You are on page 1of 33

Archives of Computational Methods in Engineering

https://doi.org/10.1007/s11831-020-09487-1

ORIGINAL PAPER

Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey


Faruk Baturalp Günay1   · Ercüment Öztürk2 · Tuğrul Çavdar2 · Y. Sinan Hanay3 · Atta ur Rehman Khan4

Received: 3 February 2020 / Accepted: 27 August 2020


© CIMNE, Barcelona, Spain 2020

Abstract
Localization has become an important area of research with the development of wireless communication technologies. Of
particular prominence within this area is Vehicular Ad Hoc Networks (VANET), which plays an important role in many
applications such as vehicle tracking, accident prevention, and efficient transportation. GPS technology, which can easily be
integrated into vehicles has been instrumental in the development of localization-based applications. Furthermore, owing
to the improvements in the wireless communication technologies such as Internet of Things (IoT), VANET localization can
be implemented in places where GPS is not available. In this survey, we begin by providing a summary of the use cases of
localization in VANET and highlight various techniques that are proposed in the literature. Moreover, we classify the stud-
ies in this area with respect to their methodologies, and discuss their respective advantages and disadvantages. Finally, we
provide an overview of potential research topics for future research.

1 Introduction of the close proximity between vehicles. The accuracy of


guidance operations carried out on electronic maps relies on
Low-cost and accurate localization is a crucial requirement the quality of the captured location information. However,
for various areas of usage including vehicular networks [1]. there are difficulties in locating vehicles in VANETs, due
With rapid urbanization and the emergence of transporta- to mobility of vehicles and the weakness of signals caused
tion systems in metropolitan areas, localization has become by environmental objects (fading and Doppler effects). A
instrumental in navigation systems of vehicles. Vehicular detailed list of published papers focusing on this aspect is
Ad-Hoc Network (VANET) is a temporary network environ- provided in this paper.
ment where vehicles communicate and share information In order to use VANET for localization, nodes and vehi-
with each other [2]. Recently, VANET has been a very active cles should first identify their own locations. This can be
area of research to tackle the traffic problems in congested achieved by leveraging either GPS signals or manually
areas of cities. Estimating the location of the vehicles in located Road Side Units (RSU). Also, a vehicle in VANET
motion can provide an overview of traffic on the network can determine its location by estimating the distance
and prevent traffic accidents [3]. It can also provide signifi- between neighboring nodes and by replacing this distance
cant fuel savings by reducing the disruptions in transporta- information with nearby nodes in the network as shown in
tion services. In dense traffic conditions, capturing accurate Fig. 1. This type of relative localization is a widely used
location information is highly critical for VANETs because technique in Mobile Wireless Sensor Networks [4]. There
are many studies in the literature focusing on the VANET
localization issue with conventional techniques (GPS, map
* Faruk Baturalp Günay matching, dead reckoning, etc.) or limited usage area (Vehi-
baturalp@atauni.edu.tr
cle Collision, Autonomous Vehicles) [5, 6]. However, there
1
Department of Computer Engineering, Ataturk University, is a lack of detailed analyses and comparison of conventional
25240 Erzurum, Turkey methods and newly designed IoT the integrated methods
2
Department of Computer Engineering, Karadeniz Technical such as RSUs. Also, more research is needed for the usage
University, 61080 Trabzon, Turkey of filters (Kalman, Particle filter, etc.) and their role in coop-
3
Department of Computer Engineering, Erzurum Technical erative localization for VANET localization. Unlike previous
University, 25050 Erzurum, Turkey studies [3, 7] one of the remarkable points in this paper is to
4
College of Engineering and Information Technology, Ajman
University, 346 Ajman, UAE

13
Vol.:(0123456789)
F. B. Günay et al.

2 Applications of VANET Localization

Although localization is used in many different areas, it


can be classified under two broad categories for VANET:
military and civilian use. For the former one, there is a
wide range of applications such as positioning and track-
ing of vehicles, search and rescue operations, finding the
position of targets, guided control and landing/launch-
ing of aircraft in fierce weather conditions. Cartography,
vehicle tracking systems, tourism, agriculture, forestry
are prominent areas of civil use. For both of the afore-
mentioned areas, accurate location identification is criti-
cal under any circumstances. Localization in VANETs is
mainly for civilian use, but it must be done with no latency
Fig. 1  General VANET localization scheme and high accuracy.
Localization in VANETs that is mostly used for civil-
ian use has to be made quickly at a high accuracy level
focus on cooperative localization and its data fusion which because it is a dynamic network. The previous applications
is more popular these days. had several features including accident prevention, crash
The key motivational point to conduct this study is to report, vehicle speed, and further location estimation,
remedy the deficiency of a collected work to summarize the and the current ones have got several objectives such as
studies on localization in VANET including recent studies autonomous parking, precise navigation, location estima-
[3, 5]. Another reason is to evaluate and sum up them in tion and accident prevention inside the tunnels. There are
an innovative classification depending on the content and also different applications regarding driver assistance and
functional status of the growing body of evidence rather than cooperative driving. In the following sections, the primary
the traditional ones. Below a summary of the contributions application areas are listed where VANET based localiza-
of this paper is provided: tion is utilized.

• As IoT technologies are widely used in many applica-


tions, access to vehicles is now easier than ever and more 2.1 Accident Prevention Applications
information can be collected from them in a more eco-
nomical way. In this paper, VANET localization appli- In accident prevention applications, communication proto-
cations are explained from the point of IoT technologies cols such as IEEE 802.11 b/g are used, and vehicular sen-
as well as other aspects distinguishing this survey work sors are chosen for crash location. The precise position can
from existing ones. be determined by including speed and heading information
• Several parameters (i.e. usage of RSU, Cooperative with localization data [5, 8]. For position estimation and
Localization methods, communication technologies) and accident avoidance inside tunnels, several methods are
their corresponding effect on VANET localization are utilized including Dead Reckoning (DR) and RSUs. This
investigated in great detail in this work. way, communication protocols can be implemented in a
• A detailed summary of all methods in VANET localiza- more efficient way where GPS signals are not available.
tion proposed in the literature is provided in this paper. Another application is alerting drivers to avoid traffic light
• Real tests on the city environment as well as simulations violations. In this application, warning messages are trans-
in the studies are explained in this paper. mitted to a vehicle, if the traffic signs are not followed.
In addition, a safe trailing distance application provides
This study is organized as follows: Sect. 2 briefly introduces drivers with warnings when the distance between vehi-
the applications of VANET localization. Section 3 gives the cles is smaller than the required distance. Furthermore, the
detailed description of basic localization technologies and left turn assistant collects information on intersections and
methods of VANET. Section 4 explains Cooperative Local- helps vehicles to turn left and complete their turn safely.
ization Term. Section 5, 6 and 7 review all categories of Apart from these, intersection collision warning, blind-
VANET localization algorithms in detail. Section 8 provides spot capture, and pedestrian crossing are considered as the
detailed information about the existing problems and future other applications in this group [9].
direction for localization algorithms. Finally concluding
remarks are made in Sect. 9.

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

2.2 Public Safety Applications the velocity based on the road direction and conditions,
through intra-vehicular communication.
Public Safety Applications can help emergency response As mentioned above, the localization process is needed
teams reaching an accident site as soon as possible. For in the VANET applications of vehicle maintenance and
this purpose, an ambulance, trying to reach an accident diagnostics, accident prevention, public safety, and sign
area, can communicate with all vehicles on the road and extension, or it makes these more efficient, and even the
send a message to clear the lane [10]. The emergency vehi- integrated use of multiple localization techniques increases
cle signal preemption changes traffic lights at the junc- the efficiency. V2V communication within VANET during
tions where the ambulances approach, turning red light to the localization is a key factor to facilitate it at the same
green through vehicle-to-infrastructure (V2I) communi- time. In these respects, it is necessary to utilize localization
cation system and help the vehicle moving forward with- in different VANET applications. Naturally, this necessity
out getting stuck in traffic. In addition to lane changing varies from application to another. For instance, in coopera-
warning, visibility enhancer warns drivers of bad weather tive adaptive cruise control or vehicle following application,
conditions and facilitates driving in these conditions. Fur- localization information is needed in high accuracy. It is
thermore, road condition warning application reports road essential that the jointly moving vehicles are sharing their
conditions such as ice, gravel, snow, and rain, etc.. Also, individual location with one another. However, in order to
it informs all vehicles which approach to those road zones achieve effectiveness in the applications or processes that
through VANET. Lastly, the SOS service and post-crash should be more sensible, such as vehicle collision warn-
warning applications are also considered to be part of this ing or automatic parking, ultrahigh localization accuracy is
category [9]. required. Owing to the high accuracy localization the target
may avoid crash or damage before it’s happening via this
kind of application.
2.3 Sign Extension Applications

These applications are designed to provide careless drivers 3 VANETs Localization Techniques
with warnings to follow traffic signs and watch for road-
side hazards [11]. In these applications, the infrastructure- In this section, we cover various techniques used in
to-vehicle (I2V) communication is used to transmit mes- localization.
sages at the frequency of 1 Hz and within the minimum
range of 100 to 500 m. Thus, drivers can recognize nearby 3.1 GPS and DGPS
places such as hospitals, schools, etc. and have a chance to
reduce their speeds by receiving notification before they Global Positioning System (GPS), at its simplest form, is a
approach a hazardous hairpin bend. Broadcasting warning satellite network used to determine the position on Earth’s
messages to vehicles like the wrong way, overpass heights, surface. Its working principle is based on determining the
etc. are included in Sign Extension applications. position by computing the differences between the devices
and the satellites. The GPS satellite network belongs to the
U.S. Department of Defense and transmits to the Earth the
2.4 Vehicle Maintenance and Diagnostics location information that they are constantly estimating.
Applications Each satellite completes one round around Earth in nearly
12 h [13]. From 1978 till the 1990s, a network of 24 sat-
The aim of these applications is to disseminate to driv- ellites (21 actives, 3 spares) was established. The number
ers the messages of vehicle conditions and their possible further increased in time, particularly in recent times. From
breakdowns in an approximately 400 m coverage zone. the orbits they travel at a height of about 20.000 km, the
From other vehicles, information can be delivered through satellites transmit low-power radio signals to the Earth’s sur-
different applications in VANET [12]. Other than safety, face. They are located in a way that every region throughout
several applications including cooperative cruise control the globe can be seen by at least four satellites. To realize
offer drivers to use their own vehicles more easily. It is the positioning activity, the GPS receiver needs to process
aimed to ensure that a group of vehicles autonomously the signals from the satellites and estimate its own location.
travel in the fleet at a speed inputted before setting out For the computation process, a technique called
by the driver, free from driver’s intervention. In the Time of Arrival (ToA) is used, and as a result, the GPS
application, Vehicle-to-vehicle (V2V) communication is receiver finds out the latitude, longitude, and altitude of
used, and partial automatic modifications can be made on its own position. However, the time equalization problem
occurs due to the mobility of the satellites during signal

13
F. B. Günay et al.

transmission. It is solved by leveraging an atomic clock 3.2 Data Matching Techniques


that every GPS satellite contains. Using this clock, any
time-lag can be avoided between the satellites, or the sat- The methods of Map Matching and Dead Reckoning fall into
ellite and the receiver [14]. The first GPS satellite was this category because they serve a function by matching or
launched by the US for military purposes and used for a integrating the subsequent location data. Map Matching is
long time for the same purposes. Russia also started to a process in which the geographic coordinates of any object
develop its own satellite system, GLONASS and experi- that are estimated using GPS are matched to the real world
enced drawbacks on coverage zone throughout the Earth map. The estimated position points are recorded in a list
until the mid-2000s [15]. In addition, nowadays, the Euro- representing the motion of a user or a vehicle, and thus a
pean Union, China, and India have their own positioning logical model is designed inconsistent with observations.
systems, named Galileo, BeiDou and NAVIC, respec- Map matching algorithms can be divided into real-time and
tively. Furthermore, Japan developed a system, Quasi- off-line algorithms. The first group relates the latest GPS
Zenith Satellite System, to improve the accuracy and reli- information to the map while the latter is used after data are
ability of GPS, and it began to operate with its 4 satellite recorded and then matched with the map [17]. For the map-
constellations. The advantages and disadvantages of GPS matching technique, different positions obtained at a certain
technology are shown in Table 1. time interval are used to identify an approximate route. In
Differential GPS (DGPS) is an improved method of recent times, with the advent of using high-quality digital
GPS which utilizes reference ground stations besides GPS maps, the obtained results can be matched with the loca-
satellites. This method is based on determining the dif- tion from the navigation system, and so sensors are better-
ference between satellite and actual pseudoranges. Also, calibrated for localization. Moreover, in the cooperative map
those stations transmit the mentioned difference to other matching technique, a more precise map matching can be
stations and it is evaluated as a correction signal. This accomplished by receiving information from other vehicles
way decreases localization error from meter to centim- in VANET, unlike traditional map matching methods [7].
eter-level. A disadvantage of this method is the neces- Dead Reckoning (DR) is the procedure in which a vehi-
sity to set up fixed reference nodes for giving differential cle estimates its current position based on its known posi-
information [14]. tion prior to a certain period of time. In a local coordinate
system, DR solution can be made through trigonometric

