Content uploaded by Francisco J. Martinez
Author content
All content in this area was uploaded by Francisco J. Martinez on Feb 09, 2018
Content may be subject to copyright.
VEACON: a VEhicular ACcident ONtology Designed
to Improve Safety on the Roads
Javier Barrachina, Piedad Garrido, Manuel Fogue, Francisco J. Martinez
University of Zaragoza, Spain
Email: {barrachina, piedad, m.fogue, f.martinez}@unizar.es
Juan-Carlos Cano, Carlos T. Calafate, Pietro Manzoni
Universitat Polit`ecnica de Val`encia, Spain
Email: {jucano, calafate, pmanzoni}@disca.upv.es
Abstract
Vehicles are nowadays provided with a variety of new sensors capable of
gathering information about themselves and from their surroundings. In a
near future, these vehicles will also be capable of sharing all the harvested
information, with the surrounding environment and among nearby vehicles
over smart wireless links. They will also be able to connect with emer-
gency services in case of accidents. Hence, distributed applications based
on Vehicular Networks (VNs) will need to agree on a ’common understand-
ing’ of context for interoperability, and, therefore, it is necessary to create a
standard structure which enables data interoperability among all the differ-
ent entities involved in transportation systems. In this paper, we focus on
traffic safety applications; specifically, we present the VEhicular ACcident
ONtology (VEACON) designed to improve traffic safety. The instances of
our ontology are composed by the information collected when an accident
occurs, and the data available in the General Estimates System (GES) acci-
dents database. We assess the reliability of our proposal using both realistic
crash tests, and vehicular network simulations, based on the ns-2 simulation
tool. Experimental results highlight that both nearby vehicles and infrastruc-
ture elements (RSUs) are notified about and accident in just a few seconds,
increasing the emergency services notification effectiveness.
Keywords: Vehicular Networks, Ontologies, Intelligent Transportation
Systems (ITS), Vehicular Accidents, VANETs
Preprint submitted to Journal of Network and Computer Applications November 25, 2011
1. Introduction
Currently, one of the most important factors of globalization is trans-
portation. Although the purpose of transport has not changed with global-
ization, the factors triggering the emergence of a global transportation system
(e.g. volume, capacity, speed, and efficiency) have evolved. Moving goods
and people as quickly as possible all around the world requires advanced in-
tegrated transportation systems (Di Lecce and Amato, 2009). Information
Technology (IT) and transport infrastructure help to manage transportation
systems in an accurate and effective manner. Intelligent Transportation Sys-
tems (ITS) will play a leading role in our society, especially in scenarios such
as warning drivers about vehicle accidents in real time, efficiently managing
vehicle information required by governments and authorities, or even being
able to offer drivers a variety of added services.
The specific characteristics of Vehicular Networks favor the development
of attractive and challenging services and applications. However, distributed
applications based on Vehicular Networks need to agree on a ‘common under-
standing’ of context for interoperability on a contextual level. We consider
that some semantic web ideas can be applied to modern transportation sys-
tems to build up such context. The Semantic Web (Berners-Lee et al., 2001)
is an extension of the traditional web which allows machines to interpret
the meaning of data thanks to the use of ontologies. An ontology is a de-
scription of a small part of the real world, including the types of items that
appear in this world, the relations among them, the leading elements, and
their restrictions. Typically, an ontology is defined as a formal specification
of conceptualization (Gruber, 1995).
In this paper we focus on safety applications. Specifically, our aim is
to improve traffic safety by using an ontology-based approach in Vehicu-
lar Networks. To that end, we propose the VEhicular ACcident ONtology
(VEACON), a novel lightweight ontology proposed with the aim of success-
fully sharing and reusing knowledge about traffic accidents. VEACON allows
to efficiently structure and encode the information collected by sensors in the
vehicle, enabling the interoperatibility among all the agents involved in mod-
ern ITS (i.e., vehicles, RSUs, emergency services, and authorities).
Nowadays, vehicular networking technologies allow a vehicle to alert emer-
gency services in case of an accident. Although there are many solutions that
2
relay on Vehicular Networks for that purpose, there are fewer solutions based
on semantics to send accident information to the emergency services. In this
paper, we explore the use of a formal ontology framework for sending critical
information captured by vehicles involved in road accidents. This informa-
tion will not only be sent to the emergency services, but also it will be shared
among the nearby vehicles. Hence, this warning information will be used for
different proposes such as: (a) preventing new accidents (avoiding that other
vehicles collide with the vehicles already involved in the accident), (b) help-
ing to allocate resources for a rescue, and (c) maintaining statistics on road
accidents, which allows fast database searches and the creation of prediction
models to estimate the severity of future accidents. This estimation could
be done with data mining classification models by combining the proposed
ontology with existing databases (Chong et al., 2005).
