CHAPTER ONE

INTRODUCTION

1.1 Background

Several studies have shown that speed is one of the most important factors in road
traffic accidents because higher speed increases the accident risk and outcome severity
(Nilsson, 2004). For instance, when a pedestrian is hit by a motorized vehicle at a
speed of 30 km/h, there is higher probability that the pedestrian will survive compared
to if the pedestrian is hit at a speed of 50 km/h. Therefore, higher speed is linked to
higher accident risk, resulting in higher probabilities of fatal and serious accidents.
Moreover, speed variability has also been considered to increase the accident risk as
well (Figueroa & Tarko, 2005). For instance, speed variability increases the
interactions among drivers which can result in greater lane changing manoeuvres, risky
overtaking decisions or rear-end collisions.

However, the speed and speed variability are inherently associated with driver
behaviour. They are the result of driver decisions combined with the road environment,
traffic conditions and driver preferences. Due to individual driver preferences, the
speed choices of drivers, under the same circumstances, are very heterogeneous.
Furthermore, Elvik (2010) states that drivers’ speed choice is not objectively rational
(i.e., decision taken with incomplete information). The basis of his statement is
described below:

a) Drivers ignore environmental impacts of speed

b) Drivers do not accurately perceive the relationship between speed and travel
time

c) Drivers underestimate the increase in accident risk due to higher speed

d) Drivers underestimate the collision speed when an accident is inevitable.

e) Drivers’ speed preferences are very heterogeneous.

1
An attempt to harmonize the heterogeneous speed choice preferences includes, for
instance, posted speed limits (PSLs) which aim to constrain speed choices to safe
levels depending on road characteristics and environmental conditions. Low PSLs are
typically applied on urban roads due to the conflicts between motorized vehicles and
vulnerable road users (VRUs). Normally in urban areas, the PSL is set to 50 km/h.
Besides, in many countries the PSL on urban roads has not been changed in many
years. Traditionally, in the US a design-safety principle, which uses the 85 th percentile
of drivers’ free-flow speed (FFS) distribution, is applied to select an appropriate speed
limit (Cameron & Elvik, 2010).

The FFS distribution reflects the chosen speed (desired speed) of drivers who are not
hindered by the vehicle in front. Therefore, the FFS distribution reflects the impact of
the road characteristics and traffic regulations on drivers’ preferences. Once observed
or estimated, the FFS distribution can be used for different purposes as described
below.

a) For design purposes, the 85th percentile of the free-flow speed distribution is
used to establish speed limits (Cameron & Elvik, 2010).

b) In simulation models, it is used to fill up the network with vehicles, the so

called “warm up” period (Caliper, 2008; PTV, 2005).

c) In “before-after” traffic safety studies, it is used to estimate the impact of a

specific measure on road safety improvements (Bang et al., 1995; Hydén et al.,
2008; Bang & Silvano, 2012).

d) In performance and capacity analyses, it is used to establish the “ideal” traffic

conditions to

be adjusted by road-traffic factors (Bang, 1994; Bang et al., 1998; HCM, 2010).

2
However, in the last decades, studies have shown that there is an increasing disrespect
towards PSLs around the world (Mannering, 2009). In other words, an increasing
number of drivers are not complying with PSLs and it is being accepted as common
behaviour. This problem is of great concern in the traffic safety community, especially
in urban areas due to the proximity between motorized vehicles and VRUs sharing the
same urban space. Therefore, efforts to lower speed levels and speed variability
through road redesign are underway. “Sustainable Safety” (Wegman et al., 2008) was
coined with the objective to prevent road crashes from happening and whenever this is
not possible, to reduce the seriousness of injuries to a minimum. This is also the final
goal of the Swedish Government stated in the Vision Zero as “Nobody should be
killed or seriously injured in a road traffic accident”. Consequently, in urban areas,
the aim is to reduce speed levels and speed variability to tolerable levels for the human
body.