Table 1  GPS Advantages and disadvantages


Advantages of GPS Disadvantages of GPS

Provides high accuracy positioning (Up to 50 km in the area has an error rate of 6–10 m, GPS satellites are under the control of the United
100–500 km in area 7–10 m, 1000 km in area 10–9 m) [16]. States. Because of the permissions required to use
GPS satellites, the position in the sky changes due
to the decision made within a day or two, causing
the different signals to be picked up, resulting in
poor positioning accuracy [16]
Low transaction complexity Sensitivity of the GPS receiver may be reduced
due to problems such as multipath and noise.
Especially in urban areas (between high build-
ings), satellite signals are blocked by hitting high
buildings and a significant decrease in the number
of satellites seeing the area results in low position-
ing accuracy. In addition, satellite signals can be
reflected from the surface of the buildings, causing
multipath effects and may reduce the positioning
accuracy. To overcome these problems, different
techniques and methods such as DGPS, GPS and
Inertial Navigation System (INS) or GPS and Dead
Reckoning (DR) have been proposed
It does not need a directional antenna, so it is enough for the receiving antennas to see the Receiver antenna should have a clear view of the sky
sky to receive the satellite signal.
GPS accuracy is largely independent of weather conditions
The measurement is possible throughout the day
Three-dimensional position can be obtained
Easy to use and install

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

calculations [18]. (x,  y) represents the coordinates, and PL(d) = PL(d0 ) + 10 × n × lg(d∕d0 ) + X (3)
(0,0) the origin in this system. The following equations can
explain the DR method [18] : where PL(d0 ) is the path loss of radio signal propagation per
k=1
unit road length, X is the random number of Gauss distribu-
∑ tion, d is the distance between receiver and transmitter, and
xi =x0 + si × cos𝜃i (1)
i=0
n is weakness factor of the environment. With this, we can
simply find out the RSSI (dBm) value (dBm) as below [21].
k=1
∑ RSSI = P𝛾 + G𝛾 − PL(d) (4)
yi =y0 + si × sin𝜃i (2)
i=0 In the above equation, P𝛾 is wireless signal power in dBm
where x0 , y0 ) is the initial vehicular position at the time of and G𝛾 is the gain of transmitting antenna.
t0 . si and 𝜃i are respectively the shortest way and the changed Every chip producer describes their own value of “RSSI
direction based on the former position ( xi − 1, yi − 1). Max” in accordance with the IEEE 802.11 standard, which
In other words, DR represents localization for a target can be within the range of 0 to 255. For instance, Atheros
that is out of a node’s coverage area. The DR method makes uses a scale from 0 to 60, and Cisco from 0 to 100. Usually,
calculations on a map using data such as location, head- RSSI value is converged to transmitter power as long as the
ing (movement direction), speed, distance, time, and wind. target gets close to RSU chip or access point. Signal power
Several techniques were suggested for localization processes can be expressed as dBm or RSSI, and however, these are
to run in the DR zones [18, 19]. For example, the vehicular different measurement units. The difference between them
position can be estimated based on the latest known position represents power as dBm mW (miliwatt) while RSSI is a
and the current speed and direction of the vehicle for a short relative value. Basic RSSI scheme is illustrated in Fig. 2.
time when GPS is unavailable. It has a disadvantage that the TDOA is based on calculating the distance between trans-
localization accuracy declines due to the errors accumulated mitter and receiver nodes by measuring the time difference
during a long period when it is employed. Furthermore, in of the receiver node and anchor nodes. The signals emitted
a relevant study reviewed before, it is showed that it can be from the target point, arrive at different times to at least 3
utilized in high accuracy and feasibility in a traffic environ- synchronous receiver antennas located at different places.
ment without GPS signals [19]. By computing the time difference of arrived signals to the
receivers, the target location is estimated.
Compared to the TDOA method, TOA uses the abso-
3.3 Lateration Techniques lute arrival time of a signal from transmitter to receiver,

Lateration techniques leverage radio signals, such as signal


quality, the arrival time of a satellite signal, and the time
difference between the arrival times of signals. Hence, in the
current study, we use these techniques for both cooperative
and non-cooperative localization types. These techniques
based on different algorithms have the ability to position
the nodes in VANET through satellite signals, which are
explained in detail below.

3.3.1 Distance‑Based Lateration Techniques

RSSI is a method used to measure the received signal qual-


ity with a device at an exact signal power value. By means
of this method, the received radio signal power and the lost
signal propagation model are utilized for distance measure-
ment and location estimation. There are mainly three factors
for signal weakness: slow fading, multi-path fading of the
radio signal and the path loss of wireless signal propaga-
tion. Furthermore, there are the effects of wireless channel,
interference, noise, and the radio path loss equation can be
shown such below [20]: Fig. 2  Basic scheme of RSSI lateration

13
F. B. Günay et al.

rather than the time difference measured for the signal to Those are utilized as well as other techniques that are model
depart from one station and arrive at another one. In this or graphic-based, having objective of high-level control [23].
respect, it is also called Time of Flight (TOF). The distance Likewise, these are used in VANET and they assist in solv-
between receiver and transmitter can be computed based on ing issues such as localization and data fusion. In the fol-
the frequency and speed of the radio signals. Here at least lowing sections, various filtering techniques are discussed.
three anchor nodes should be used. The synchronization of
the node to locate with anchor nodes is the key point. It 3.4.1 Kalman Filter
is unlikely that both receiver and anchor nodes have high-
precision synchronous timers. Thus, it is more practical that The Kalman filter is an algorithm that estimates the next
either its location must be determined, otherwise, anchor state over the noisy data, even though called a filter, based
nodes receive time data from a common timing source. The on current and previous states. Kalman filter is used in many
distance between these nodes can be estimated directly using fields, e.g., aviation and control theory. In control theory,
arrival time while moving at a constant speed. The arrival a Kalman filter is a recursive filter that evaluates system
time data from two base stations tighten up the circle for the dynamics and noise measurements of the state vector of a
desired position, and it is necessary to import data from a dynamic system model. The Kalman filter is a minimal vari-
third one in order to find the exact position. ance filter when it is noisy Gaussian and is minimum vari-
ance linear filter when it is not noisy Gaussian [24]. Several
3.3.2 Angle‑Based Lateration Techniques methods in literature are based on estimation of the distance
between two nodes for GPS localization [25–28]. These are
In addition, AOA measurement (also called Direct of derived from any of radio positioning techniques including
Arrival, DOA) is a method of determining radio frequen- Time of Arrival (TOA), Angle of Arrival (AOA), Received
cies propagation direction or high signal power in a series Signal Strength (RSS), and Time Difference of Arrival
of an antenna. The target is estimated with the intersection (TDOA) [29]. However, localization in radio positioning
of two angular lines, each of which is formed by the radius methods may result in two types of errors: multiple path
of one main station. AOA estimates direction by evaluat- effects or non-line of sight (NLOS) and hardware measure-
ing TDOA on the array elements. TDOA measurement is ment errors. Typical localization techniques integrate data
usually made in a way that computes the phase differences of GPS receiver and movement measurement of the vehicle.
of antenna array elements. In AOA, delayed arrivals are Nevertheless, these fail to produce high accuracy of locali-
directly measured for each point and it is converted to an zation from outdoor environments when the vehicle passes
AOA measure. They perform well in the situations of Line through a site producing multipath signals. Such a contribu-
of Sight (LoS), however, the accuracy and precision reduce tion of Kalman filter is undeniable.
when signal reflections occur. For high accuracy positioning,
at least two receiver sensors (triangulation) are required. In 3.4.2 Extended Kalman Filter
Fig. 3, AOA and TOA methods are presented [22].
As a kind of Kalman Filters, Extended Kalman Filter (EKF)
3.4 Filters estimates recursive states. If the system conditions and
observation models are linear functions of the state and the
Real-time localization problem has been adapted to land, initial uncertainty is Gaussian, Kalman filters may provide
sea and air platforms and state estimation methods which optimal solutions. EKF can be applied to nonlinear systems
consist of parametric filter approaches such as Kalman fil- and it may linearize them with the first-order Taylor Series
ter based Extended Kalman Filter and Distributed Kalman expansions. Also, EKF has computational efficiency because
Filter; and non-parametric methods such as Particle Filter. of Kalman filter [27]. But this feature may find out only
unimodal distributions and it can be applied to only local
localization. Moreover, it may fail in the relocalization of
nodes in VANET because of the difficulties in solving some
situations like symmetry.

3.4.3 Bayesian Filter

It is noted that the Bayesian filter is a common algorithm


for estimation; it determines the probabilities of several
beliefs to allow a vehicle to gather its position. The algo-
Fig. 3  a Angle Of Arrival (AOA) b Time Of Arrival (TOA) [22] rithm provides a framework to estimate the state of the

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

system based on probability theory. Bayesian filters can connections. Another advantage is that mobile devices con-
estimate a state of the dynamic system in noisy ambients. sume less power than a transmitter or a satellite.
The state is a person’s or location of an object and loca-
tion sensors provide information about the state in location 3.6 Image/Video Processing
estimation for pervasive computing. The state could be a
simple 2D position or a complex vector including 3D posi- Another method uses the captured image and/or video
tion, yaw, roll, pitch, and linear or rotational speeds [30]. data for positioning [34]. With this method, the geometric
This method generally has the capability of determining parameters of the vehicle including the width of the strip,
the orientation and position of a node in VANET [31]. the road gradient, the distance between the lane and the
vehicle, the direction of the vehicle and the camera tilt
angle can be precisely calculated. However, this method
can produce a huge amount of data and cause congestion
3.4.4 Particle Filter in the network. Moreover, the processing the images and
videos should be fast enough. Delays due to processing
The Particle filter is an implementation of the Bayesian time may decrease the accuracy of the positioning.
Filter and a kind of learning algorithm. It uses a sequen-
tial Monte Carlo method and it is usable for any kind of
nonlinear distribution and probability. Also, this method 3.7 Road Side Unit‑(RSU)
can be utilized in noisy systems. Monte Carlo methods are
used for updating density values over time. This approxi- RSUs in VANET are fixed infrastructure communication
mation is nonparametric and it may be more effective than nodes present at roadsides. As shown in the study, they
EKF with Gaussian. Also, this method can be applied eas- are individually used in types of localization, coopera-
ily in localization techniques. In this method, the size of tive and non-cooperative. RSU structure has a key role in
the particle set affects the accuracy of localization. So, VANET for different reasons. First, they send important
increasing the number of particles allows a more accurate data to the vehicles within the network. The second favor
localization. However, the localization cost will increase is that they transmit the received messages to end-users.
with this technique. The last advantage is that they have the ability to connect
the vehicles to the Internet. Since VANET is a network
fully mobile with dynamic topology and latency, RSUs
substantially facilitate communication as well. To consider
3.5 Cellular Localization all these, RSU structures improve VANET coverage and
network performance. Also, the communicative perfor-
This technique is based on using a mobile cellular infrastruc- mance of the network vehicles may depend on the number
ture in the urban environment to calculate the location of an of RSUs in VANET. Furthermore, RSU installation and
object. Cellular data may not be available everywhere. It operation increase VANET costs and partly aggravate its
may also vary depending on the network mobility related to execution. Therefore, a balance should be made between
the Doppler effect and the location of the connected object. RSU coverage area and installation cost. In this context,
Therefore, the cell coverage area should be large enough the idea of the limited number of RSUs to be used in
[32]. Mobile cellular systems require a communication infra- underpopulated settlements has become important [35].
structure consisting of a number of cellular base stations Apart from these, RSUs can not only send location data to
distributed over a closed region for working properly [33]. the vehicles in VANET but also communicate information
A cell typically uses a different set of frequencies than the on road, weather conditions and traffic situation.
neighboring cells in order to prevent interference and ensure The RSUs generally utilize high-frequency DSRC
the guaranteed quality of service within each cell. When communication (i.e. 5.9 GHz) to meet the requirement for
combined, these cells provide radio coverage in a wide a quick response like collision avoidance and have low
geographical area. In this regard, different types of port- latency. GPS signals can easily be broken or blocked, and
able transceivers, e.g. mobile broadband modems, pagers, that’s why the use of the RSU can help intra-vehicular
mobile phones, tablets, and laptop computers, can access the communication on highly dense traffic roads for more
network and communicate over the base stations at any loca- accurate localization. A critical security problem for
tion in the network. Base stations in cellular networks can localization techniques is to have accurate and reliable
be used for location estimation of above-mentioned devices. position data from the vehicles. Fake location informa-
One advantage of positioning with a cellular network is tion has serious effects on the performance and safety of
that a single transmitter can be used for a large number of

13
F. B. Günay et al.

location-based applications. The RSUs currently used are • Positioning Module A relative map is designed and sim-
available in different cities and regions throughout the plified.
world. • Globalization This module is where the related position
values are adjusted to other relative maps, which offers
the best choice of gateway and transfer of the data to the
3.8 Other Techniques other ones.