This paper is organized as follows: Section 2 reviews the related work
regarding the use of ontologies applied on ITS. Section 3 presents VEACON,
our proposed ontology. In Section 4, we assess the feasibility our proposal by
doing some real experiments as well as carrying out some simulation tests.
Finally, Section 5 concludes this paper.
2. Related Work
For the proper operation of traffic safety systems, we must consider two
different factors: (i) vehicles must be able to communicate among them in
order to share information, and (ii) the shared information should be under-
stood by all the entities involved in transportation systems. The first fac-
tor has been widely studied by the wireless networking research community
(Martinez et al., 2010b; Bakhouya et al., 2011; Antolino Rivas et al., 2011;
Daeinabi et al., 2011). However, the second factor has not been studied to
the same extent.
Regarding the use of semantics in vehicular environments, some authors
have worked on the integration of transportation systems information and
semantics. Zhai et al. (2008b) presented an ontology for structuring data
traffic. Zhai et al. (2008a) introduced a knowledge navigation system with
urban traffic information based on the XML Topic Maps technology, enabling
intelligent information retrieval through association between topics. These
different works highlight the importance of using ontologies in ITS, however
they do not provide any ontology specially designed for ITS safety.
3
Regarding ITS safety, Eigner and Lutz (2008) showed the need for onto-
logical context models for VNs safety environments, and how all the com-
ponents of the system would be able to understand one another through
these models. They considered that vehicles should incorporate a variety
of sensors to get data from the vehicles themselves, as well as from their
surroundings. In addition, information obtained by these sensors could be
shared with other vehicles using VNs. The authors showed that vehicular
applications can benefit from the inherent characteristics of ontological mod-
els such as distributed composition, partial validation, richness and quality
of information, as well as a certain level of formality. Additionally, authors
proved that calculations on the model are still fast enough to fulfill real-time
requirements imposed by the active safety systems of vehicles. However, they
did not build a specific ontology. More recently, Kannan et al. (2010) pro-
posed an ontology modelling approach for assisting vehicle drivers through
warning messages during time critical situation. Authors focused on gen-
erating the alert messages based on the context aware parameters such as
driving situations, vehicle dynamics, driver activity, and the environment.
Although all the above presented works proposed ontological models for
warning messages using Vehicular Networks, none of them enriched their
proposal with historical information to estimate the severity of accidents.
3. VEACON: Our Ontology for Vehicular Networks
From the point of view of Communications and Information Technolo-
gies for Vehicular Networks, ITS applications will rely on efficient vehicular
communications and smart exchange of information among all the entities
involved, i.e., vehicles, RSUs, emergency services, management authorities
and police. When a traffic accident occurs, a crucial issue that should be
addressed in transportation systems is to collect as much information as pos-
sible, since vehicles should rapidly warn nearby vehicles and the emergency
services to obtain a quick and efficient response from them. However, the
information usually collected in accidents is neither structured nor does it
present relationships between their basic elements. We propose to organize
this information by using an approach based on the Semantic Web, where
the information can be obtained through various techniques such as ontolo-
gies, classifications, taxonomies, thesauri, or Topic Maps (Garshol, 2004).
We consider that the use of ontologies is the more common and versatile
technique to organize such kind of contents.
4
Figure 1: VEACON ontology components.
Our system gets the information from warning messages exchanged among
vehicles and emergency services. This information should be, on the one
hand, concise enough to avoid irrelevant information, but, on the other hand,
it should not ignore any information that might be useful for the emergency
services to determine the most suitable set of resources. Thus, the deliv-
ered information should include: data about the conditions under which the
accident occurred, data about the occupants of the vehicle, as well as a de-
scription of the security systems included within the vehicle. These data will
be sent to the emergency services to provide a more detailed view of the
conditions of the accident before their arrival.
In this work, we use an ontology based technique to group all these in-
formation sources, while allowing to make inferences over the collected data.
5
An ontology formally represents knowledge, and it can be used to reason
about the entities within that domain, and may be used to describe the
domain, so we can elaborate estimations about different factors of the acci-
dent (impact severity, passenger injuries, and so on). Basically, an ontology
consists of three parts: classes and instances of real-world items, relations
among these items, and rules for modeling knowledge and complex behav-
iors (creation, restraint and response). Specifically, we propose the Vehicle
Accident Ontology (VEACON), a novel lightweight ontology proposed with
the aim of sharing and reusing knowledge about the vehicles involved in road
accidents. VEACON meets our requirements since it: (i) promotes interop-
erability between different knowledge bases, (ii) provides an infrastructure or
cooperative system, (iii) facilitates the information sharing, and (iv) enables
domain knowledge reuse. VEACON consists of a set of classes representing
the categories of the entities of interest in the ITS domain, the attributes
which define properties of those classes, and the relationships between those
entities.