Technology development is being used as well to sustain speed levels and reduce
speed variability. For instance, there are systems interacting with drivers to keep speed
thresholds such as intelligent speed adaptation (ISA) or active accelerator pedal (AAP)
(Várhelyi et al., 2004). Other systems inform drivers on traffic conditions attempting
to stabilize speed levels and harmonize speed variability. More complex systems share
traffic information among the parties involved. These systems are known as
Cooperative Systems. They include vehicle-tovehicle (V2V), infrastructure-to-vehicle
(I2V), and vehicle-to-infrastructure (V2I) communication systems (Böhm et al., 2009).
Therefore, the use of technology development is valid and useful to maintain speed
levels and reduce speed variability.

1.2 Problem statement

The relationship between speed, speed variability, and traffic safety is a research topic
with multiple dimensions and complex interactions. Researchers have agreed on the

3
fact that speed preferences are heterogeneous and difficult to harmonize. Enforcement
measures are expensive and their influence limited in time and space. Geometric
redesign impacts are local as well. Besides, technological advancements are increasing
the workload on drivers making the driving task more demanding. Consequently,
driver behaviour in terms of speed choice and speed variability need further
investigation to evaluate the performance and safety impacts of traffic management
measures, road geometric characteristics and in-vehicle warning technological
developments.

1.3 Objectives

Large studies were conducted investigating the impact of posted speed limits, on-street
parking, and road capacity in urban areas. Moreover, the impact of driver warning
systems was investigated. The thesis is a collection of articles based on the author’s
participation in these projects with the main objective as described below.

To investigate factors influencing speed choice and resulting safety impacts. More
specifically, the objective is to provide a better understanding of the impact on speed
choice of various factors and their effectiveness.

Studied factors

 Impacts of speed limits.

 Impacts of road characteristics.

 Impacts of in-vehicle warning systems.

In order to achieve the objectives, extensive field measurements were conducted. In

total more than 100 sites in 11 Swedish cities were studied. Each site was surveyed
before and after speed limit changes from 50 km/h to 40 or 60 km/h with one year

4
between the measurements on each site. The data collection was normally performed
mid-block on urban road segments far away from signalized intersections and
roundabouts. Typically, each measurement took place on a weekday from 8:00 to
17:00 hours. Only free-flow speed observations were used for analysis (time headways
> 10 seconds). In Sweden, pneumatic tubes are being used by the Traffic
Administration and Local Agencies to monitor traffic conditions; hence drivers are
used to and aware of the purpose of such devices. A distance of 3.3 meters, calibrated
for Swedish conditions, was used between the tubes. The data were stored in TMS07
data loggers connected to the pneumatic tubes. The loggers recorded axle passage
times (precision of 1 millisecond) from which the following results could be obtained:
vehicle passage time (front axle), speed, travel direction, vehicle type, time headways
and traffic flow.

The analysis aims to investigate the impact of the posted speed limits on speed choice
on urban roads. Moreover, the investigation incorporated the road geometric
dimension to explore trade-offs and interactions with traffic safety. The analysis
explored how the mean free-flow speed and speed variability were influenced by the
speed limit changes and different road characteristics such as link length, land use,
road layout, street function, number of lanes, median, lane width, on-street parking,
parking intensity, sidewalks, bicycle lane, bus stops, presence of heavy vehicles,
number of driveways, traffic flow, etc. The estimation procedure was conducted by
means of linear regression and general statistical analysis.

The analysis aims at exploring the free-flow speed distribution on urban roads,
specifically focusing on time headway thresholds to discriminate free-flow vehicles
from constrained vehicles. In order to determine the free-flow speed distribution and
the probability of being constrained, a probabilistic model was developed with time
headway as the decision variable to establish the threshold between free-flow and
constrained vehicles. It was also assumed that the mean free-flow speed was a function

5
of different road characteristics. The probability to be constrained was modelled by
means of a logit model.

This paper aims to investigate speed behaviour under the impact of infrastructure-to-
vehicle (I2V) cooperative system at the aggregate level. A factorial experiment was
designed with two factors: traffic demand and system penetration. To replicate speed
behaviour with and without the system, speed distributions from a simulator
experiment were used. A motorway of 4 km was built in the VISSIM simulation
software. Indicators such as speed, density, delays and travel times were chosen to
evaluate and compare the motorway performance with and without the system.