In the Nano Integration method, it is designed to use Carbon In the GeoLV technique, geometry-based localization can
Nano Tube Network integrated with nanodevices. They offer be performed for GPS outage conditions. In the present
millimetric precision in such areas as collision avoidance, study, the current position can also be found using the pre-
collision alarm, and following lane for VANET [36]. sent and past vehicular dynamics and road trajectories of
The location of a target is found by using a signal inter- the vehicles [39]. It is worth to explain another technique,
ference-noise ratio from Exchange messages in the first Monte Carlo Localization. This method is used in various
method of Weighted Localization (WL). In the second tech- studies and anchor box localization based on the min/max
nique, localization is made through the measurement of the values of node or vehicle in a coordinate system is essen-
distance between both SINR and neighborhood vehicles. tial. In other words, the intersectional area of anchor nodes
There is a detail for determining those weights in weighted is the potential area where a non-localized node can be
centroid localization (WCL). If a vehicle gets closer to the found and is relatively enlarged using the straight edges.
anchor node, SINR value of this connection and weight of Thus, mobile network localization can be provided in a
anchor node increases. There are primarily 3 general steps more realistic way and with high fault tolerance. Localiza-
in WCL: tion through the Monte Carlo technique is shown in Fig. 4.
Anchor box at any “t” time is defined as below [40]:
1. Each node periodically broadcasts beacon message.
2. Weights are assigned to nodes according to both
At = [x(t)min , x(t)max , y(t)min , y(t)max ] (5)
approaches above. where min and max values represent the minimum and
3. The identified weights will be used for the WCL locali- maximum operators for x and y coordinates, and in the fol-
zation. lowing figure anchor box made for three anchor nodes are
illustrated.
As for Optimized Weight Localization, all the nodes in the At places like tunnels, cities with large buildings, can-
Network essentially transmit beacons periodically. This yons, parking lots where GPS is not available or not suffi-
includes the data of sender id, position, turning angle, and ciently precise for local applications, the location accuracy
SINR and RSSI values. A table for neighborhood vehicles can be increased by using any known infrastructure such
are prepared using the beacon messages received as well as as Localization services, RADAR, UWB, WSN or WiFi
the containing information of them. The general weighted Localization.
function is equal to the sum of the parameters of distance
and turning values, and then location is made according to
these weights [37].
NLOS problem is dealt with in a different angle in the
Geographical Information Enhanced Cooperative Localizer
(GIECL) technique. First of all, area sampling is performed
by using a method built on geographic information including
building locations, and so NLOS links are described. Both
LOS probability of each link and the current vehicular posi-
tions are computed using geographic data. In another study,
the Relative Cooperative Positioning (ReCoP) algorithm
requires three modules [38].

• Partition Module Here, exchange of information is made


with neighborhood vehicles, and the evaluation between
them is realized through the AOA method. After that,
clusters are formed based on their status of the neighbor-
hood within the network, and one candidate for cluster
head is assigned to develop a cluster member map. Fig. 4  Monte Carlo localization [40]

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

4 Cooperative Localization (CL) estimates are shared with all neighbors and express the
probability density function of that node in VANET.
Cooperative Localization (CL) can be described as the pro- CL accuracy is not only influenced by the number of
cess of more accurate localization in a combination with techniques combined but also some other factors such as
different localization techniques with fusion. In cooperative their compatibility, landmark density in VANET and their
localization, nodes aim to find their location by measuring individual achievements. They are related to the hardware
relative distances with respect to each other in a coopera- of vehicles or locations. There are also several methods to
tive fashion. The additional information acquired from these be applied in the form of mathematics and software such
cooperative measurements between pairs of nodes enhances as filtering or algorithms.
the accuracy of the localization system [41]. In VANET, In the classification of VANET localizations, a dif-
there are some information sources for the localization of ferent methodology has been followed. In recent years,
moving vehicles. Firstly, the position and velocity of mov- many articles have been published about Non-Cooperative
ing vehicles can be provided with an odometer or vehicular Localization as well as Cooperative Localization. Besides
sensor measurements. Positions of other vehicles may be this, the rising number of studies on Cooperative Localiza-
also obtained and used. Finally, position measurements will tion has brought out the requirement to divide into GPS
track mobile vehicles, relative to the position of roadside based and non-GPS based ones. In this respect, the only
unit (RSU), or base station (BS). Some of the studies about classification by name of the techniques may not be ade-
CL are shown in Table 5 in Sect. 7. quate for this issue. Instead, techniques are subdivided by
Kalman Filter, Particle Filter, and Spawn Algorithm are their attributes or characteristics. Likewise, the studies on
important techniques in Cooperative Localization. Kalman GPS based techniques are classified by use of the filter.
and Particle Filters are explained in the Filters Section. Following this division, the techniques and the related
Spawn algorithm basically uses factor graphs and sum studies are summarized. In the NON-GPS based coopera-
product algorithm (SPA). In the Spawn algorithm every tive techniques, using built-in technologies or accuracy
node or vehicle first obtains position information from escalating techniques is highlighted to enhance the accu-
the last localization. Then a node receives messages from racy in VANET structures. In Fig. 5, a general classifica-
neighbor nodes in the network and updates its position tion of VANET localization techniques is presented.
estimation according to this new information. These new

Fig. 5  Localization algorithms classification in VANETs

13
F. B. Günay et al.

Table 2  Studies on filter usage


Type of filter Fusion type Requirements Accuracy (m) Landmark Den. References

Particle GPS+PF GPS 0.5 – [68]


Kalman GPS+Vehicle Sensors+KF GPS+Vehicle Sensors 1 – [44]
Kalman GPS+Vehicle Sensors+KF GPS+Vehicle Sensors 0.88–7.2 High [45]
Extended Kalman GPS+V2V+EKF GPS+V2V 0.12 High [57]
Unscented Kalman GPS+V2V+UKF GPS+V2V 1.58–9.8 High [63]
Bayesian GPS+CMM+BF GPS+V2V 2.74–10.47 Average [72]
Particle GPS+TOA+PF GPS+V2V+V2I 2.07–3.54 Average [70]
Particle GPS+TOA+PF GPS+V2V 2.28–3.83 Low [65]
Kalman GPS+IVCAL+PPD+KF GPS+Vehicle Sensors 0.05–1.89 Low [42]

5 GPS Based Cooperative Localization 5.1.1 Kalman Filter (KF)

GPS usage has more efficient and successful accuracy values Inter-vehicular Communication Assisted Localization Using
when cooperatively used with other methods. With this, it Previous Path Detection (IVCAL+PPD) algorithm [42]
is more useful than its single-use and contributes to the sus- makes localization real by integrating the estimates of GPS,
tainability of VANET localization in GPS outage. It means direction, speed and previous position in a hybrid way. In
that GPS location found just before outage is at the same the study, firstly the fusion of GPS information with the data
time the latest status auxiliary to localization immediately of direction and speed is made. If multipath is recognized
after the outage. GPS based CL techniques are shown in in the signal, the most efficient three neighborhood anchors
Table 2 in Sect. 6. in VANET are identified and positioning and accuracy
enhancement are performed once again. Then, the previous
position estimates within a limited time, such as the last
5.1 Filter Usage one minute, are gathered with the Previous Path Detection
(PPD) algorithm and averaged. By combining with the cur-
Both Kalman and Particle filter techniques are based on the rent estimate, the past data have been utilized. The algorithm
Bayes filter approach. In the section where filter usage is has not been tested in real life, however, the methods of
considered in VANET localizations, the systems including GPS, KF, IVCAL, and IVCAL+PPD are compared in real-
a variety of filters are explained. In Table 3 in Sect. 5.2, CL istic simulations. The most effective result is achieved with
techniques with filter usage are presented. IVCAL+PPD in an accuracy error of 0.05 m.

Table 3  GPS based cooperative Accuracy (m) Network density Landmark density Type of localization References
localization
0.02–0.4 Low Low UWB (Time Of Flight) [89]
0.8–7.2 High High Na [45]
0–3 Low High RFID [101]
4–7 Average Average IEEE 802.11p [102]
0.28–1.11 Low Low IR-UWB [59]
1.7 Low Low radio-based ranging sensor [71]
0.05–1.89 Low Low IVCAL+PPD [42]
2.07–3.54 Low Average TOA+PARTICLE FILTER [70]
0.8–4.0 Average High GPS+DSRC [109]
0.5–2.0 High High GPS+WAVE(IEEE [61]
802.11p)+MAP MATCH-
ING
0.7–6.1 Low Low GPS+RSSI [93]
2.28–3.83 Low Low GPS+TOA+PF [65]
3.81–15.75 High High GPS+DLEA+KF [96]

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

In [43], VANET localization is improved by non-coop- with an orderly IVCAL scheme. CWS-IVCAL verifies the
erative features like traffic lights, inactive cars or people. So reliability of GPS in urban environment and the flexibil-
it is called Implicit cooperative localization. Also, above- ity against wrong intra-vehicular distance estimates. Con-
mentioned features are fused with V2V communication sequently, it was concluded that the constraint weighting
links. Moreover, Gaussian Message Passing algorithm is scheme optimizes the localization process thanks to use of
applied for this study which integrates GPS measurements an innovative assurance measurement and described that it
with Kalman filter. Many VHF (Vehicle to feature) methods is effective for the soundness and accuracy of localization.
are used in this paper and those measurements are almost The subsequent vehicular position data in VANET sys-
noisy. So with the construction of consensus by applying tem are discrete, and therefore update frequency of loca-
those VHF techniques more accurate localization can be tion information should be efficiently decreased in order to
obtained. If those features would be applied increasingly, lessen the burden of communication and data management.
localization results could be more accurate. For instance, The next study [47] suggests that a location updating model
localization root mean square error has been about 0.2 m will be designed based on the equations taking into account
when 200 features were applied in Urban Canyon. However, the Kalman filter difference, and an updating strategy of
RMSE was acquired approximately 4 m for only 5 features vehicular position data based on Kalman filter, with repeat-
in urban canyon too. These features were applied in an urban ing package. A decision algorithm for repeating packages
canyon, but not in a rural area. to avoid package losses is designed to present the vehicular
The researchers in [44] propose to send collision alarm position updating data. In addition to this, the model not
to drivers by integrating the movements of the positioned only produces location updating information according to
neighborhood vehicles by means of vehicular sensors, GPS the predefined position but also makes decisions on package
and wireless communication. The vehicular sensors consist repeating mode of the distance between two neighborhood
of wheel velocity sensors, steering angle coder, and fiber vehicles. As a result of the simulation, it is shown that pack-
optic gyro. The information received from these sensors and age repeating location updating strategy can substantially
steering angle coder were combined in a Kalman filter. In the reduce Messages Generation Density (MGD) of the vehicles
study, a dynamic cycle model was used as an example, and and increase the Packet Delivery Ratio (PDR) to keep band-
in urban areas the data obtained at low and high speed dur- width, which decreases database problems and enhances the
ing a trip of 60 km were processed, such as stops, junctions, reliability of network communication.
U-turns, lane changing, etc. These are verified with the facts Rohani et al. [48] suggested cooperative vehicular locali-
that the filter makes dead reckoning, disseminates the speed zation based on a decentralized Bayesian approach that a
and involves the collision alarm. Namely, real-time position- vehicle can integrate GPS localization with other vehicles’
ing structures designed and deployed on the vehicles were GPS data and vehicular distance measurements. The aim of
compared. An extended Kalman filter was integrated into the study is to improve GPS localization and eliminate data
vehicular sensors, including wheel velocity, yaw rate gyro, dependency due to the fact that this is common in proba-
DGPS observations, and the steering angle. For the cur- bilistic approaches[49–51]. It has been also proven that the
rent study, the mean number of satellites is less than 5, and suggested method has no over convergence problem, unlike
hence GPS location errors are badly computed with greater other methods. The new localization estimates produced
than 2 to 3 m on average. A similar study aims to mini- using that method was employed a Kalman filter and the
mize the multipath problem depending on the localization results were compared to GPS measurement positions. Con-
precision of VANET vehicles. The suggested Inter-Vehicle sequently, it was shown that this significantly decreases the
Communication Assisted Localization (IVCAL) technique positioning uncertainty along with localization error.
utilizes communication between VANET vehicles to obtain The study of Ghaleb et al. [52] aims to develop a highly
more information from the vehicles in the neighborhood. accurate system under dynamic and unstable circumstances.
The simulation results indicate that the positioning error is For this purpose, innovation-based Adaptive Estimation
reduced up to 53% compared to that of the traditional tech- Kalman Filter (IAEKF) structure was used to help localiza-
niques [45]. tion measurements be utilized together with vehicular kin-
A study using KF structure [46] suggested the technique ematic information so that this was partly renewed according
of Constrained Weighting Scheme of Inter-Vehicle Commu- to VANET localization, named as EIAE-KF in the study.
nication Assisted Localization (CWS-IVCAL) that utilizes Based on the obtained results, satisfaction charts were estab-
intra-vehicular communication to enhance the accuracy of lished with regards to the localization conducted and the
vehicular localization. CWS-IVCAL fulfills the procedures variety of noise levels, and the EIAE-KF method achieved
in consideration of the uncertainty of dead reckoning for the the best outcome in terms of satisfaction.
vehicles with IVCAL as localization technique. The sug- The published paper of Sunil and Rekha [53] is on vehic-
gested method was tested through simulation and compared ular localization made through Kalman filters. In the study,