Figure 1 shows the basic VEACON lightweight ontology structure, which
groups the available information into four different areas: Vehicle, Accident,
Occupant and Environment. As for the languages, we decided to use the
Ontology Web Language (OWL)1to create XML-based messages, since it is a
flexible and expressive language which provides a basic syntax to describe the
relationships between entities. Listing 1 shows an example of a VEACON-
compliant warning message.
1<?xml ve r si o n=” 1 .0 ” ?>
2<rd f: RD F x ml ns=” h t t p : //www . o wl−o n t o l o g i e s . com/ Ve h i c l e C r a s h . ow l#” . . . >
3<o w l : C l a s s r d f : I D=” Oc cu pa nt ” />...
4<o w l : O b j e c t P r o p e r t y rd f: I D=” t ak e s p l a c e i n ”>
5<r d f s : d o ma i n r d f : r e s o u r c e=”#Ac c i d e n t ”/>
6<o w l : i n v e r s e O f>
7<o w l : O b j e c t P r o p e r t y rd f: I D=” c a n o c u r r ” />
8</ o w l : i n v e r s e O f>
9<r d f s : r an g e rd f : r e s o u r c e =”#E nvi ro nm en t ”/>
10 </ o w l : O b j e c t P r o p e r t y>. . .
11 <Oc cu pa nt r d f : I D=” Oc c up an t 2 ”>
12 <i l l n e s s xml : la ng=” e s ”>No</ i l l n e s s>
13 <pr e g n a n t r d f :d at at yp e=” h t t p: //www. w3 . o r g /2 00 1 / XMLSchema#bo o l e a n ”>t r u e</ pr e gn a nt>
14 <bl o o d x m l: la ng=” e s ”>A+</ b lo o d>
15 <we i g ht r d f : d a t a t y p e =” h t t p : / /www. w3 . o rg /2 0 01 /XMLSchema#f l o a t ”>56 . 4</ w ei g ht>. . .
16 </ Oc cu pa n t>
17 </ rd f: RD F>
Listing 1: Example of a VEACON OWL-based warning message.
1http://www.w3.org/TR/owl-ref
6
Due to privacy requirements of the collected data, especially the medical
information of the passengers, all the messages generated with the VEACON
ontology are encrypted using the Advanced Encryption Standard (AES)
(Daemen and Rijmen, 2002) before being sent by the vehicles.
3.1. Getting information into our Ontology
Nowadays, vehicles incorporate a series of sensors to obtain information
about different areas. Examples of them are crush zone crash sensors (Breed,
1991), occupant position sensors (Breed et al., 1998), rain sensors (Petzold,
1999), or seat belt tension sensors (Husby and Simpson, 2000). Therefore,
it is possible to get key information from these sensors when an accident
occurs.
Furthermore, we consider that future vehicles will get additional informa-
tion from the environment and its occupants, since vehicles will be provided
with sensors capable of knowing if there are pedestrians or cyclists involved in
an accident, and also information regarding the health of the occupants such
as blood group or heart problems. Currently, there are some approaches
addressing these issues; for example, the Ford Motor Company (2011) is
designing seats that can monitor the driver’s heartbeat in real time. For per-
sonal information and health data of the occupants, we consider that each
occupant could have this information on his cell phone, and the vehicle would
collect this data when boarding.
To allow estimating the severity of the accidents, our proposal also uses
the General Estimates System (GES), an historical database maintained by
the (National Highway Traffic Safety Administration (NHTSA), 2011), which
contains information related to previous traffic accidents, obtained from a
sample of Police Accident Reports (PARs) collected all over the USA roads.
To protect individual privacy, no personal information such as names, ad-
dresses or specific crash location is coded.
3.2. VEACON Fields
For our proposed ontology, we selected a number of existing fields in the
GES database, and we have also added others that we felt necessary. The
selection was specifically made considering the data that can be significant
when an accident occurs.
We have grouped the information into four areas: (i) Vehicle, which con-
tains the characteristics of the vehicle and data for identification; (ii) Acci-
dent, which collects the location and time of the crash, the characteristics
7
Table 1: Vehicle dataset.