1.4 Limitations

The surveyed sites were proposed by the involved municipality authorities with the
aim to reduce or increase PSLs according to pre-specified criteria stated in Rätt fart i
staden (“The right speed in the city”), Vägverket & SKL (2008). Typically, PSLs
reduction or increase was proposed respectively where it was suitable according to
traffic and environmental current conditions. Therefore, random selection with control
group approach was not applied. Besides, the objective to relate speed choice with PSL
changes and road environmental characteristics entailed that urban roads with heavy
traffic were avoided due to our need to observed as many as possible free-flow
vehicles.

Driving simulators are being used to replace or supplement field experiments in order
to overcome research obstacles. For instance, future scenarios, redesign aspects, risky
driver behaviour can be investigated in a safe and controlled environment. However,
these advantages can be offset by poorly calibrated or unrealistic behaviour in
simulators. Even though driving simulators are calibrated and validated, there is
always an embedded uncertainty in their results since drivers may not behave or react
in the same way as they would do in the real world.

6
 CHAPTER TWO

LITERATURE STUDY

Speed Choice

Hypotheses have been put forward to explain the processes behind the driving task. In
the literature these hypotheses attempt to explain how drivers make decisions while
driving. Early studies suggested that drivers attempt to maintain constant levels of
stress along the entire route (Taylor, 1964). Meanwhile, Naatanen & Summala (1974)
postulated that drivers take decisions eliminating any risk perception of an accident.
Researchers later on agreed on the fact that drivers have psychological or motivational
thresholds while driving. There are two main hypotheses behind the driving task:

i) Drivers react to an accident-risk level. The model is called “Risk Homeostasis

Theory” (RHT) introduced by Wilde (1982). The author states that drivers react
according to a risk-level which they attempt to sustain. The risk-level is influenced
by their preferences and experiences on risky situations which is determined by
trade-offs among expected advantages of risky behaviour, expected advantages of
cautious behaviour and expected costs of those behaviours (Winter et al., 2012).

ii) Drivers react to task-difficulty levels which drivers expect to maintain. The
most prominent model is the Task-Capability Interface (TCI) model proposed by
Fuller (2005). The author states that the interaction between task demand and

7
 driver capability determines the different levels of task difficulty. According to the
theory, speed choice is the key factor to maintain difficulty levels under certain
thresholds. Other factors influencing the difficulty level are the road environment
and other road users. Drivers’ capabilities are influenced by training, education
and experience (Winter et al., 2012).

The two approaches above have produced a lot of debate in the scientific community.
For instance, Tarko (2009) argued whether it is risk or difficulty levels which
determine the speed choice. According to the author, it is reasonable to think that
drivers drive as fast as they feel comfortable thus drivers’ speed choice is optimal from
the point of view of driver preferences and their subjective judgement of safety and
speed enforcement. Based on these aspects, the author has put forward a model to
predict speed choice based on utility-based theory. The author defines a trip disutility:

Trip disutility = subjective time value + perceived risk + perceived enforcement

(1)

Therefore, drivers speed preferences are influenced by travel time, attitudes towards
safety, and attitudes towards law enforcement. For instance, depending on the trip
purpose and departure time, the subjective value of travel time for a given driver
changes which might induce the driver to accept higher risks and disregard law
enforcement (e.g. drive faster, violate PSLs).

According to Elvik (2010), there is bipolarity within the driver population which is
composed of two groups (faster and slower driver groups). The author states as well
that the speed coordination between the two groups is difficult since no group can be
better off by changing unilaterally its speed choice (Nash equilibria). Moreover, the
author states that the speed choice might be determined by a small group of drivers i.e.,
depending on how sensitive drivers are; they may change their speed choice and join

8
either the faster or slower drivers group. This sensitivity is related to speed choice
stability i.e., same speed choice under different situations. Haglund & Aberg (2002)
investigated the speed choice consistency on rural roads. The authors found that speed
choice was stable for the same location but not among locations. In other words, if a
driver chooses to drive fast at one location, it is very likely that the driver will do so
repeatedly at that same location.