13
F. B. Günay et al.

they allow to improve the results without filling the memory given for comparison. Naturally, the system with low rate of
obtained using Kalman channel. The penultimate situation error provides better localization. The study of Hoang et al.
is not kept in memory, which is a key advantage for built-in [59] aims to reduce localization errors through integration
systems in which GPS and video images are used together. of GPS positions with V2V IR-UWB distance measurements
The researchers in [54] emphasize the observation of corre- and turning information from the IMUs. In addition, it was
lation between successive localization measurements. Then tried to decline Geometric dilution of precision (GDOP) in
those measurements are integrated for modeling localization localization. Different problems including distance-based
error. Yule-Walker equations determine the correlation of CL limits, IMU turning integration and lane borders inte-
past and future positions of vehicles. Also, p-order Gause- gration, were intended to solve based on the applied method,
Markov method is applied for future location prediction of using definite-time distance measurement with IR-UWB
vehicles and utilizes past positions of vehicles. That method technology, modified EKF, quality gyroscope, camera and
compares the datasets which contain the mobility traces of digital map. Localization results were obtained through the
two vehicles. above-mentioned technologies used in the problem solving,
Another paper [55] proposes a plan to identify the pres- individually or in group. According to these, the package for
ence of multipath conditions and mitigate their effects on which all technologies are utilized and GPS+UWB+IMU
vehicular localization using the suggested IVCAL localiza- package are so close to one another in localization perfor-
tion technique. In order to achieve maximum accuracy, the mance. Plus, the simulations are not performed for different
location uncertainty indexes of vehicles exchange with the speeds.
position data.The improved localization performance was Golestan et al. in [60] proposed a method for localization
evaluated in both open areas and urban or canyon environ- through vehicular movement model-based (dynamic model)
ments [55]. V2V communication and data fusion. The system designed
specifically for this purpose includes context-aware informa-
5.1.2 Extended and Semi‑Extended Kalman Filter tion processing unit (CAIPU), data fusion unit (DFU), and
(EKF‑SEKF): Cognitive Gateway. DFU combines data from sub-levels and
transmits them to the CAIPU together with an estimate Ack-
Relating to this method, Jaiswal and Jaidhar aimed localiza- erman movement model was applied, which does not just
tion for nonlinear vehicle movement using Extended Kalman consist of gas and brake but also have such information as
Filter (EKF) in their study [56]. The nonlinearity indicates velocity, steering angle, etc through sensors. AOA and TOA
the speed variability of vehicles inside a city. Nonlinear measurements are used for localization, and communication
vehicle movement was analyzed especially in an urban envi- is provided using Dedicated Short Range Communication
ronment. It was assumed in the study that all vehicles have (DSRC). For estimation of all vehicles’ current situation and
an omnidirectional antenna and communicate both RSU location, Bayesian filter, Extended Kalman Filter for recur-
and also other vehicles through the IEEE 802.11p protocol. sive state estimating, and Unscented Transform methods are
Additionally, it is assumed that the values of latitude, longi- utilized to make adjustments to the results. Furthermore, the
tude, velocity, acceleration, and steering angle are observed combination of TOA and AOA is a significant difference. In
by GPS and inertial navigation systems on the vehicles. The the study, two test cases were made in itself: one of them is
study professed that the EKF method is more efficient than straight road and the other is curvy road. And the simula-
KF, and then the comparison was made between these two tions were conducted using 5, 7 and 9 cars, and the least
methods and better localization results obtained. Further- localization error is 0.77 m.
more, it was observed in the localization through EKF that In study [61] all the nodes have synchronous Positive
the simulation results for the highway environment were Orthogonal Code (POC) MAC to deliver safety packets
more successful than those for urban environment [57]. In to other nodes. Each node has one POC codeword. It is
another study [58] trace traffic times for intra-vehicular dis- assumed that all VANET nodes are both frame-synchronous
tance estimate and reduced inertial sensor system (RISS) and slot-synchronous. They can use TDOA due to the known
are utilized for localization as integrated. Location errors times of transmitting and receiving the messages synchro-
from this integration are limited by means of the applied nous. Only a semi-extended Kalman filter was used in the
EKF method. The paper employed scenarios in which sev- study, in which both the received GPS locations and the
eral vehicles could be connected to GPS for a short time. approximate distances between the vehicles through the
The localization system additionally requires no synchro- POC were utilized. The study was tested in two cases: one
nization in the use of trace traffic, and this feature makes a is the situation of single road and low dense traffic, and the
difference. The percentages in VANET of the vehicles with other is multipath with different traffic density. In both, it is
the ideal GPS update were ultimately 25, 50, 75, and 100. assumed that the intra-vehicular communication range can
The study results differently evaluated (0 to 5m) are also be maximum of 50 m, and the vehicles are heterogeneously

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

distributed on road. Both have 5 types of results: Semi- was suggested to integrate GPS absolute localization with
Extended Kalman Filter with map matching, Parker’s the data computed using a vision system producing the
method with map matching, Semi-Extended Kalman Fil- position and heading information of the vehicle on road.
ter without map matching, Parker’s method without map A great number of outcomes are acquired through the sys-
matching, and VLOCI2 algorithm. The accuracy ranking tem deployed onto an empirical vehicle on real roads and
was simulated as 0.8, 0.9, 1.0, 1.3, and 2.0  m, respectively. in real-time, and the best standard deviation became 48 cm
In the multipath situation, they are found as 0.5, 0.6, 0.7, 1.0, along with road axis and 8 cm along with axis normal. These
and 1.6 m, respectively. It is arguable of how much possible results are found by making applications along the road of
is to employ TDOA in such a simple way. 4 km in an urban area including a curvy road as well as
single-and double-lane roads. The obtained results indicate
5.1.3 UKF the requirement of integration of the information from differ-
ent sensors to mitigate the uncertainty of and the non-access
In [62], a new approach is suggested for vehicle localization to GPS data for vehicular localization. In another study, this
of using Unscented Kalman Filter (UKF). The kinematic method is also applied. [69].
parameters of vehicles (i.e. position, velocity, movement, In [70], simulations were performed with 3 to 11 nodes.
acceleration, etc.) are regarded as a multi-dimension datum. The most successful method in all cases the particle filter
As a result of the simulations, it was shown that the sug- method was followed by the GPS and TOA measurements,
gested approach could be more accurate and more reliable respectively. Particle filter for 11 nodes with 2.49 m error
than GPSIDR system and Extended Kalman Filter (EKF) and 4.32 m for 3 nodes. Also, GPS system is considered
based solution. For vehicular localizations, the 10 m preci- to be disabled and the simulation was also performed for
sion of GPS is insufficient to apply it for VANET’s security this situation. 11 nodes and 3 nodes results are 5.13 m and
applications. In the proposed algorithm, both vehicular data 6.25 m respectively. Again a simulation in a tunnel was
and movement models are taken into account, and UKF is made and where 3-4 nodes and RSUs are used. The particle
used to achieve better localization accuracy. In the study of filter and only TOA have 3.86 m. and 17.25 m. accuracy
Cruz et al. [63] UKF was employed as a second-step filter. respectively [70].

5.1.4 Particle Filtering 5.1.5 Sequential Monte‑Carlo Probability Hypothesis


Density (SMC‑PHD) Filter
An alternative approach is proposed in [64] to perform
vehicular localization through the RSU structures inside The researchers in [71] propose to make a cooperative,
tunnels. The main difference is the necessity of the locali- multi-sensor, multi-vehicle, high-accuracy localization. All
zation without GPS connection. However, the obtained the vehicles contain GPS-module, radio-based distance esti-
data from Particle filtering and Inertial Measurement Units mation systems. The intra-vehicular communication distance
/ Wheel Speed Sensors (IMU/WSS) were combined based is 300 m. To determine the target status and number through
on localization. Different localization infrastructures were the Bayes filter, the condition of the vehicles is predictable
tested within ToF-based ranges and with ITS-G5 RSS in a with Sequential Monte Carlo-Probability Hypothesis Den-
tunnel. Localization errors were found depending on their sity (SMC-PHD) filter. Maximum RMSE is found as 4.7 m
filter-fusion situations are between 0.1 and 2.0 m. when the standard deviation is 0.83, and RMSE with the
Particle filter and weight update were used for improv- standard deviation of 0.26 as 1.7 m. Furthermore, it was
ing accuracy and based on absolute position, vehicle motion observed that more accurate localization was made than the
and mutual positioning in this study [65]. TOA was applied only GPS option. GPS usage in every vehicle can be consid-
for range finding. TOA, GPS and TOA, GPS, Particle with ered as a disadvantage.
particle filters were shown in simulation results. Results are
categorized into two kinds, GPS and GPS outage. For GPS 5.1.6 Bayesian Filtering
results, 3, 5, 7 and 9 nodes are used in results and 3.83,
3.30, 2.76 and 2.28 m positioning errors are acquired with To investigate Bayesian Filtering, the related study [72] aims
particle filters respectively. On the contrary GPS, outage to develop a cooperative map matching method (CMM),
caused less accurate results and 13.04, 9.68, 9.21 and 7.41 m obtain pure GPS data between vehicles and design a precise
were obtained for position errors respectively. Simulation map. At the first step, the vehicles transmit GPS data to
was tested for Istanbul with Sumo. another through CMM and VANET. The other vehicles using
The closed localization systems have been reached up this information improve localization by means of more pre-
to the still acceptable performance [66, 67] and outdoor cise mapping. As a second step, dynamic base station DGPS
localization have remained a challenge. In this study [68] it will be utilized, which help the vehicles, particularly those

13
F. B. Günay et al.

which newly join VANET, to correct their GPS Pseudorange other strategies for all the metrics studied including Package
errors in a decentralized way. That pseudorange errors are Delivery Rate, MAC cost, package delay, and average route
distance measurement errors between GPS satellites and length. In the study of Abumansoor and Boukerche [75],
vehicles. The difference of this study from others is that The NLOS condition is dealt with, and location verifica-
there is no need for distance information between the related tion protocol between the neighborhood vehicles to provide
vehicle and others, and the pseudorange errors of different VANET localization services. The position of any node can
vehicles and mapping uncertainties are eliminated through be verified as well as the safety measures for the improve-
the CMM and DGPS. In the best scenario to compare with ment of data integrity through this multihop-based approach.
the only GPS and Single Vehicle MM (SVMM), location
error was found as 14.06 m for GPS, 5.02m for SVMM, and 5.2.2 RSSI in GPS‑Based Cooperative Localization
2.74 m for CMM with 10 vehicles. The method was simpli-
fied with a lack of intra-vehicular distance measurement, One other preferred technique is RSSI in Non-Filter usage
and however it is remained unclear whether localization is like the other methods. The first study in this kind [76] sug-
more successful when it is applied. There is an obligation gests a distributed algorithm that is made through a radio-
that every vehicle has a GPS module, and it is economically based spacing technology, using control movement estimates
disadvantageous as well. in order to realize vehicular localization between the neigh-
In [63] localization has been tried to be performed with borhood nodes. In this method, the algorithm establishes a
both vehicle communication and mobile phone sensors. The cluster minimizing excesses of Euclidian distance between
two-stage Bayes filter technique was used for localization. the vehicles because the intra-vehicular distance estimates
The first filter is the Kalman filter used to estimate the angle include noise factor, and helps a position of a vehicle within
of rotation thanks to the sensors in the mobile phone, and this cluster to be accurately localized. The suggested algo-
the other filter is the particle filter structure that combines rithm works in three steps:
the map information, GPS location, and inter-vehicle RSS
measurements. As a contribution to the study, the location 1. Initialization : Each vehicle evaluates its distance from
of other unknown vehicles in the absence of GPS is to help the neighbors and then shares this information with the
localization and real-world data is tested with urban sce- other vehicles.
narios. Urban scenarios are also supported by 30-minute 2. Attenuation : Each vehicle sensitizes the initial estimates
driving. Localization errors have been between 1.58 and to obtain a final estimate of its neighbors’ positions. The
10 m. [63]. aim is to minimize the residuals of the intra-vehicular
The researchers in [31] propose to use both EKF and PF Euclidean distance and the measured length.
which finds good results even in sensor failures. GPS sen- 3. Iteration : In a mobile environment, the first two steps
sor is used for some time to measure the accuracy of the are repeated per T seconds.
proposed model.
Consequently, a new distributed solution was proposed for
5.2 Non‑Filter Usage localization to be used as a collateral way to GPS, and it
was shown that the GPS accuracy can be improved using
5.2.1 Dead Reckoning (DR) radio-based spacing technology (i.e. RSSI). Even in greater
error levels for vehicular localization (e.g. GPS accuracy
While DR can provide more accurate and robust localiza- underestimating localization), the algorithm was shown to
tion, the disadvantage is their uninterrupted and real-time come closer to its real position with relatively higher accu-
information which have received from sensors. The study racy. The researchers in [77] propose to exploit RSSI for
using Position Based Forwarding (PBF) [73], analyzed the position estimation via vehicular networks. Also, the pro-
effects of DR strategies on location accuracy between bea- posed method is a self-correcting technique that works on
cons. For this purpose, simulations are performed to inves- different topology scenarios and tries to develop techniques
tigate the effect of different DR strategies on the protocol for reduction of localization errors. Self-correcting locali-
performance and the relation between cost and performance zation scheme was used for solving important problems
is explained. In the previous study [74] it was shown that of RSSI. RSSI limitations are destructed with this method
Position-Based Forwarding is superior to Topology-Based in sparse networks. Generally, localization with lateration
routing methods using vehicle highway scenarios, and this needs 3 nodes at least. SCL-VNET method provides the
was compared to Active, Passive Anticipation and Defensive necessity of 2 nodes. According to this method 2-hop, near-
Neighbor Selection. As a result, it was observed that Active est nodes to 1-hop nodes should be used for localization.
Anticipation reduced average localization error to less than There are 3 phases of this process. In the communication
50 cm and showed that it exhibited better performance than phase, communication and listing processes of all nodes are