Field Description
Chassis Vehicle chassis number
Make Manufacturer of the vehicle
Model Vehicle model
Model Year Vehicle model year
Body Type Vehicle body type
Trailer If vehicle is towing trailing units
Num occupants Number of vehicle occupants
Haz Mat If vehicle is carrying hazardous materials
Haz Mat T Hazardous materials type
Emcy use If vehicle is on an emergency run
Spec use Vehicle special use category applied
License Plate Vehicle plate number
of the collision, and the caused damage; (iii) Occupant, which collects occu-
pants’ personal and medical information, their location within the vehicle,
and the safety systems deployed; and (iv) Environment, which contains in-
formation about road, weather and lighting conditions.
In the set of data related to Vehicle, the fields used are those indicated
in Table 1. For this dataset, we used available fields from the GES database,
and added two new fields: License Plate and Chassis. We believe that they
are necessary since they provide a unique identifier (Chassis), and allow the
emergency services to quickly recognize vehicles at the scene of accident
(License Plate).
In the set of data related to Accident, the fields used are those indi-
cated in Table 2. For this dataset, we used fields from the Accident and
Vehicle dataset from the GES database, and added two new fields: Coor-
dinates and Max Acceleration. We consider that they are useful to locate
the crash site (Coordinates), and to obtain a measure of the impact sever-
ity (Max Acceleration). Note that, if using the appropriate technology, the
system can also determine the value of the Non Invl field, which indicates
whether people outside the vehicle (pedestrian or cyclist) were involved in
the crash. This information could be very useful for emergency services to
decide the rescue resources required. We did not include a field indicating
8
Table 2: Accident dataset.
Field Description
Time Time when crash occurred
Coordinates Crash point coordinates
Speed Vehicle speed at the crash moment
Point Of Impact Point of impact for the crashed vehicle
Rollover If vehicle has overturned
Dam Area Dam area for the crashed vehicle
Non Invl Number of non-motorists involved in the crash
Fire If vehicle is in fire
Max Acceleration Vehicle maximum deceleration during the crash
the number of vehicles involved in the crash because all the collided vehicles
will send their own messages.
In the set of data related to Occupants, the fields used are those indicated
in Table 3. For this dataset, we used basic fields from the GES database,
and also added nine new fields: Blood,Allergies,Illness,Medication,Preg-
nant,Weight,Age,Sex and Id. The first eight fields could be previously
stored on the mobile phone of each passenger. Then, in case of an accident,
emergency services will receive all this individualized medical information,
thereby allowing emergency services to identify each person.
Finally, Table 4 shows the different fields related to Environment. For
this dataset, we used fields from the Accident dataset available in the GES
database.
3.3. Qualitative Comparison of Similar Existing Ontologies
Table 5 presents a summary of the VEACON comparison we made with
respect to other existing ITS ontologies. We have structured the comparison
in eight different categories: (a) the source of their attributes, (b) if they
used any historical database, (c) if they support accident severity prediction,
(d) the tagging language used, (e) the software frameworks used, (f) if data is
grouped into classes, (g) if they present the relationships, and (h) the method
selected for the evaluation.
As shown, VEACON is the only ontology that uses historical data for
its design, enabling the prediction of accidents severity, which in our opin-
ion makes the difference, since nowadays traffic accidents cause millions of
9
Table 3: Occupant dataset.
Field Description
Id Occupant identifier
Airbag Airbag availability/function in the seat position of the occupant
Rest Sys Restraints that are being used by the occupant immediately prior to the crash
Seat Pos Occupant seating position
Blood Occupant blood type
Allergies Occupant allergies
Illness Occupant illness
Medication If occupant needs specific medication or treatment
Weight Occupant weight
Age Occupant age
Sex Occupant gender
Pregnant If occupant is pregnant
people killed or severely injured. Moreover, in contrast to VEACON, the
rest of studied ontologies have not been evaluated under real testbed crash
environments, and using vehicular simulations.
4. Validation of Our VEACON Proposal
In vehicular environments, wireless technologies enable peer-to-peer mo-
bile communication among vehicles (V2V) and communication between ve-
hicles and the infrastructure (V2I). We think that the combination of V2V
Table 4: Environment dataset.
Field Description
Speed Limit Roadway legal speed limit
Surface Cond Roadway surface condition
Road Profile Roadway profile
Road Align Roadway alignment
Weather Atmospheric conditions at the time of the accident
Light Cond Light conditions at the time of the accident
10
Table 5: ITS Ontologies Comparison.
Description VEACON Eigner and Lutz
(2008)
Kannan et al.
(2010)
Where does it select
attributes?
GES database enriched At their own discretion At their own discretion
There is historical
data to compare ac-
cidents?
Yes, the GES database No No, it is designed to
support a Driver Assis-
tance System
Does predict it the
damage from acci-
dent?