Speed Limits

Overriding posted speed limits has also been a concern for traffic safety researchers.
Speed behaviour towards speed limits has been thoroughly investigated (Haglund &
Aberg, 2000; Wallén-Warner & Aberg, 2008). Mannering (2009) showed that there is
a trend all over the world to drive faster than the posted speed limits. Williams et al.,
(2006) conducted a study to characterize speeding drivers. The authors defined
speeders as drivers driving 15 mph (24 km/h) above the speed limit and relatively
faster than surrounding vehicles. The results showed that speeding drivers are
characterized as follows: young, male driving a newer vehicle. Furthermore, the use of
GPS on-board units facilitates data acquisition and temporal and spatial aspects of
speeding can be captured. Greaves et al., (2011) state that crosssectional studies do not
present the whole picture on speeding. They used geocoded data from a sample of
more than 130 drivers and self-reported speeding behaviour. Their results revealed that
aggressiveness and excitement were related to speeding, whereas risk aversion was
related to not speeding.

Theories have also been developed to explain why drivers override speed limits. The
theory of reasoned action introduced by Ajzen & Fishbein (1980) was first applied to
explain speed choice. Haglund & Aberg (2000) summarize it as: the “intention” to
perform an act which determines the outcome behaviour. This intention is influenced
by attitudes and subjective norms. Later on, Ajzen (1991) refined the theory of

9
reasoned action and introduced the theory of planned behaviour (TPB). Letirand et al.,
(2005) summed it up as: the speed choice is determined by behavioural intention
which in turns is influenced by attitudes and subjective norms. Attitudes are beliefs
towards, e.g., speed limits and subjective norms are the beliefs or influence of others,
e.g., other drivers’ speed choice or family members. A study was conducted by
Wallén-Warner et al., (2008) where the TPB was applied in Sweden. The authors
found that the most positive belief to exceed the speed limits in urban and rural roads
was: “Makes me follow the traffic rhythm better.” The most common belief to exceed
the speed limit in urban areas was: “Being in a hurry” whereas for rural roads was
“road being good - straight & wide” and “free from vulnerable road users”. More
importantly, the authors state that changing the speeding behaviour is a difficult task
(normative and control beliefs should be changed). Therefore, the authors state that to
change speeding behaviour, the lifestyle of the drivers must be changed. They
concluded that intelligent speed adaptation and in-vehicle warning systems at high
penetration levels might be effective solutions to reduce speed levels and speed
variability.

In-Vehicle Systems

Speed choice can be impacted by in-vehicle systems. For instance, intelligent speed
adaptation (ISA), active accelerator pedal (APP), and warning systems are being used
to influence speed choice. Wallén-Warner et al., (2008) reported a study where ISA
was used by almost 300 drivers for about 6 to 12 months. Results surprisingly showed
that travel times either did not change or were slightly reduced. The conclusion was
that the drivers with ISA drove more calmly but more effectively. Hjälmdahl &
Várhelyi (2004) and Várhelyi et al., (2004) evaluated the effects of APP on driver
behaviour. The results showed that APP has a great potential to reduce speed levels
and speed variability suggesting that traffic is improved as well. The mean speed with
the system showed compliance with the speed limits for all roads considered. Time

10
gaps also increased when driving with the system which can be considered as traffic
safety improvement. The authors warn for behavioral adaptation in the long run. The
driver-acceptance for the system was high especially on urban roads.

Other systems influence speed choice by means of warning messages about oncoming
traffic events. These systems are being investigated to improve traffic safety as well.
Especially on motorways these systems can improve the overall traffic operation and
safety. For instance, dangerous situations can occur when high-speed vehicles
suddenly encounter low-speed vehicles due to traffic events like congestion, accidents,
etc.

Urban Speed Models

What characterizes an urban road is the fact that many types of road users interact with
each other. Therefore, urban road traffic safety is important since vulnerable road users
(VRUs) meet motorized vehicles. Besides, different types of activities take place along
them, namely: pedestrians and bicyclists heading to work, school, shopping,
recreational activities, etc. According to the literature linear regression models are
mostly used to evaluate the influence of different road characteristics and side-friction
events on the mean speed or speed percentile (commonly the 85th percentile) of
vehicles in urban road segments (Ericsson 2000; Aronsson & Bang 2006; Wang et al.,
2006; Hansen et al., 2007). Some of the road characteristics that are being investigated
include:

 Road width

 Number of lanes

 Street function

11
  Median

 On-street parking

 Sidewalk

 Traffic flow

 Intersection density

 Number of driveways

 Pedestrians

 Land use

Typically, the mean speed is modelled as a linear function of several road

characteristics as shown in the equation below:

V β ∑βX ε (2
)

Where:

V = space mean speed or 85th speed percentile

β = constant or intercept

β = regression parameters

X = explanatory variables

ε = error term

Self-explaining roads

12
Self Explaining Roads (SER) is a concept which has been developed during the last
decades which describes the capability of a road to support appropriate driver
behaviour while driving on it. This is basically done through the road design and
traffic management characteristics. In particular, the speed choice must be appropriate
depending on the traffic environment and road conditions. For instance, an urban road
where pedestrians walk along the road must have special features to signal to the
drivers that a lower speed should be selected. A highway must have features to ensure
that high speeds can be used without high risk, Theeuwes & Godthelp (1995). The
SER concept is related or complemented with the Forgiving Roads (FR) concept,
which accounts for the fact that drivers make mistakes. However, those mistakes
should not lead to fatalities or serious injuries, i.e. if a driver makes a mistake while
driving, the road should give the driver a chance to correct it or at least to reduce the
severity of an accident.

Wang et al., (2013) have made an extensive literature review addressing the impacts of
traffic and road characteristics on road safety. Speed, density, flow and congestion
were used as traffic characteristics. Road characteristics included horizontal and
vertical alignment, link length, curvature, and number of lanes. The authors concluded
that the effect of speed on road accidents needs to be further examined; likewise,
traffic congestion and horizontal curvature need further investigation since mixed
effects were found in the literature. Additionally, the authors stated that models other
than the Negative Binomial (NB) model should be applied to understand better the
effect of road accidents e.g. spatial models, random and multilevel parameter models,
etc.

Marshall & Garrick (2011) investigated the impacts of network design on traffic safety
in urban areas. Crash data were collected and spatially correlated to street network
characteristics. Characteristics such as street density, street connectivity and street
network patterns (i.e., neighbourhood types) were considered. The authors applied the

13
negative binomial (NB) generalized linear regression model to the crash data. They
found that the total number of accidents was reduced as intersection density increased.
On the other hand, the number of accidents increased as link connectivity increased.
Similarly, increasing the number of lanes resulted in an increase in the expected
number of crashes. On-street parking was found to increase total number of crashes
and more severe crashes but no significant association with fatalities. The authors put
forward the suggestion to look not only at individual street characteristics but to take
into account the network characteristics as well, i.e. neighbourhood type, in terms of
building a safer and more efficient transport system. Figure 1 shows the spatial
patterns due to neighbourhood and street network characteristics considered in the
study by Marshall & Garrick (2011).

FIGURE 1 Spatial pattern in urban environments.

Another characteristic of urban environments is that economic and human activities

take place along them. Such activities are likely to develop on-street parking,
especially, in city centres. Therefore, many studies have been conducted to investigate
the impacts of on-street parking (Shoup, 2006; Marsden, 2006; Ibeas-Portilla, 2009).

14
On-street parking affects traffic safety as reported in the Handbook of Road Safety
Measurements (Elvik et al., 2004). For instance, parking-related accidents, reported to
the police in Norway during 1991 – 1995, accounted for 2.4% of the total number of
injured accidents. According to the Handbook, the most common accident types where
on-street parking was present are summarized below:

 Hitting parked vehicles (30%)

 Hitting pedestrians crossing the road between parked cars (25%)

 Collision while overtaking parking cars (15%)

 Accidents when leaving the parking position (8%)

The authors compared different studies where a before and after evaluation was
conducted regarding parking policy introduction including time-limited parking
restrictions, banning on-street parking, parking layout and one-side parking. The
authors found that time-limited parking and banning on-street parking appeared to
reduce the number of accidents. Changing from diagonal to parallel parking seemed to
have the same effect. Contrary, one-side parking was found to increase the number of
accidents compared to two-side on-street parking but the effect was not significant.
Furthermore, marking parking places appears to increase the number of accidents as
well.