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

realized. Selection of 3 1-hop nodes or less than 3 1-hop is determined to depend on such situations as Sybil devia-
and farthest 2-hop beacon nodes are made in the estimation tion, observation period, witness number, etc., however, a
phase and finally correction and alignment phase is carried localization accuracy was not explained [79].
out. According to simulation results, 0.44 and 0.42 m are
acquired for average localization errors in urban and subur- 5.2.3 Map Matching
ban area respectively. Above mentioned results are valid for
high traffic density but average localization error increase to The project named Cooperative Vehicular and Infrastruc-
0.6 m to moderate and low traffic density. ture System, have the following functions: It offers a new
To have a safe and rapid transportation system in the localization method to make the connection between vehi-
VANET structure, it should be known of where traffic prob- cles, pedestrians, roads, and traffic data. Also, it shares traffic
lems may occur including any vehicle or crashed or broken information which is taken through wireless communica-
vehicle. Like tunnel, in these cases, localization problems tion, distribute local map data of RSU and Wi-Fi and uses
occur and then a method without GPS should be utilized map matching algorithm with the On-Board Units (OBU)
to localize vehicles. In a relevant study, a new technique equipped with GPS receiver and Wi-Fi devices [80]. This
is proposed to be a solution for this issue and based on the method helps vehicles to find the true path and also the true
communication interval. For the Optimized Disseminating lane even on the single-lane roads. Map Matching method is
Alarm Message (ODAM) algorithm, a simulation was real- used also in the study of Abumansoor et al. [75].
ized using 150 nodes on a two-way road of 20 km. In the In their study, Bauer et al. [81] evaluated shadow maps
simulation, speeds are different. The best result was obtained for safe and real-time vehicular localization in urban areas,
in the way that 40% of the nodes are equipped with GPS in the reference of shadow mapping rather than classical
[78]. digital 3D maps. The proposed approach proved that single-
Another published paper on this issue aims to create frequency GPS receiver enhanced the accuracy and integ-
a lightweight security system and determine Sybil nodes rity without additional physical sensors in urban areas. The
through a distributed and localized approach. The useful evaluated results are generated from different test drives.
characteristics in this technique RSUs are and traffic pat- Nowadays, accurate localization is a key necessity in the
terns. The method can be divided into three steps: areas of comfort and safety for the intelligent Transporta-
tion System (ITS). NLOS measurements are fairly made and
1. The nodes periodically transmit beacon messages, can be derived from the positioning solution. Shadow map-
receive them from all their neighbors, and signal ping designs a compact map to explain signal coverage for
strengths of the related messages are kept in memory. each satellite in a certain region under one service provider.
To realize this, there are three sub-steps. Claimer: Each Afterward, that is transferred onto the vehicles available in
node transmits a beacon message for the discovery of VANET by a RSU. As a result, the study has introduced a
neighbor in a periodical way, and the message has the concept of shadow maps to efficiently display satellite signal
position and id data such as GPS Position. Witness: It coverage in urban areas and developed an interval of locali-
measures signal strength and memorizes the related zation accuracy and estimation integrity. Furthermore, it is
information to the neighbors. Then, a beacon is trans- highlighted that the maps in shadow can substitute for the
mitted with an attached list of neighborhood informa- conventional digital map databases, with modern VANETs,
tion. Verifier: After receiving the beacon message, node which requires its deployment onto the vehicles and keep-
stays at the verifying interval and during this period of ing updated.
time, it can more accurately achieve localization than In another study, a multi-sensor fusion strategy is pro-
that made by claimer, comparing with the messages of posed for a new path matching technique designed to
witness process from the neighbors improve the characteristics of real-time navigation in the
2. A node can find more accurate location of its neighbors advanced driver assistance systems. [82]. This is a useful
following after collecting sufficient signal strength data. approach to eliminate uncertainties in the path matching.
3. If the neighborhood node at the second step is recog- Multi-sensor and multi-path estimation have been made
nized as being Sybil node, it will try to find out whether using a hybrid Bayesian Network. In fact, real-time locali-
the other nodes are Sybil using node classification algo- zation on digital path maps helps to drive assistance mod-
rithm computed. ule for accurately position the vehicle, facilitating processes
(i.e. route estimation, Adaptive Cruise Control (ACC), col-
There are two different scenarios for simulation. In the first lision alarm, lane departure warning, etc.) and promoting
one, the maximum signal distance is 200 m, its distance at Advanced Driver Assistance System Applications (ADAS).
the claimer point is 200 m, and the number of witnesses is The path matching problem is a localization problem based
escalated to 30%. As a result of the study, the detection rate on dynamic vehicular localization. In the previous studies,

13
F. B. Günay et al.

errors were not explicitly modeled, the methods were com- suggested such as Kalman filters and Markov Localization.
prised linking dead reckoning and path geometry. The topol- Out of Markov techniques, Monte Carlo localization algo-
ogies proposed by Taylor, Blewitt, and Greenfeld are used to rithms (MCL) are remarkable, renowned as Particle Filters
eliminate outliers [83, 84]. or Condensation Algorithms [88] which function by repre-
The main contributions of the study are the formaliza- senting the posterior estimates of potential locations with the
tion of path matching technique in the context of Hybrid weighted samples or the particle clusters [89].
Bayesian Network and its empirical verification with real
facts. Another aspect of this approach is its flexibility and 5.2.6 TOA and TDOA in GPS‑Based Cooperative Localization
modularity so that we can easily integrate other sensors.
The researchers in [90] propose a location verification algo-
5.2.4 AOA in GPS‑Based Cooperative Localization rithm is proposed to prevent specific attacks in VANETs
(e.g. spoofing attacks). The suggested algorithm, Coopera-
In a study of this technique [85] it was aimed to integrate tive Location Verification (CLV), is primarily based on the
the cooperation VANET scenario well-designed through a use of two vehicles (Verifier and Cooperator) to complete
signal processing at a low level on the physical layer, and the verification of one vehicle (Prover). Verifier and Coop-
hence an adaptable system to varying road conditions was erator transmit a message to the Prover, and it disseminates
designed, in which data exchange is minimum. Therefore radio frequency-based location information. According to
the covariance matrices of adaptive control with the EKF simulation results, it was proven that that approach is better
method and the variables of beacon package purchases were than Secure Location Verification (SLV) and Greedy For-
used. The study suggests that it is effective and successful warding Algorithm (GFA).
even in the environment of lightweight communication and Agarwal et al. used [91] the methods of TOA (time of
at low computational complexity and remarks that it is dif- arrival) and TDOA (time difference of arrival) for VANET
ferent from this aspect. localization. In the scenario tried to realize, vehicular cloud,
roadside cloud, and central cloud servers were utilized for
5.2.5 Ultra Wide Band (UWB) real-time road safety. By means of OBU and vehicular cloud
networks, localization can be achieved in the spaces with-
The relevant study aims to fulfill localization with data out GPS like tunnels. The vehicles within the clustered net-
fusion, fed by the local and VANET neighbors utilizing works act as service providers for other vehicles with OBU.
UWB (TOA) measurements and GPS-oriented information The used technique for localization is virtual beacon-based
[86]. In the paper, three methods are primarily used: DGPS, method indeed, which vehicle B that can know its own loca-
Tight CL, and INS aided tight GP. In this context CL can tion through GPS outside of the tunnel communicates to
be made, either debug from pseudorange errors or Kalman vehicle A inside the tunnel and estimates their distance and
filter or more robust CL through Doppler Shifting computa- upon three repetitions of this way, can find out the posi-
tion are conducted, and the tight-CL method is employed. tion of vehicle A. However, heading estimate can be made
It is seen that better interpretation and conclusions may be with built-in cam in the OBU. In the study, it was described
dependent to the number of GPS satellites. As expected, the that OBU tech continued to be improved, and additionally,
increased number of satellites has positive effects on the a mobile application was developed to transmit the relevant
results, and however, as it rises, the achievement rates of data to a range with a radius of 10 km via central cloud
these methods get closer to each other. As understood from system and to cover the position and report to the police.
this, the method was designed in tests with a satellite number In the study [92] it is aimed to perform localization with
that can be accepted as low as 5 became more successful no need for GPS by means of clustering and link prediction.
than the other one. The effectiveness of T-CL with UWB TOA and TDOA methods are employed for inner-tunnel
was compared to the other three methods: DGPS, T-CL and localization when GPS is unavailable. Unit Disk Model is
T-CL with INS. Localization accuracy was found to be 1.32, the channel model applied to the study. According to this
1.40, 1.42, and 1.52 m, respectively. Tests were conducted in model, two vehicles can communicate to each other as long
an NLOS environment, which is a disadvantage. Presenting as their Euclidean distance is less than communication
the measurement results owing to the number of GPS satel- distance. Thus, these can pass by one another and remain
lites is considered a true approach. unstable. In the process of link prediction, four attributes
The probabilistic approaches used for localization prob- are utilized: degree centrality (or the number of neighbor-
lem are based on a posterior probability distribution estimate hoods with other nodes), closeness centrality (or the inverse
out of possible vehicular locations. One of these approaches proportionate of the average of the distances from the related
is Kalman filter [87] that makes Gauss distribution reason- node to all the other nodes), betweenness centrality (or the
able. Different variations on this matter were previously ratio of the shortest path communication through the related

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

node), and straightness centrality (one of centrality indexes). In a thesis related to the method [95] it was examined
Centrality based approach differs the study from others, and with types of Weighted Centroid Localization used to get
the disadvantages of the study are lack of conclusion and no weights applied. The first technique is Link Quality Distance
numerical result obtained. (LQD) in which Heading and Map are utilized to determine
Another relevant study [93] aims to control the posterior them for vehicles based on the quality of intra-communica-
unmanned vehicles with WSN for a mobile vehicle fleet tion channels. LQD is calculated with SINR/di. For Head-
on which the only first one has a driver. The system was ing, if vehicles are moving towards the same direction, the
designed in the way that the whole fleet follows the lead- related weight element is recorded as 1 or otherwise 0. For
ing vehicles along the path via WSN. In this pursuit, it was map, if they are in the same map, it is 1, or if not, 0. In the
assumed that the fleet primarily had two types of mobil- paper, it was anticipated that RSS errors could occur and
ity. The proposed Straight Path technique is used when the weight identification was made with Fuzzy Logic. Beacon
Anchor nodes located through GPS available on the fleet messages are transmitted from the node to all the others and
are in the same line. If not, or in the case of road bend, tri- a localization chart is established through them. A fuzzy rule
lateration techniques are applied or mobile nodes between a base is made with RSS (distance) and Heading values from
greater number of anchor nodes are positioned. the received signals and the result appears as the weight fac-
In trilateration or multilateration, the least mean squares tor. Finally, the position data taken are adjusted through data
(LMS), Tulip and Cramer–Rao techniques were used. verification for better localization. WCL, Fuzzy WCL, and
Cramer-Rao is the major method to produce the most accu- Mobility Data Verification WCL methods were compared
rate results out of the applied ones and becomes a success- with each other. Those three methods were compared to cen-
ful method especially for the vehicles having 1 m following troid localization and accuracy errors were presented both on
distance. At this point, the location errors made approxi- average and at the maximum that average values were 22, 21,
mately ranging from 0.7 to 3.5 m can be considered a suc- and 1 m respectively and maximum levels were 60 and 1 m
cessful result. As the following distances between vehicles for WCL and Mobility Data Verification WCL respectively.
increase localization errors increase too. Tulip method is In [96], Weighted Least Squares Double Difference
another method to apply to the system through trilateration. (WLS-DD) and Distributed Location Estimation Algorithm
Localization errors were found to be variable between 0.90 (DLEA) algorithms are proposed for improving the accu-
and 4.0 m. It was observed that LMS method has the highest racy of localization. First, WLS-DD detects the intervehicle
localization errors between the applied techniques for mobile distances based on the sharing of GPS pseudorange meas-
vehicles. Localization errors were observed to be nearly 0.85 urements and a weighted least squares method. Based on
to 6.10 m by varying distance levels. In addition, the dis- this distance information and the sharing of inaccurate GPS
tances of anchor nodes to mobile nodes were arranged as 15, pseudorange measurements, DLEA improves the accuracy
25, and 35 m, and the lowest levels of location error were of vehicle positioning. The results show that 4–15 m locali-
determined using the TOA method as 0.853, 1.5702, and zation error of GPS can be reduced to 1.5–6 m by employing
1.9671 m respectively, depending on the above distances. this method.