Yes, using historical
data
No, it is only designed
to prevent accidents
This ontology is not
specific for traffic acci-
dents
Tagging language OWL OWL OWL
Software used Prot´eg´e Not specified Prot´eg´e
Does it present
the ontology
relationships?
Yes No Only partially
System Evaluation Crash tests and net-
work simulations
Simulations using the
Virtual Traffic Simula-
tor (VISSIM)
Ad-hoc simulator
Map Topology for
Validation
Real roadmaps Synthetic roadmaps Synthetic single, 2-way,
and 4-way roads
and V2I communications can propel our communication capabilities even fur-
ther, improving the traffic safety under Intelligent Transportation Systems
(ITS). To verify that our ontology works correctly in Vehicular Networks, we
performed two different kinds of experiments: (i) real crash tests involving
Vehicle-to-Infrastructure (V2I) communications, to verify that the message
using our ontology proposal is correctly sent to the emergency services in
case of an accident, and (ii) vehicular network simulations, to study how
VEACON messages would be propagated to the rest of vehicles in terms of
V2I and V2V communications, in a realistic urban environment.
4.1. Real Crash Tests
To prove the feasibility of our ontology, we performed several crash ex-
periments in the facilities of Applus+ IDIADA Passive Security Department
11
sited in Santa Oliva (Tarragona, Spain)2. This laboratory is one of the most
sophisticated crash test laboratories in the world, and is an official center for
approval under the EuroNCAP: European New Car Assessment Programme
(2011). Due to the cost of using real vehicles in the collision experiments,
tests were performed using a platform (known as “sled”) which is able to
simulate different kind of vehicles and impact severities in traffic accidents.
Figure 2 shows the sled used in our tests. As shown, a series of weights
were added to accurately simulate the behavior of a conventional vehicle.
Figure 3 details the electronic components used to implement the OBU
prototype on the platform. Validation experiments consisted in front, side
and rear-end collision tests with different severities. The classification of
the severity of the collision is dictated by the EuroNCAP and RCAR tests
(RCAR, the Research Council for Automobile Repairs, 2011). In our ex-
periments, the On Board Unit (OBU) installed in the sled collected all the
information provided by the sensors, built the warning message according to
our VEACON ontology, and sent this alert information at the collision time
by using wireless communications. An external computer acted as a Road
Side Unit (RSU), in charge of receiving the warning messages broadcasted by
vehicles, and forwarding them to a suitable Public Safety Answering Point
(PSAP) or 112 Service Center.
The results obtained in the real crash tests were very promising. Figure
4 shows some of the acceleration pulses recorded by the OBU and sent to the
RSU for three different front crash tests. Although different types of vehicles
were tested, the figure only includes those corresponding to the large family
car segment. The OBU is in charge of determining the severity of the direct
impact, but interpreting acceleration values is not trivial since the received
pulses have a very limited duration, and also because both their amplitude
and duration should be considered in the classification. As shown in Figure
4, using simple acceleration thresholds to distinguish acceleration pulses is
not enough (e.g. the minor accident has a peak deceleration that is greater
than for the severe accident). However, we discovered that the value of the
integral function defined as the variation of acceleration over time allows
simple and accurate pulse classification since it accounts for both amplitude
and duration of the pulse.
The experiments performed in real crash tests proved that our system was
2http://www.idiada.es
12
Figure 2: Sled used in our crash tests.
Figure 3: Close-up of the electronic components installed on the sled.
13
-35
-30
-25
-20
-15
-10
-5
0
5
10
0 0.05 0.1 0.15 0.2 0.25
Acceleration (G)
Time (s)
No accident (15 km/h)
Minor accident (40 km/h)
Severe accident (64 km/h)
Figure 4: Vehicle acceleration pulses during front crashes in the large family car segment.
able to collect all the information provided by in-vehicle sensors, to build
the VEACON-compliant message, and to communicate with the RSU in
every tested situation without message loss. Moreover, the OBU was able to
accurately determine the impact severity by using the integral approach in all
cases, generating an adequate warning message, and sending it to the nearest
RSU. Warning messages were broadcasted and successfully received by the
RSU, and the contained information was correctly extracted and interpreted
(see Figure 5). Consequently, we conclude that the VEACON ontology can
be successfully used to notify accident situations in real environments.
4.2. Network Simulation Tests
To study how messages built using VEACON propagate in a vehicular
network scenario, simulations were done using the ns-2 simulator. We im-
proved the simulator by including the IEEE 802.11p standard closely, which
defines enhancements to 802.11 required to support ITS applications (IEEE
802.11 Working Group, 2010). In terms of the physical layer, the data rate
used for message broadcasting is 6 Mbit/s, as this is the maximum rate for
broadcasting in 802.11p. The MAC layer was also extended to include four
different priorities for channel access. Therefore, application messages are
categorized into four different Access Categories (ACs), where AC0 has the
lowest and AC3 the highest priority.