Accident Risk Models

An extensive literature review was conducted by Vadeby & Forsman (2012) on

accident risk models where speed is related to the accident risk. The authors classified
risk models according to individual and aggregated risk models. Individual models
attempt to predict the individual accident risk based on the individual speed choice
whereas aggregate models aim at predicting the total number of fatal, injured and

15
damage outcomes based on the speed distribution. However, more attention is placed
on aggregate models. A prominent model of this type is Nilsson’s power model. The
author derived the model investigating accidents due to speed changes on rural roads
in Sweden (Nilsson 2004). Basically, the change in the mean speed is the key factor
which predicts the change in the accident risk. However, according to Cameron &
Elvik (2010), Nilsson’s power model is not applicable on urban roads due to the much
higher speed variability in urban environments. Nilsson’s power model has been
derived from rural roads which do not present fast changing conditions on the road
environment. The authors proposed that the mean speed should be supplemented by
the coefficient of variation (c.v.) or other alternatives such as the proportion of
vehicles exceeding the speed limit, the mean speed of vehicles above the speed limit or
the proportion of vehicles exceeding the speed limit by 25 km/h. According to the
authors, these indicators have the same explanatory power for urban roads.
Consequently; the authors proposed an adjustment of Nilsson’s power model to be
applicable to urban roads.

Another approach is the exponential model proposed by Hauer (2009). According to

the author this model fits accident data better due to the fact that the model does not
depend on the initial speed as the power model does. A comparison of both models
was conducted by Elvik (2013), who found very similar results from both models, but
the exponential model fitted the injury data more accurately. Thus author suggests that
the exponential model should be used instead of the power model in urban
environments. However, further analysis and refinement is needed in the relationship
between speed and traffic safety. For the sake of information the power and
exponential models are presented below respectively cited from Elvik (2013).

"# $%&'() -# . .

16
 · !"# $ *(&+)(, (3)

234546789 :⁄ <·=4>546>?@
/01 (4)

Where, AMF is the accident modification factor related to speed change. A and B are
parameters estimated by regression analysis. C is the speed before change and C 6 is the
speed after the change. Figure 2 presents the curve fitting of both models against actual
accident data of fatal accidents and injury accidents.

(a) Power vs. Exponential (fatal accidents) (b) Power vs. Exponential (injury
accidents) FIGURE 2 Power vs. Exponential models (Elvik 2013).

Other accident risk models take into consideration the flow rate and road
characteristics. For instance, Greibe (2003) developed a simple flow function model to
compute the expected number of accidents per km per year as follow:

E8µ< 2.44 H 105KQ.MN (5) Where, Q is the

traffic flow expressed in annual average daily traffic (AADT). Furthermore, the author
developed an accident risk model where different explanatory variables were included

17
such as speed limit, road width, land use, on-street parking, etc. The research was
carried out for urban road links. The accident data was collected from official
statistical databases from the police and the help of Danish Municipalities. The model
is presented below:

E8µ< a8QP< 6 βQ 6 β: 6 βK 6 βR 6 βN 6 βS (6)

Where, the parameters α, p and β ′s are estimated from the data and Q is the average
annual daily traffic (AADT).

CONTRIBUTIONS

The main contributions of the research are summarized below.

3.1 Paper I: The impact of speed limits and road characteristics on drivers’ free-flow
speed in urban areas.

Paper I (Silvano & Bang) describes a follow-up study of changing the posted speed
limit (PSL) in urban areas in Sweden. The study aimed at determining the impacts on
free-flow speed of changing the base PSL of 50 km/h to 40 or 60 km/h. Moreover, the
study aimed at identifying and quantifying road design characteristics which might
influence the drivers’ speed choice. Disaggregate and aggregate levels of analysis were
conducted using crosssectional speed measurements before and after the PSL change.
The study conducted on the disaggregate analysis verified that PSLs had an impact on
drivers’ speed choice. The analysis suggested a speed threshold of 45 km/h to bring in
a speed decreasing effect from PSL reductions.

The results from the aggregate analysis indicate that PSLs had a statistically significant
influence on drivers’ speed choice. More importantly, the paper identified that PSL
changes had greater impacts on higher network classes. For instance, a PSL increase of
10 km/h, brought about a change of 2.3 km/h for arterial roads; whereas, the change

18
was 1.7 km/h in main roads. Additionally, the analysis contributed by identifying and
quantifying road characteristics influencing the speed choice on urban roads such as
link length, area type, network class, carriageway width, on-street parking, sidewalk
and number of exits/entries.