5.2.7 Weighted Localization (WL) 5.2.8 Grid‑Based On‑Road Localization (GOT)

In [94] two weighted localization (WL) methods were To overcome the unavailability of GPS signals, this study
applied to VANET. The centroid localization diversification [97], suggests that the vehicles with or without GPS signal
is different from other studies. The methods of Weighted on low-density traffic or in tunnels.Vehicles should be inte-
Localization (WL) and Weighted Localization Distance grated into VANET, change location, and distance informa-
(WLD) dealt with in the study were compared to Centroid tion , and have a new Grid-based On-road localization sys-
Localization and Relative Span Weighted Localization tem (GOT) be developed to assist one another. This method
(RWL) in terms of accuracy, WLD accuracy error was found evaluates several fuzzy geometrical relations between the
less than 30 m, the second-best method was RWL with the vehicles and applies a grid-base approach to estimate the
accuracy of a bit more than 30 m, and they were followed by vehicular position, and thus beacon nodes and error propa-
WL and centroid localization respectively. As a simulation gation, two significant challenges in path localization, are
environment, a Manhattan Grid (3X3) with an edge length handled. Consequently, it is shown that the proposed model
of 1000 m was selected, and active working time was 30 offers high localization accuracy as well as a little burden
minutes. Examined in the study, centroid localization and its of communication and estimation in both tunnel and city
derivatives, WL and WCL are the fast estimation methods scenarios.
based on the rapidly changing VANET structure, which is Within the vehicular range, the system receives the loca-
a key advantage. tion and position information and these are transferred if its

13
F. B. Günay et al.

distance to the related node is greater than the communica- even Non-GPS vehicles can locate themselves by means of
tion threshold. If a vehicle simply knows the positions of RFID readers that they contain, using the position informa-
and the length from the neighbor vehicles, it should have tion from RSU or RFID tags on other GPS vehicles. The
at least three vehicles’ location data to activate Reckoning authors showed with both simulations and the tests made
mode. This feature constitutes the essential of the previous on test-beds of a real urban environment that the system that
localization papers [98, 99]. The tunnel scenario of GOT they have suggested reduced GPS –oriented errors. With
was evaluated together with city and hybrid scenarios on the test results, the best performance is obtained with the
the real maps [100]. According to the simulation results, in deployment of one 4-tag cluster per 5 m. The weaker side of
the tunnel scenario, although distance measurement error is this method is the requirement of additional RFID devices
at a high level, 30%, a vehicle 100 m farther from the tun- for their placement on roadsides.
nel exit can locate itself with an error less than 12 m and a
delay less than 15 m. In the city scenario, on high-density 5.2.11 Tight Integration
traffic, only 10% of the vehicles can estimate their position
and when distance measurement error is at a high level, 30%, A new CL method is proposed to improve relative localiza-
the proposed system measures localization error less than tion between two vehicles within VANET using the received
10 m. 40% of the vehicles have location information while data from GPS [102]. This method is similar to differential
more than 90% of the other vehicles can effectively manage solution principles such as differential GPS (DGPS) and
to locate their positions. however shows better performance than DGPS and increases
relative localization accuracy and precision nearly by 37%
5.2.9 Nano Integration and 45% respectively. Technically for localization estima-
tion, several radio techniques were used, including the differ-
In the paper of Nasr and Mohamed, the aim of the theoretical ence between nodes, time of arrival (TOA), Time Difference
model is to achieve a higher accuracy rate and performance of Arrival (TDOA), and Received Signal Strength (RSS).
level along with low cost and energy consumption through This method [103] is very similar to the above study and
vehicular localization. In a article [36], this suggestion pro- it’s analysis was made based on the number of satellites
vided theoretical researches and mathematical formulations, connected with GPS module. Also, individual results were
and simulations were made except for a physical applica- obtained only for GPS outage times. Accordingly, accu-
tion. An integrated nanotechnology model was developed racy errors were between 1.3 and 1.52 m and dramatically
to determine exact positions in the near distance. It was increased 15 s after GPS outage.
aimed to make a-few-meter distance estimate with one-or-
two-millimeter deviation. Furthermore, a theoretical model 5.2.12 Real‑Time Kinematic (RTK)
was developed for Vehicular Carbon Nano Tube Network
(VCNT). According to a study on this method [104] it is consid-
ered that the GPS itself is insufficient for high precision in
5.2.10 RFID VANET localization. Therefore, localization was aimed to
have higher precision in the exchange of GPS locational
VANET applications including safe driving and disaster measurements and the evaluation of their differences by
recovery typically involve high position accuracy. However, means of DGPS and real-time kinematic calculations. For
the conventional Global Positioning System (GPS) cannot VANET communication ITS-G5 protocol was utilized,
provide this high-level accuracy. Apart from GPS, Lee et al. which helps transmit messages with 10 Hz and dynamic
put anchors on roadsides using the DGPS concept in order to information such as acceleration and speed. Test roads,
discover the GPS errors [101]. These anchor points are the county routes and highways were assigned as application
RFID (Radio Frequency IDentification) transmitter devices areas, and tests were conducted at intervals of 30 km/h, max
placed at the roadside which propagate the data of physical 100 km/h and 80 to 180 km/h. Correction values were got
position, and this method is called RF-GPS. The proposed between 0.08 and 2.96 m. Naturally, the designed the real
system computes GPS errors using location information time kinematic system has a greater part in these location
from static RFID tags with a range of 3 to 4 m. Those RFID error corrections.
tags are disseminated to the neighbor vehicles and hence
they also recognize the GPS errors. This is useful infor- 5.2.13 Interference‑Aware RSSI‑Based Localization
mation for them because similar deficiencies occur on the (INTERLOC)
GPS receivers in the vicinity. The main contribution of this
study is to exclude signal strength and propagation char- The researchers in [105] propose an RSSI-based localization
acteristics for localization. Through the proposed system, in which other mobile nodes are used for localization of a

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

mobile node. Also, it was aimed that that node can adjust lead to non-deliberate localization errors in the case of error
itself according to heterogeneous ambient noise sources. In accumulation and when the initial reference point is wrong.
addition, a system was designed to discover the attacks of Differently, it used a webcam to estimate the intra-vehicular
Sybil nodes. To know the noise levels, it is important to distance and was able to manage thanks to knowing the ID
identify all the initial levels and then make it adapted to and the recorded scale of the vehicle. The vehicles’ ID data
keep pace with the variability on the mobility-based levels, are transmitted via DSRC, and their identification is made
which is caused by noise grades and may update localization through a vision scan. For example, one vehicle has a plate
results. After being updated, all nodes make an advertise- numbered 4567-AB, and in this case, it is enough to visually
ment about their positions. The received locations are con- identify the number of 4567 and math with the ID. By means
sidered Sybil nodes if not within the updated localization of the IPC algorithm designed in the study, it is aimed to
area. In the study, there is no need for a reference point like remedy these deficiencies. The IPC algorithm has primarily
RSU, which makes the system more economic. A compari- the following steps:
son was made with POEST and FRIIS methods. In POEST,
vehicular localization is performed using the RSSI values 1. The GPS positions of all the vehicles are collected, and
from the RSUs. FRIIS offers distance measurement with the the relative measurements for angles and distances are
received signal power to the transmitted signal power ratio, made.
however neglecting noise sources and path losses. INTER- 2. New circle radius values are found, which helps intra-
LOC became more effective than these two methods, with a vehicular distance to be calculated.
detection success rate of nearly up to 60% on a 900-vehicle 3. According to the resulting status, new positions are
environment while the others brought success with about organized.
40%.
Moreover, the study not only explains the methodological
5.2.14 Multi Dimensional Scaling (MDS) theory but also exemplifies it in quantities. The IPC algo-
rithm developed was compared first to the single GPS and
For the convex optimization problem, multi-dimensional then to both GPS and the previously used method of the dis-
scaling (MDS) is a method to present a variety of data tributed location estimate algorithm (DLEA). Hereunder, out
objects as a geometric graph. MDS is also a method in which of the average location errors found in the tests conducted
measurement is made in a multi-dimensional space com- for the intravehicular distances of 5 to 30 m, the lowest rate
mencing with Dissimilarity matrices, especially distance was observed in the IPC, for 5 to 10 m DLEA provided bet-
matrices. In a relevant study, the convex optimization based ter localization than GPS and for 15 to 30  m GPS was more
real-time vehicular localization was conducted, in which precise in localization. For example, for a 30 m distance,
time is divided into slots. In general, less-dimensional the location errors of IPC, GPS and DLEA methods were
Euclidean space points are embedded and point-to-point respectively found 7.67, 9.25, and 12.32 m. Its advantages
distances are considered as a real dissimilarity. Accord- are being real-world tested and the ingenuity of the method
ing to the obtained localization results, the lowest RMSE and the disadvantage is that the tests were conducted at low
(m) error rate is 1 m and there is no requirement for GPS speeds (20 to 40 km/h), and thus the applicability of the IPC
and RSU. In the simulations made, better outcomes were algorithm is assisted with a webcam at high speeds could be
achieved compared to Parker’s algorithm. This superiority questionable.
remains between 0.3 to 1.3 m by communication distance.
In comparison with other studies, it is suggested that this is 5.2.16 Parked Cars
efficient for the mobile VANET [106].
The core idea of this study is the localization of parked
5.2.15 Improving Localization in Real City Environments cars on streets or roads [108]. These vehicles are utilized
(IPC) as anchor nodes. When the engine is off, the following two
conditions are to be met: V2V communication could be
Another technique [107] aims to decline localization accu- made with the other vehicles on traffic and the OBUs can be
racy error by way of cooperative positioning in a simple easily used. In the study, why parked cars should be utilized
manner, even with only two vehicles. While doing this, it was clarified. As statistically explained, one car is parked
doesn’t need a digital city map. When the single GPS is for 6.64 h, and 69.2% of them are on the streets. Further-
used in VANET localization systems, an error of 5 to 30 more, the vehicles start to use GPS following the long-time
m occurs. It is considered to minimize the errors caused parking. It is assumed that the vehicular communication dis-
by GPS, using the past traveling records through the DR tance is 400 m. A localization simulation was designed in
system, which is not totally errorless though, that’s it may the study with 5 parked and 15 moving vehicles. Accuracy

13
F. B. Günay et al.

was obtained as 4.06 m while it was found 9.04 m using the between vehicles. Each equipped vehicle has a 48-bit MAC
traditional CL method. The study could have given better address with the IEEE 802.11p standard. Also, they have
information in regard to theoretical and localization data. 100 ms periodic time which they transmit their updated posi-
tion. Generally, this system provides a future position esti-
5.2.17 Absolute Sum of Single Differencing (ASOSD) mation with last GPS positions, distance measurements from
neighbor vehicles with vehicle speeds on time slots. Average
According to this new method, the aim of Mahmoud et al. localization errors are about 0.65 m with 100% equipped
is to achieve localization through an angle approximation vehicles and 1.4 m with 10% equipped vehicles.
method that is artificially developed when no efficient con-
nection is made by GPS satellites [109]. The blocked pseu- 5.2.20 Other Methods for GPS Based Localization
dorange is produced using the data of visible pseudoranges
transmitted via DSRC communication between the target It is possible to enhance the safety level without many costs.
vehicle and the vehicle with multi-constellation GPS. The Several vehicles may decline the performance of applications
artificial pseudoranges were established by using a geo- by deliberately transmitting incorrect positions in bad inten-
metric approach which are actually the blocked pseudor- tion. In addition to the malicious vehicles, a fake position
anges. These are artificial candidate pseudorange (ACP) could be produced due to the breakdown of the localization
and the optimum pseudorange is selected through either device and sensory system (e.g. GPS receiver). In considera-
absolute sum of double differencing (ASODD) or absolute tion of these problems, an RSU method-based localization
sum of single differencing (ASOSD). Generating pseudor- technique was proposed [112]. The number of RSUs will be
anges according to the choice of methods is considered the greater because of many intersections in urban areas, which
originality that belongs to the study. The assistant vehicles escalates the costs. In another study [113], the suggested
respectively remain 10 and 25 m far from the target vehicle method improves localization accuracy in some regions with
in the simulations. In addition, the vehicles move in a tan- the range of three RSUs. Furthermore, it decreases the nec-
dem arrangement. For localization accuracy results average essary steps for location estimation and incorporates the full
ACP and ASOSD selection based simulations were made, path into the coverage area. The purpose of another study is
and minimum accuracy errors were 4.0, 1.8, and 0.8 m, to transmit safety messages to the vehicles through the RSUs
respectively. [114]. Low-density traffic brings about network fragmenta-
tion leading to research difficulty that is critical for safety
5.2.18 Geographical İnformation Enhanced Cooperative applications. That’s why Raut et al. proposed a system to
Localizer (GIE‑CL) find out relative positions among multiple vehicles using
Big Circle Algorithm [114]. RSU can disseminate the details
A relevant method proposes to perform cooperative localiza- of the broken vehicles and the affected vehicles from the
tion with enhanced geographic data [110]. TOA method pro- multipoint alarm system to the specified ones, which will
duces the likelihood of LOS above a certain threshold value prevent the propagation issue. The study allows to establish
with respect to updating localization. The study was simu- communication with each other within the range of 100 to
lated through NanoLOC transceiver and TOA methods, and 150 m and aims to develop a mobile network in a large area.
20 measurement points were determined. Localization was A routing protocol is required to deal with the problematic
made with the only GPS, least squares (LS-CL), extended network fragmentation and transmit multipoint signal rout-
generalized approximate message passing (EGAMP-CL) and ing messages to the vehicles. Therefore, special vehicles
GIE-CL techniques, and average accuracy errors were found receive early warning messages using partial localization.
12.814, 9.5651, 10.1386, and 6.0091 m respectively. The These results from transmitting emergency alert messages
study’s essential advantages are the fixed geographic blocks onto the vehicles in the danger zone and the recorded early
and locations and their capability of being pre-installed. This warning messages onto the crashed vehicle in the related
method is preferable because it is advantageous in high area. Multicasting is preferred because all the vehicles in
mobility. the network are not influenced when any event like a crash
occurs and early warning messages are not disseminated.
5.2.19 DSRC/WAVE Thus, a state of panic is avoided, which can appear with
transmitted messages for early warning to the needless vehi-
In this study, each equipped vehicle its position by compil- cles, as well as dissemination problems and traffic jams in
ing neighbor vehicles positions with distances to the target the road scenario.
vehicle and it’s own GPS position then realizes better locali- Hashni et al. [115] suggested fusion technique depend-
zation [111]. For non-equipped vehicles GPS positions are ing on the scenario for VANET localization that GPS will
required related to the absence of DSMC communication be utilized for localization in outdoor environments. In the