14
Figure 5: VEACON-compliant information received by the 112 Service Center and pre-
sented in a web interface.
15
The purpose of the 802.11p standard is to provide the minimum set of
specifications required to ensure interoperability between wireless devices
attempting to communicate in potentially rapid changing communication
environments. For our simulations, we chose the IEEE 802.11p because it is
expected to be widely adopted by the industry.
We want to evaluate whether or not our proposal ontology could affect
to the dissemination of warning messages in Vehicular Networks.
We tested our proposed ontology by evaluating the performance of a
Warning Message Dissemination mechanism where each vehicle periodically
broadcasts information about itself, or about an accident. These messages
are built according to our VEACON ontology.
Our simulations have been carried out in two different scenarios of 4 km2,
obtained from real maps from New York (USA) and Rome (Italy). As shown
in Figure 6, the New York map presents the longest streets, mostly arranged
in a Manhattan-grid style, while the city of Rome represents the opposite
situation, with short streets in a highly irregular layout.
To increase the realism of our simulations, we used Citymob for Roadmaps
(C4R)3, a mobility generator based on SUMO (Krajzewicz and Rossel, 2007).
C4R includes all the original characteristics from SUMO (collision-free ve-
hicle movement, multi-lane streets, etc.). In addition, it is able to define
attraction and repulsion points which simulate areas with different vehicle
densities, something very common in real cities. Regarding the radio prop-
agation model, the network simulator was also modified to make use of our
Real Attenuation and Visibility (RAV) scheme (Martinez et al., 2010a), which
proved to increase the level of realism in VANET simulations since it accounts
for the effect of obstacles (e.g., buildings) in radio signal propagation when
simulating urban scenarios.
We simulated a frontal impact scenario where two vehicles are involved.
The first vehicle is a family car with two occupants, and expressing all the
information required, according to VEACON, a message of 13 KBytes. The
second vehicle is a minivan with eight occupants, which required up to 18
KBytes to code the data for all passengers. Each simulation run lasted for
450 seconds. In order to achieve a stable state, we collect data only after
the first 60 seconds. All results represent an average of over 30 executions
with different scenarios (maximum error of 10% with a degree of confidence
3C4R is available at http://www.grc.upv.es/software/
16
(a) (b)
Figure 6: Scenarios used in our simulations: (a) fragment of the city of New York (USA),
and (b) fragment of the city of Rome (Italy).
of 90%). Table 6 shows the parameters used in the simulations.
In order to determine the feasibility of VEACON in different situations,
we present the results obtained when considering both V2I and V2V commu-
nications. We consider that some factors, such as the density of vehicles, the
density of RSUs, or the map topology, should have a significant impact on the
performance of our ontology-based warning message dissemination scheme.
Therefore, we performed different experiments by varying these factors, and
studied their effect on the following metrics: (i) the notification time (i.e.,
the period elapsed between the time when a warning-mode vehicle requests
for help, and the time when any RSU receives the warning message, deliv-
ering it to the next Public Safety Answering Point (PSAP) or 112 Service
Center), (ii) the percentage of RSUs receiving the warning messages, (iii)
the warning notification time (i.e., the time required by nearby vehicles to
receive a warning message sent by a collided vehicle), and (iv) the percentage
of vehicles receiving the warning messages. These metrics are crucial when
assessing with the usefulness of our studied system, since a warning message
delivered too late is useless when facing dangerous situations, and nearby
vehicles must be informed about these situations.
17
Table 6: Parameter Values for the Simulations.
Parameter Value
number of vehicles 50,100,200, and 400
simulated cities New Y ork and Rome
simulated area 2000m×2000m
number of collided vehicles 2
warning packet size 13 and 18KB
packets sent by vehicles 1 per second
warning message priority AC3
normal message priority AC 1
mobility generator C4R
mobility models Krauss and Downtown
MAC/PHY 802.11p
radio propagation model RAV
maximum transmission range 400m
4.2.1. V2I Communications Results
Regarding V2I communications, Table 7 shows the minimum notification
time and the reachability (i.e., the percentage of times that warning messages
reach any RSU), when varying the number of vehicles, the number of RSUs,
and the simulated roadmap. As shown, the simulated roadmap affects both
the warning notification time and the percentage of RSUs receiving the warn-
ing messages, especially when the vehicle density is very low (Rome shows
higher notification times, but, in contrast, it shows a higher percentage of
successful RSU notifications). When 400 vehicles are simulated, notification
time is slightly higher in Rome, since the topology is more complex than New
York. However, results show that in complex roadmaps like Rome, the per-
centage of receiving RSUs is higher compared to New York. We think that
this demonstrates that V2I communications can play an important role in
such complex scenarios. Moreover, as expected, results reflect that increasing
the density of vehicles highly increases the chances for warning messages to
reach any RSU, i.e., the emergency services notification effectiveness.