3.2 Paper II: The Free-Flow Speed Distribution on Urban Roads. A probabilistic
approach to model time headway thresholds.

Paper II (Silvano, Farah & Koutsopoulos) describes a methodology to determine the

free-flow speed distribution on urban roads. There is a study conducted by Vogel
(2002) who proposed a 6-seconds threshold to discriminate free-flow from constrained
vehicles; however, it is a fixed threshold approach conducted with data collected in
one single site. As shown from the literature, the threshold is likely to vary depending
on traffic conditions, road characteristics and driver behaviour. The main contribution
of the paper is the probabilistic approach put forward to identify the free-flow speed
distribution on urban roads which takes into consideration a broad range of road
characteristics using midblock data from several sites. Consequently, the methodology
produces different probability curves depending on site characteristics. For instance,
low-speed sites present lower constrained probabilities compared to high-speed sites.
The logit approach for the estimation of the probability the driver is constrained is a
promising approach.

3.3 Paper III: Simulation-based evaluation of I2V systems impact on traffic

performance.

Paper III (Silvano, Farah & Koutsopoulos) describes a factorial experimental design
with two factors: traffic demand and I2V system penetration with three levels each.
The paper contributes to the understanding of the impacts of in-vehicle warning
cooperative systems on traffic performance and safety. Such systems are at a

19
developmental stage and difficult to evaluate in real environments. Therefore,
simulation models are commonly being used. The main contribution of the study is the
evaluation of the impacts of the system penetration levels. For instance, lower
penetration levels resulted in worse traffic performance compared to not having the
system at all. The study also confirmed that congestion drastically reduced the impacts
of the system due to constrained speed choice. Moreover, driving with the system was
characterized by smoother speed decelerations when approaching critical
incident/accident events.

FURTHER RESEARCH

A great body of research has been conducted understanding safety aspects of urban
roads. However, there are still factors and dimensions which need to be further
investigated. Besides, models and methodologies applied to urban roads can also be
improved. The application of the exponential model and power model on urban roads
is still subject to discussion. In other types of accident risk models, the network level
and spatial patterns (i.e., neighborhood types) need to be incorporated.

Speed choice can also be further investigated. For instance, further statistical analysis
is needed to better understand the behaviour of faster drivers (>85 th speed percentile)
or of drivers at higher speeds than the posted speed limit (>50 km/h PSL). Studies
have shown that there are drivers whose speed choice is influenced by the speed of
other drivers (faster – slower driver group), thus further studies must be conducted to
understand drivers’ speed choice in greater detail. ISA, APP, and in-vehicle warning
message systems have potential for keeping speed choice more harmonized and under
certain desired levels.

The methodology put forward to estimate the free-flow speed distribution needs
further investigation as well, such as, its applicability to other types of transportation
facilities, e.g., rural roads, freeways, would be very informative. The use of

20
cooperative systems and their influence on driver behaviour also needs to be better
understood. For instance, analysis at the micro level, e.g., car-following and lane
changing behaviour under the impact of cooperative systems is required.

REFERENCES

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social
behaviour. Englewood Cliffs, NJ: Prentice-Hall.

Ajzen, I. (1991). The theory of planned behaviour. Organizational Behaviour and

Human Desicion Processes, Vol. 50, pp. 179-211.

Aronsson, K.F.M. & Bang, K.L. (2006). Speed characteristics of urban streets based
on driver behaviour studies and simulation including pedestrian movements.
Transport Research Board. 86th Annual Meeting. Paper Number 07-1490.

Bang K.L. & Silvano, A.P. (2012). Utvärdering av nya hastighetsgränser i tätort.
Slutrapport. Traffic and Logistic Division, Royal Institute of Technology,
Stockholm, Sweden.

Bang, K-L, Carlsson A., Palgunadi. (1995). Development of Speed-flow Relationships

for Indonesian Rural Roads using Empirical Data and Simulation.
Transportation Research Record 1484, Transportation Research Board, National
Academy Press, Washington D.C.