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

Table 4  Non-GPS based Accuracy Network density Landmark density Kind of localization References
cooperative logalization
20–50 m Low Low Vehicular Sensors+Fuzzy Logic [124]
0.1–3 m Low Low Vehicular Sensors+TDOA [125]
24–98 m High Low OWL+Vehicular Sensors+RSSI [37]
7–28 m NA Low OWL+Vehicular Sensors [126]
NA Average Average RSSI+RSU [128]
30 m High High RSSI+WL [129]
NA Average Low RSSI+RSU [130]
10 m NA Low MCL+RSU [131]

Table 5  Cooperative localization
Protocol Fusion type Requirement Accuracy Landmark Den. References

– GPS,DR GPS 4.0–5.0 m Average [73]


NA GPS,RSSI GPS 2.0–3.0 m Average [76]
UWB GPS,UWB GPS,UWB 0.02–0.4 m Low [89]
NA GPS,wheel speed sensors GPS,Vehicular Sensors 1m Na [44]
NA GPS,INS,KF,Neural Network GPS,Vehicular Sensors 0.8 –7.2 m Average [45]
NA GPS,RSU,V2V,V2I,INS GPS,RSU,Vehicular Sensors 5–50m Low [31]
802.11p GPS,V2V,DR,MAP M. GPS,V2V NA Low [75]
ZigBee GPS,ACCELEROMETER GPS,V2V About 10 m Low [132]
802.11p RSU,DR RSU,V2V 3.75 m Low [119]
802.11p GPS,RSU GPS,RSU,V2R 0.74 m Average [112]
NA GPS,DR,V2V,INS,IEA-KF GPS,V2V,Vehicular Sensors 0.1– 3 m Average [52]
NA GPS,CMM,BF GPS,V2V 2.74 –10.47 m Average [72]
UWB GPS,UWB,EKF GPS,V2V 0.28 – 1.11 m Low [59]
DSRC GPS,V2V,V2I GPS,V2V,V2I,CAIPU 0.77 m Low [60]
802.11p V2V,INS,Fuzzy Logic V2V,Vehicular Sensors 20 m High [124]
802.11p GPS,WAVE,MAP M.,Semi EKF GPS,V2V 0.5–2.0 m High [61]
NA GPS,RSSI,Self Correction GPS,V2V 1–4 m Average [77]
802.11p GPS GPS,V2V 0.65 m High [111]
802.11p GPS,V2V,V2I GPS,V2V,V2I,KF 0.2 m Average [43]

open areas where GPS communication is made, both GPS of RSSI measurements. The second is the varying standard
directly and the fixed base station are used to communi- and average deviations based on the distances. The error
cate. By comparison of these two, better localization was rate is rising in the measurements for distances of more than
aimed. In another scenario, simulations were realized in a 120 m. RSSI measurement error rates were from 14 to 16 m
closed area of 4500 x 3400 m. Here, the Bluetooth, Wi-Fi, in the range of normal measurements.
and cellular localizations were individually implemented in In a recent paper [117] the main goals are to detect GPS
this zone. The designed system was tested in accordance position errors and deploy RSU in that region based on the
with the epidemic, ProPhet and Spray, and Wait routing greatest GPS errors. GPS integrated 316 taxicabs roamed for
protocols. The most successful protocol, Spray and Wait, 1 month in Rome. The cars downloaded the data of id, loca-
became significant in both latency and delivery. Yokoyoma tion, and time at intervals of 7 s through an Android applica-
et al. aimed in a study [116] to evaluate the distance between tion. The errors above 20 m were deleted in the system and
the vehicle and RSU by way of vehicular communication ignored. The measurements for more than 120 m distances
and recorded RSSI values and datasets. Based on this situ- contain higher error rates. RSSI measurement error rates
ation, RSSI behavior was characterized. In the study, two were 14 to 16 m in the normal range. In the study conducted
traditional analyses were made for RSSI characterization. on Vehicular Sensors [118], the RSUs are taken as reference
The first is the analysis of the quantity of beacon per unit nodes. The distinctive part of this study is the economic
length for both vehicle and RSU, including the distributions choice of both the GPS modules in the RSUs and the speed

13
F. B. Günay et al.

sensors on the vehicles, which are odometers and heading under the G5 standard by the European Telecommunications
sensors. NS3 is a simulation environment and IEEE 802.11p Standards Institute (ETSI) or Basic Safety Messages (BSM)
is used as the communication protocol. In the simulations, defined by the US Society of Automotive Engineers (SAE),
error rates were determined as 2 degrees for turning and measurements are made. In the study, it is highlighted that
as 10% for velocity. Location errors of GPS modules can the localization of an irregular pedestrian is determined with
approximate to 15 m. Consequently, the mean of localization an error rate of 1 m. With this, it is suggested that the result
errors were simulated at 0.91 m and the standard deviation will be better because the vehicular movement is simpler,
as 0.48 m. The tests performed in the Matlab and NS3 envi- and the IMU allows for better localization on vehicles. Three
ronments assumed that the vehicles go at the constant speeds based searching algorithm was used to find the best parking
of 90 and 100 km/h, that’s linear movement was anticipated. area out of the accessible ones. One contribution of the study
is the capacity to establish the WSN with a minimum-level
infrastructure. As understood in the study, localization is not
6 Non‑GPS Based Cooperative Localization an outcome but just a process towards the end. That’s why,
localization results are not discussed, and the proportional
Hardware technologies (i.e. RSU, Cellular Loc, etc.) are outcomes related to the parking lots are considered.
available to replace GPS for VANET localization when there Another relevant study [121] provides the optimal
are no systems such as GPS. Additionally, non-GPS based K-coverage to use of the minimum number of RSUs. For
methods including Weighted Loc and Monte Carlo may offer the purpose, the GDOP method was employed. In the GDOP
good accuracy for localization. However, the lack of GPS is computation, two different ways were preferred: First, in the
not just empirical. The unavailability of GPS in such areas RSS based method, spacing techniques were applied for the
as tunnels, mass buildings, and woods is not arbitrary but a distances between RSU and vehicles. GDOP can also be
localization scenario that may be usually. The studies in this computed through the Hybrid TOA/TDOA method for the
section include both cases. This type of localization without vehicles containing a uniform linear array (ULA) antenna.
GPS and so GDOP problems can be considered an advan- GDOP is computed with the obtained information of dis-
tage. Non-GPS based CL techniques are shown in Table 4 tance and angle using array process techniques. Follow-
in Sect. 7 below. ing this, optimization is made in a range with K coverage
through asynchronous particle swarm optimization (APSO),
6.1 RSU in Non‑GPS Based CL and the appropriate RSU coverage was achieved. The origi-
nality of the study is to compute RSU coverage, GDOP, and
The current non-GPS based techniques involve communi- APSO and to reduce the number of RSU to perform localiza-
cating with at least two RSUs increase system costs due tion. It was compared to uniform interval placement (UIP)
to many RSUs needed. In this study, in order to localize method and in the case of fitness when K=5, had superiority
the vehicles via communication with single RSU, partial to it 18.5% versus 10.1%. Lack of accuracy level to display
use of Dead Reckoning and a localization method using two efficiency in the localization is a deficiency. In [122] the sub-
TOAs were proposed [119]. The results showed that the rec- ject is unrelated to direct localization, in essence, explains
ommended method is 13.1 to 20.1% better than techniques the efficient multicast method developed for location man-
using one RSU, and meanwhile, full dependency on Dead agement system within VANET. There is also an additional
Reckoning appeared. When both sides used two RSUs the study on the excess of the transmitted and received data in
result obtained was 3 to 8.9% worse. Also, DR and RSU the tracking process in VANET. Its aim is to make packet
aided localization methods with the vehicle to RSU com- load reduction, using the communication techniques of RSU,
munication were studied, and a localization algorithm was WAVE (Wireless Access in Vehicular Environment – IEEE
proposed to avoid the localization spoofing attacks [19]. 802.11p) [123]. To review more studies using RSU, see [40].
In [120] a system is designed to take into consideration
the heterogeneous vehicular distribution and help to find the 6.2 Vehicular Sensors
best parking lot for an autonomous vehicle, as a result of a
common approach whose determinant parameter is the ratio As a method of integrating sensors to a vehicle provides
of the empty parking lots communicable at that time to the low-cost and continuous localization. The researchers in
total number. All the empty lots in the parking area are not [124] propose to design an intelligent localization method
available due to the coverage zone. One of the essentials of through fuzzy logic using the neighbor vehicles’ location
study is that vehicles can make localization through latera- information. The considered parameters for the localized
tion via RSU in parking garages. They use the properties vehicles are distance and heading in the fuzzy logic. In this
of inertial measurement unit (IMU) and distance measure- sense, RSSI in the received beacon signals is considered as
ments. With the Cooperative Awareness Messages (CAM) distance parameter. In addition, the heading information of

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

the neighbor vehicle is included in the transmitted beacon less. Non-use of GPS is an advantage, however accuracy
message. The neighbor vehicles are weighted based on their success cannot become at the desired levels [37]. Altoaimy
distances identified with this method. The closer the vehicle and Mahgoub discuss the data accuracy from other vehicles
is, the bigger the weight is, or vice versa. Previously, these and nodes for localization and deal with the correction of the
weights were found as the ratio of the difference between incorrect ones in their study [126] In the system weighted
the related vehicle and the closest ones distances to that localization was completed. The elements of the weight are
between the farthest and the closest ones’ distances. Accord- distance information, SINR ratio and heading information
ing to the rule base established, distances are displayed with to the neighbor vehicles.
“vvclose”, “vclose”, “close”, “med”, “far”, and “vfar”. For Moreover, the localization process is related to a beacon
heading, only two situations are available: the same or oppo- message, with which the transmitter’s id, position, time,
site. Weighted centroid localization is carried out through velocity, heading, and road information are received. The
the trilateration method by means of the obtained output errors in transmitted data cause faulty localization. To avoid
membership functions. The main difference of the study this, the system is divided into three separate processes: data
constitutes the weighted localization made using fuzzy logic. gathering, data verification, and localization. In the first
The usage of manhattan grid in the 1000 m simulation area method, data are collected and recorded from the neighbor
makes it different as well. Its effectiveness is compared to vehicles, in the second confirmed and checked for its reli-
that of centroid localization. The highest accuracy level ability and then lastly the above-mentioned localization is
reached up to nearly 20  m while it became 60 m for centroid made. Four parameters such as position, heading (turning),
localization without weight. Due to the lack of use of RSU, map (path) and timestamping accuracy values are acquired
GPS, and filter localization accuracy is very low. to establish a trust value. When 50% of the vehicles have
The aim of another paper on this method is to design wrong information and verification is undone, the average
a system to fulfill CL in the spaces where GPS connec- localization error was measured 24 m, or if done, 7 meters.
tion is unavailable [125]. The followed method can shortly The maximum error for either situation was respectively
be explained such as; Each transmitter node periodically observed 28 and 14 m.
requests for the data of position, turning, and speed from
its neighbors. After these are delivered, both the neighbor’s 6.4 TOA in Non‑GPS Based Cooperative Localization
and the transmitter’s positions are updated. At this stage, the
calculation of azimuth and speed are made using the infor- The relevant study aims to carry out 3D localization using
mation obtained from odometer and gyroscope. After that, omnidirectional antenna and eliminate collision threat
the distances of the nodes are determined through the TDOA with Doppler domain analysis [127]. While doing this, any
method. Finally, advanced location estimation is made with DSRC protocol is required. The received data of location
Kalman Filter. Three scenarios, filtered or unfiltered, are and movement of the vehicles through Basic Safety Message
simulated. They produced the results picked at the speeds of (BSM) are not needed and the only receipt of the message
3, 7, and 11 m/s, respectively. The best localization results is enough. BSM refreshes 10 times per second and trans-
are the filtered ones, and among these, the best accuracy was ceiver structures require using IEEE 802.11p. The respective
0.10 m got with the speed of 3 m/s. For speed simulations inter angles of the vehicles are identified via omnidirectional
with 7 and 11 m/s in the filtered systems, these levels were antenna and the TOA method is used for distance measure-
observed 0.2 and 0.3 m respectively. ment. In the system, Doppler domain analysis was conducted
for the vehicles to avoid a collision. In addition, collision and
6.3 Optimized Weight Localization (OWL) fake alarm probability computation were not made as well as
accurate measurement. In this respect, the study will be use-
In this study, it is aimed to perform localization using the ful to be briefly mentioned. Collision probability was found
SINR ratio calculated with exchange messages and the dis- as 60.87%, and fake alarms as 41.7%. Moreover a relevant
tance found between the neighbor vehicles. The used meth- study is available [27].
ods are Optimized Weighted Localization (OWL), Centroid
Localization, Relative Span Weighted Localization (RWL), 6.5 RSSI in Non‑GPS Based Cooperative Localization
and Weighted Localization Using Distance (WLD). Locali-
zation errors are from 24 to 98 m based on the number of The combination of VANET and Wireless Sensor Networks
vehicles lined up in VANET. The most successful localiza- (WSN) is called Vehicular Ad Hoc and Sensor Networks
tion method is OWL, followed by WLD, RWL, and centroid (VASNET). The sensor nodes with high mobility are deployed
localization ranked in order of success. Furthermore, it was onto the vehicles in VASNET. This study involves the design
observed that when the vehicles head for the same direc- of new protocols, for operating VASNET results in a num-
tion, localization estimation is more effective, or otherwise ber of challenges such as frequency assignment, standards