4.2.2. V2V Communications Results
Regarding V2V communications, Figure 7 shows the obtained results
when varying the scenario topology and the vehicle density.
18
Table 7: V2I Simulation Results.
New York Rome
Vehicles RSUs Notif. time (s) Reach. (%) Notif. time (s) Reach. (%)
50
1 0.383 5 12.919 25
2 0.674 5 6.669 55
4 0.790 30 6.188 65
8 0.790 30 4.841 75
100
1 1.383 50 1.724 95
2 1.146 65 1.427 95
4 1.107 70 1.456 95
8 1.247 70 1.283 95
200
1 1.147 80 1.643 100
2 0.954 85 1.348 100
4 0.815 85 1.216 100
8 0.850 85 0.822 100
400
1 1.377 90 2.008 90
2 1.162 90 1.720 100
4 1.130 100 1.513 100
8 0.929 100 1.193 100
As shown, both factors have a high impact on the performance. The
selected map has a great influence on the percentage of vehicles receiving
warning messages and on the warning notification time, especially when the
vehicle density is low. When only 50 vehicles are simulated, warning messages
reach only 5.50% of vehicles in Rome and 31.60% of vehicles in New York,
where the long and regular streets allow easy propagation of the wireless
signal. The system requires 1 second to reach 5%, and 15% of the total
number vehicles, respectively.
For higher vehicle densities, the differences between the maps are reduced,
the percentage of informed vehicles increases (e.g., when 400 vehicles are
simulated, warning messages reach 99.35% of the vehicles in the New York
scenario, and 90.30% of the vehicles in Rome), and the system needs less
time to inform the same percentage of vehicles (e.g., when 200 vehicles are
simulated, the system only requires 1.5 seconds to reach 60% of the vehicles
in New York, and 2.3 seconds to reach the same percentage in Rome).
19
0
20
40
60
80
100
0 5 10 15 20
% of vehicles receiving the warning messages
Warning notification time (s)
400 vehicles
200 vehicles
100 vehicles
50 vehicles
(a)
0
20
40
60
80
100
0 5 10 15 20
% of vehicles receiving the warning messages
Warning notification time (s)
400 vehicles
200 vehicles
100 vehicles
50 vehicles
(b)
Figure 7: Warning notification time when varying the density of the vehicles and the
simulated roadmap: (a) New York, and (b) Rome.
20
5. Conclusions
In this paper we present VEACON, a Vehicle Accident Ontology for Ve-
hicular Networks. VEACON allows to efficiently structure and encode the
information collected by in-vehicle sensors, enabling the interoperatibility
among all the agents involved in modern ITS (vehicles, RSUs, emergency
services, authorities, etc.). VEACON combines the information sensed from
the accident with the available data in the GES database to offer rich and
structured information to the parties involved in traffic accidents manage-
ment.
VEACON provides an ontology based approach for faster data searching
and improved understanding between vehicular applications.
To verify that messages structured by using VEACON are correctly trans-
mitted using VANETs, we performed two different tests. On the one hand,
crash tests proved that the OBU correctly estimates the severity of the
accident, and our system was able to collect, build, and communicate the
VEACON-compliant messages with the RSU without message loss. On the
other hand, by using simulations we demonstrated the feasibility of our sys-
tem in terms of V2I and V2V communications. Experimental results high-
light that both nearby vehicles and infrastructure elements (RSUs) are no-
tified about and accident in just a few seconds, increasing the emergency
services notification effectiveness, and thereby validating the proposed ap-
proach.
Acknowledgments
This work was partially supported by the Ministerio de Ciencia e Inno-
vaci´on, Spain, under Grant TIN2011-27543-C03-01.
References
Antolino Rivas, D., Barcel´o-Ordinas, J., Guerrero Zapata, M., Morillo-Pozo,
J., 2011. Security on VANETs: Privacy, misbehaving nodes, false infor-
mation and secure data aggregation. Journal of Network and Computer
Applications 34 (6), 1942–1955.
Bakhouya, M., Gaber, J., Lorenz, P., 2011. An adaptive approach for infor-
mation dissemination in Vehicular Ad hoc Networks. Journal of Network
and Computer Applications 34 (6), 1971–1978.