Bang, K.L. (1994). Indonesian Highway Capacity Manual Project Phase 2: Interurban
roads. Final Report. General Directorate of Highway, Jakarta.

21
Bang, K-L, Ronggui, Z., Huichen, X. (1998). Traffic performance and passenger car
equivalents for road links and township roads in Peoples Republic of China.
Proceedings of the Third International Symposium on Highway Capacity,
Copenhagen June 1998 (TRB + Danish Road Directorate).

Böhm, M.,Fuch, S., Pfliegl, R., Kölbl, R. (2009). Driver Behavior and User
Acceptance of Cooperative systems on Infrastructure-to-Vehicle
Communication. Transportation Research Record: Journal of the Transportation
Research Board 2129, pp. 136 – 144.

Cameron, M.H. & Elvik, R. (2010). Nilsson’s power model connecting speed and road
trauma: applicability by road type and alternative models for urban roads.
Accident Analysis and Prevention 42 1908-1915.

Caliper (2008). TransModeler. Traffic Simulation Software, User’s Guide. Version

2.0.

Elvik, R., Höye, A., Vaa, T. (2004). The Handbook of Road Safety Measures. First
Edition, Emerald Group Publishing Limited, UK.

Elvik, R. (2010). A restatement of the case for speed limits. Transport Policy 17 196-
204.

Elvik, R. (2013). A re-parameterisation of the Power Model of the relationship

between the speed of traffic and the number of accidents and accident victims.
Accident Analysis and Prevention 50, pp. 854-860.

Ericcson, E. (2000). Variability in urban driving patterns. Transport Research Part D 5

337354.

22
Figueroa, A.M. & Tarko, A.P. (2005). Speed Factors on two-lane rural highway on
free-flow conditions, 84th Annual Meeting of the Transportation Research
Board.

Fuller, R. (2005). Towards a general theory of driver behaviour. Accident Analysis and
Prevention, Vol. 37, pp. 461-472.

Greibe, P. (2003), Accident prediction models for urban roads. Accident Analysis and
Prediction 35 273 – 285.

Hauer, E. (2009). Speed and Safety. Transportation Research Record 2103, 10-17.

Haglund, M., Åberg, L. (2000). Speed choice in relation to speed limit and influences
from other drivers. Transportation Research Part F: Traffic Psychology and
Behaviour, Vol. 3, Issue 1, pp. 39-51.

Haglund, M., Åberg, L. (2000). Stability in drivers’ speed choice. Transportation

Research Part F, Vol. 5, pp. 39-51.

Hansen, G., Garrick, N.W., Ivan, J.N. & Jonsson, T. (2007), Variation in Free Flow
Speed due to Roadway Type and Roadside Environment. Presented at 86th
Annual Meeting of the Transport Research Board, Washington, D.C.

HCM (2010). Highway Capacity Manual, Volume 2: Uninterrupted flow. Transport

Research Board. Washington, D.C.

Hydén, C., Jonsson, T., Linderholm, L. & Towliat, M. (2008). Nya Hastighetsgränsen i
tätort. Bulletin 240. Lund Institute of Technology, Sweden.

Hjälmdahl, M. & Várhelyi, A. (2004). Speed regulation by in-car active accelerator

pedal Effects on driver behaviour. Transportation Research Part F, Vol. 7, pp.
77-94.

23
Ibeas-Portilla, A., Alonso-Orena, B., Moura-Berodia, J.L., Ruisanchez-Diaz, F.J.
(2009). Using M/M/ Queuing Model in On-Street Parking Maneuvers,
Journal of Transport Engineering, Volume 135, Issue 8, pp 527 – 535.

Nilsson, G. (2004). Traffic Safety Dimensions and the power model to describe the
effect of speed on safety. Bulletin 221. Lund Institute of Technology, Sweden.

Näätänen, R. & Summala, H. (1974). A model for the role of motivational factors in
drivers’ decision-making. Accident Analysis and Prevention, Vol. 6, pp. 243-
261.

Mannering, F. (2009). An empirical analysis of driver perceptions of the relationship

between speed limits and safety. Transport Research Part F, Vol. 12, pp. 99-
106.

24