shared with the transportation
response and higher response accuracy in authorities regarding effective means to ease
the driving simulation experiment. the impacts of active DMSs on highway
traffic.
5. Conclusion
Acknowledgements
T h is st udy eva luated t he i mpac ts of
a c t i v e DM S s t h rou g h bot h a s t at ic T he aut hors wou ld l i ke to t han k t he
(survey) environment and a dynamic (real University of Rhode Island Transportation
traffic) environments and explored and Center (UR ITC) and the R hode Island
evaluated possible remedies through a Department of Transportation (RIDOT)
driver questionnaire survey and a driving for their support and guidance in this study.
simulation experiment. Speed variations
were observed in both environments when References
approaching active DMSs. While the survey
results showed that text-only messages (T) Benson, B.G. 1996. Motorist attitudes about content of
were the most preferred by participants, variable message signs, Transportation Research Record,
the study’s driving simulation experiment 1550: 48-57.
suggested a graphic-aided DMS message with
partial text (GPT). This proposed display Bonsall, P.W. 1992. Research methods for the study of
could shorten the response time the most driver response to in-vehicle and roadside guidance-
and possibly help ease the speed changes. methods. In Proceedings of the Sixth World Conference on
It is believed that the results gathered Transport Research, 2515-2530.

11
 Song M. et al. Assessing and Mitigating the Impacts of Dynamic Message Signs on Highway Traffic

Boyle, L.N.; Mannering, F. 2004. Impact of traveler Peng, Z.R.; Guequierre, N.; Blakerman, J.C. 2004.
advisory systems on driving speed: some new evidence, Motorist response to arterial variable message signs,
Transportation Research Part C: Emerging Technologies, Transportation Research Record, 1899: 55-63.
12(1): 57-72.
Rodriguez, R.J. 1990. Speed, speed dispersion, and
Bruce, D.; Boehm-Davis, B.H.; Mahach, K. 2000. In- the highway fatality rate, Southern Economic Journal,
vehicle auditory display of symbolic information. In 57(2): 349-356.
Proceedings of the 14th Triennial Congress of the International
Ergonomics Association and the 44th Annual Meeting of the Solomon, D. 1964. Accidents on Main Rural Highways
Human Factors and Ergonomics Society, 230-233. Related to Speed, Driver and Vehicle, Federal Highway
Administration, U.S Department of Transportation.
Conference of European Directors of Roads (CEDR).
2003. Action FIVE - Framework for Harmonized Spell, B.A.; Ardeshiri, A.; Jeihani, M. 2014. Speed
Implementation of Variable Message Signs in Europe. Pattern Analysis in the Proximity of Dynamic Message
LV-SE2000 0826. Signs Using a Driving Simulator. In Proceedings of
Transportation Research Board 93rd Annual Meeting, No.
Dudek, C.L. 2004. Changeable Message Sign Operation 14-5486.
and Messaging Handbook, Report FHWA-OP-03-070,
Federal Highway Administration, Texas Transportation Tarry, S.; Graham, A. 1995. The role of evaluation in
Institute. ATT development, Traffic Engineering and Control, 36(12):
688-693.
Hanowski, R.J.; Kantowitz, B.H. 1997. Driver memory
retention of in-vehicle information system messages, Transportation Research Board. 1998. Managing Speed:
Transportation Research Record, 1573: 8-18. Review of Current Practices for Setting and Enforcing
Speed Limits - Special Report 254, Effects of Speed,
Harder, K.A.; Bloomfield, J.R.; Chihak, B.J. 2003. The DC, USA.
Effectiveness and Safety of Traffic and Non-traffic
Related Messages Presented on Changeable Message Ullman, B.R.; Dudek, C.L.; Trout, N.D.; Schoeneman,
Signs, Publication MN/RC-2004-27, Minnesota: S.K. 2005. A mber A lert, Disaster Response and
Minnesota Department of Transportation. Evacuation, Planned Special Events, Adverse Weather
and Environmental Conditions, and other Messages for
Harder, K.A.; Bloomfield, J.R. 2008. The Effectiveness Display on Dynamic Message Signs, Federal Highway
and Safety of Traffic and Non-Traffic Related Messages Administration, Texas Department of Transportation.
Presented on Changeable Message Signs-Phase II.
Minnesota Department of Transportation, Minneapolis. Wang, J.H.; Keceli, M.; Maier-Speredelozzi, V. 2009.
Effect of dynamic message sign messages on traffic
Hauer, E. 1971. Accidents, overtaking and speed control, slowdowns. In Proceedings of Transportation Research Board
Accident Analysis and Prevention, 3(1): 1-12. 88th Annual Meeting, 1-16.

Lave, C.A. 1985. Speeding, coordination, and the 55 mph Wang, J.H.; Hesar, K.; Collyer, C. 2007. Adding graphics
limit, The American Economic Review, 75(5): 1159-1164. to dynamic message sign messages, Transportation
Research Record, 2018: 63-71.
The Manual on Uniform Traffic Control Devices. 2009.
Federal Highway Administration, U.S. DOT.

12