21
Berners-Lee, T., Hendler, J., Lassila, O., May 2001. The Semantic Web.
Scientific American.
Breed, D. S., 02 1991. Vehicle Crush Zone Crash Sensor, Patent US 4995639.
Breed, D. S., Castelli, V., Johnson, W. C., DuVall, W. E., 12 1998. Vehicle
Occupant Position and Velocity Sensor, Patent US 5848802.
Chong, M., Abraham, A., Paprzycki, M., 2005. Traffic Accident Analysis
Using Machine Learning Paradigms. Informatica 29, 89–98.
Daeinabi, A., Pour Rahbar, A., Khademzadeh, A., 2011. VWCA: An efficient
clustering algorithm in vehicular ad hoc networks. Journal of Network and
Computer Applications 34 (1), 207–222.
Daemen, J., Rijmen, V., 2002. The Design of Rijndael: AES – The Advanced
Encryption Standard. Springer-Verlag.
Di Lecce, V., Amato, A., May 2009. A multilevel interface for an intelligent
transport system. In: IEEE Instrumentation and Measurement Technology
Conference (I2MTC’09). pp. 1397–1401.
Eigner, R., Lutz, G., March 2008. Collision avoidance in VANETs - an appli-
cation for ontological context models. In: Sixth Annual IEEE International
Conference on Pervasive Computing and Communications (PerCom). pp.
412–416.
EuroNCAP: European New Car Assessment Programme,
2011. Evaluation procedures and test results. Available at
http://www.euroncap.com/testprocedures.aspx.
Ford Motor Company, 2011. Ford researchers develop car seat that monitors
drivers’s heart activity. Available at http://media.ford.com/news/
fordresearchersdevelopcarseatthatmonitorsdriversheartactivity.htm.
Garshol, L. M., 2004. Metadata? Thesauri? Taxonomies? Topic maps!
Making sense of it all. Journal of Information Science 30 (4), 378–391.
Gruber, T. R., December 1995. Toward principles for the design of ontolo-
gies used for knowledge sharing. International Journal Human-Computer
Studies 43, 907–928.
22
Husby, H. S., Simpson, L., 06 2000. Seat Belt Tension Sensor, Patent US
6081759.
IEEE 802.11 Working Group, July 2010. IEEE Standard for Information
Technology – Telecommunications and information exchange between sys-
tems – Local and metropolitan area networks –Specific requirements – Part
11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) specifications Amendment 6: Wireless Access in Vehicular Envi-
ronments.
Kannan, S., Thangavelu, A., Kalivaradhan, R., July 2010. An intelligent
Driver Assistance System (I-DAS) for vehicle safety modelling using on-
tology approach. International Journal Of UbiComp (IJU) 1 (3), 1–13.
Krajzewicz, D., Rossel, C., 2007. Simulation of Urban MObility
(SUMO). Centre for Applied Informatics (ZAIK) and the Institute of
Transport Research at the German Aerospace Centre, Available at
http://sumo.sourceforge.net/index.shtml.
Martinez, F. J., Fogue, M., Coll, M., Cano, J.-C., Calafate, C. T., Man-
zoni, P., July 2010a. Assessing the Impact of a Realistic Radio Propa-
gation Model on VANET Scenarios Using Real Maps. In: 9th IEEE In-
ternational Symposium on Network Computing and Applications (NCA).
Boston, USA, pp. 132–139.
Martinez, F. J., Toh, C.-K., Cano, J.-C., Calafate, C. T., Manzoni, P., sum-
mer 2010b. Emergency services in future intelligent transportation systems
based on vehicular communication networks. IEEE Intelligent Transporta-
tion Systems Magazine 2 (2), 6 –20.
National Highway Traffic Safety Administration (NHTSA), 2011. National
Automotive Sampling System. Available at http://www.nhtsa.gov/NASS.
Petzold, S., 05 1999. Rain Sensor System for a Motor Vehicle, Patent US
5900821.
RCAR, the Research Council for Automobile Repairs, 2011. Design guides
and information for manufacturers. Available at http://www.rcar.org.
23
Zhai, J., Wang, Q., Lv, M., July 2008a. Application of XML Topic Maps to
knowledge navigation and information retrieval for urban traffic informa-
tion portal. In: 27th Chinese Control Conference (CCC). pp. 458–462.
Zhai, J., Zhou, Z., Shi, Z., Shen, L., 2008b. An integrated information plat-
form for intelligent transportation systems based on ontology. In: Xu, L.,
Tjoa, A., Chaudhry, S. (Eds.), Research and Practical Issues of Enter-
prise Information Systems II Volume 1. IFIP International Federation for
Information Processing. Springer Boston, pp. 787–796.
24