13
F. B. Günay et al.

for physical and communicative layers, routing, localization, the vehicles, localization is made by using the weighted least
and safety issues [128]. This localization protocol is proposed squares method. Unlike other studies using RSSI, TOA, or
vehicles in VASNET and GPS is not needed. The aim is to TDOA, the AOA data are obtained from the linear array of
exactly determine the vehicular positions via sensors placed RSUs with M elements. In the study, the effects on the out-
at road-sides, and however, its cost is very high. The protocol comes of channel contributions were aimed to deal with in a
performance was evaluated based on the variation in the Effi- realistic way. Localization errors increased by the distances
cient Road Side Sensor (RSS) to the number of nodes ratio and of the network vehicles to the RSU and approximated to
accurate localization information is transmitted by 73% when 10 m. Nevertheless, simulation results show that much bet-
the RSS ratio was 50% and by 93% at 90%. In another study ter localization can be achieved particularly than GPS as the
RSSI method is used again [37]. RSUs are deliberately got closer to vehicles, and localization
In [129] the same authors apply two weighted localization errors reduce down to less than 1 m.
(WL) methods for VANET. Another study also compares
RSSI and the other methods [130]. In this paper, localiza-
tion is made through V2V or V2R communication, IEEE 7 Non‑Cooperative Localization
802.11p, IEEE 1609 communication. DV-HOP, RSSI, and
MDS-MAP methods are applied when GPS is disconnected. In this section, the localization methods are used alone and a
The software architecture is primarily divided into three sec- fusion state does not occur. The smaller number of the used
tions. At the top layer, the operator interface exists to make technologies allows for more economic and less energy-
necessary adjustments as well as the control interface in the consuming systems. However, accuracy in these methods
middle and data interface at the bottom. System workflow cannot escalate unless landmark density is higher. Non-CL
has two stages: the first is the off-line training stage, in which techniques are presented in Table 8 below.
the OBU receives sampling information for different posi-
tions on the path and can sense the RSUs within the range. 7.1 GeoLV
In carrying out this process the periodic broadcast message
of WAVE Service Announcement (WSA) is used. In this The available information in this method are velocity, direc-
case, the surrounding RSUs are scanned, RSSI values are tion, and time ticket. They have recorded as the past data as
collected, and ultimately data of the current location trans- well as the recently identified GPS location information. The
ferred via network to a location server. The other stage is study primarily examines three localization scenarios [39];
an online stage in which all WSAs are sensed and are got
involved with the chosen network. After that, the RSUs are 1. Case of short GPS outage: The outage is so short that the
scanned, the RSSI data accumulated, and the RSSI vec- location with traditional GPS is seemingly lined with the
tor formed. Finally, the data are uploaded into the server current position.
through the RSU. DV-HOP, RSSI, and MDS-MAP methods 2. Case of longer GPS outage for Road bend position:
are compared with simulation. MDS-MAP is a successful Outage time is longer than traditional GPS location is
way when anchor number in VANET is low. In simulations, considered as being in a different line with the current
it was seen in the RSSI method that localization error tripled position.
even for a bit noise factor increase (0.1). Furthermore, the 3. Case of longer GPS outage for Straight road: Outage
MDS-MAP method produced a smaller localization error time is also longer that traditional GPS location and the
than the other two methods, in which knowing anchor nodes’ current position are considered multiple and angular.
coordinates has a little part in the localization of other nodes. This technique is used passing through perpendicular
The maximum error was 200 m in this technique while it areas.
was approximately 300 m in the others. DV-HOP is not a
recommended method in the literature to use for mobile net- Additionally, both static and dynamic relocation is made
works, rather utilized for fixed wireless networks. One of the following outage. One of the differences that the study
the disadvantages of the study is the resulting highness of makes is direction-based and geometric localization. It was
localization errors. In the study, no filter was used to posi- compared to GOT and IVCAL methods. Localization was
tively contribute to the results along with GPS. applied depending on outage times (60 to 300 s). The best
accuracy was achieved with the GeoLV technique at the
6.6 Other Methods for Non‑GPS Based Cooperative velocity interval of 20 to 80 km/h, followed by GOT and
Localization IVCAL respectively. By using the same parameters, locali-
zation delay was observed, and the same ranking was got.
The study on AOA [131] provides the angle of arrival (AOA) Nonetheless, GeoLV became much more successful at this
data by means of the beacon packages from the RSUs. On time. Both accuracy and delay simulations were conducted

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

based on the vehicular frequency (200 to 1000 vehicles), and the simulations, it was described that an estimation model
the rank remained the same. with low complexity can substantially enhance the accuracy
of neighborhood localization. The useful information to esti-
7.2 Monte Carlo mate the nearer future positions of the nodes (in a couple
of seconds) is integrated into the periodic messages with
This method suggests a distributed and range-free pursuit vehicular locations. In this way, one vehicle can localize
algorithm to trace the vehicles [40]. The word of range-free another vehicle within the coverage zone, when necessary.
means that audible anchor locations and communication For the exchange of equal-frequency messages, the average
are used together in vehicular tracking. In addition, this is error rate was reduced at a percentage of more than 50%.
a cost-efficient method that declines the expenses spent on The Vehicle Tracking Systems (VETRAC) help to moni-
GPS systems to deploy anchors citywide. The proposed tor the drivers or any third party track the position of any
algorithm was simulated, its empirical results in an intel- moving vehicle. A majority of vehicles use GPS modules
ligent city were compared to those of traditional methods, [135] as well as its use and application. In VETRAC, WiFi
and its better performance was shown. The algorithm can IEEE 802.11 b/g is used for wireless communication. The
also be combined with WiMax and 4G cellular technologies vehicular information can be seen on digital maps via the
which are used as substitutes for more energy-demanding internet or specific software. In the study, an Intelligent
GPS. Furthermore, it costs less than the GPS system for the Vehicle Navigation System has been developed to use WiFi
citywide deployment of anchors. at low traffic density and determine the optimum path for
In another relevant study, wireless network connectiv- navigation. This system can also be utilized to identify the
ity is the main requirement for VANET localization [133]. place signs for newly coming tourists. Microsoft established
There are two contributions to this paper. One of those is a partnership with different automotive manufacturers to
designing a localization method which is based on Monte develop a version of the operating system in order to use
Carlo Localization and vehicular nodes movement direction. in vehicles, and the authors picked Microsoft Windows as
Heuristics is used for realizing this process that improves platform [136]. VETRAC is a navigation system for large
accuracy and increases the number of localized vehicular campuses such as universities, airports, railway stations, etc.
nodes in a tunnel. In a connectivity based positioning sys- GPS is inactive for vehicle tracking on a campus or inside
tem, there are 2 stages. Those are VANET node deploy- a closed tunnel, and therefore WiFi was used. The Project
ment and localization stages. Tunnel shape, tunnel width and was tested with different vehicles including TATA Sumo
length, number of tubes and lanes are important factors for cars and motor-cycles in a campus where WiFi devices are
the deployment stage. Minimum VANET node number and deployed [137].
maximum coverage are the main goals of the deployment
stage. In the localization stage, there is a time period that is 7.4 ZigBee
less than the division of transmission range of Vanet node to
the maximum speed of a node. Moreover, if a node receives In the study using this system, a triangle centroid approx-
a beacon message from an anchor node, this beacon message imate location algorithm model was proposed due to the
provides its ID and location. Then VANET node stores those limitation of “distance-loss” in the wireless location model.
information as a list and forwards to other neighbor nodes in When verified in terms of distance errors and location accu-
VANET. Vehicular Node discovers the farthest and nearest racy, it was concluded that the compatibility of this method
node, so Monte Carlo location estimation can be realized is available [138]. Using the collected data in low-speed
with those nearest and furthest nodes. The minimum locali- driving conditions, location accuracy of 2m and 3m resulted.
zation error was acquired 10 m with COPO method. The position of unknown mobile node in VANET is recog-
nized through the triangle centroid method. In the study, it
7.3 IEEE 802.11 and IEEE 802. 11 b/g was evidenced that low-speed vehicles can meet the active
safety requirements.
Despite the fact that a number of vehicles are equipped with
GPS, it is impossible to accurately localize the vehicles like 7.5 RSSI in Non‑Cooperative Localization
in VANETs. However, an ignored issue is the localization
of the neighbor vehicles. In VANETs, the nodes can usu- In [139], the velocity of an object was aimed to calculate
ally move at high speed, thus locational information of a using RSSI measurement made only for the localization
neighbor vehicle may be the quick repetition of its previ- without any requirement for another sensor and the time
ous data. In the article, there is a way to solve the problem taken for synchronization. The study used three nodes in
instead of increasing the frequency of periodical exchange three different roles: mobile node, fixed node, and base sta-
messages containing a node’s position [134]. As a result of tion. Essential measurements were made between the fixed

13
F. B. Günay et al.

node and the base station. Firstly for measurement, when a classify the behavioral samples of the target. These behav-
mobile node sends a package to a fixed node, the latter pre- iors are later used to estimate the target’s next positions.
pares mobile node ID, arrival time, and the computed RSSI Based on Bayesian estimation, a Dirichlet-Multinomial
value related to this package. The mobile object’s velocity model was developed to learn the behavioral patterns of a
can approximately be calculated based on maximum RSSI moving vehicle on a dynamic metropolitan road network.
value and the period between two ticketed times as the return An empirical analysis was made by using a real-world city
time of the signal is the mobile node. The study was con- map, and the results indicated that the proposed method was
cluded that the lowest error rate in velocity calculation was successful. Another paper aims to constrain searching cover-
0.94%, and the highest rate 14.86%. In total, 120 tests were age to limit the quantity of OBUs and RSUs functioning in
conducted. The shortest distance between nodes was 0.3 m the tracking process, and thus the number of tracking mes-
and the longest 1.2 m. These experiments are performed in sages is minimized [141]. Cam-equipped OBUs function as
the lab environment, and this became a study to confirm the remote movable sensors. Tracking messages are transmitted
principles, however far away from real environment condi- between the OBUs and the RSUs. For this purpose, the OBU
tions. Other real experiment results are provided in Table 6. observations are used for an action modeling technique and
future path estimation. To obtain higher estimation accuracy,
7.6 ReCoP a new model was proposed to use a non-parametric Bayes-
ian calculation technique with Gaussian prior to the condi-
The researchers in [38] propose to make high precision tional logit system parameters. Also, this method evaluates
localization without the necessity for fixed infrastructure with an empirical analysis using a real city map. In the sug-
and GPS thanks to the Relative Cooperative Positioning gested model, the results of each simulation were weighted
algorithm. For this, one node should be assigned as head of based on datasets, and these values were compared to the
each group to fulfill one-hop communication. In the study, previously proposed D-M model [140] using the Bayesian
simulations were carried out in the NS2 environment, and method.
then different results were produced depending on a number
of different parameters. For example, when the coverage area 7.8 DR, ANN and KF
of the nodes is 50  m, the localization errors became nearly
1 m. Whereas, at the coverage area of 350 m, the errors rose The study aims to learn driver behaviors on the way through
up to 5 m. Furthermore, in this technique, the number of Artificial Neural Networks (ANN) and estimate lane chang-
Relative Maps declines as the coverage area expands. It was ing [142]. Kalman filter is used like in the other studies,
raised from 50 to 150 m of coverage area, and a decrease which can manage to recursively make the statistically opti-
in the number of Relative Maps was observed between 68 mal estimate. In addition, it may be linear optimal estimator
and 73%. in the Non-Gaussian noise. KF, DR and ANN methods were
applied, the best results were got with KF and DR, and NN
7.7 RSU in Non‑Cooperative Localization produced an undesired outcome with error rates more than
10 m. It is important to engage the ANN method in the stud-
Coming to RSU method in Non-Cooperative localization ies and simulate the city of Cologne in real conditions on
OBUs and RSUs are used to localize a mobile target or mov- the NS-2 simulator. Other simulation results are provided
ing vehicle non-cooperative with VANET vehicles, though in Table 7.
within the coverage area in this paper [140]. However, differ-
ent speed limits and varying traffic congestions on the road 7.9 Other Methods
network complicate vehicle tracking. Tracking messages
are transmitted between OBUs and RSUs in potential areas In this section, the related studies to VANET localization
where the target could be. OBU observations are utilized to that cannot be directly classified are presented. Of Cellular

Table 6  Real tests Type of motion Speed (kmph) Accuracy (m) Landmark density References

Straight, Turn 30,80,100−180 0.08– 2.96 Low [104]


Straight,Turn 20−40 6.9 −11.18 Low [107]
Straight NA 14−16 Average [116]
Straight 0.7−3.6 0.04−0.06 Low [139]
Straight,Turn NA 1.32 Low [86]
Straight,Turn NA 3.41−5.86 Low [138]

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

Table 7  Results of the simulations


Physical location Type of motion Speed (kmph) Road length (km) Accuracy (m) Landmark density References

NA Straight 50–60 5 3–16 High [55]


NA Straight 120 5 3.8–11.04 High [62]
NA Straight 0–50–100 20 NA High [143]
NA Straight 120 12 8.8 Low [19]
Zurich, Switzerland Straight,Turn 54 >10 0 - 10 High [100]
NA Straight,Turn 1–126 23.5 0.12 High [57]
Porto,Portugal Straight,Turn NA 7 1.58-9.8 High [63]
NA Straight,Turn 80 3 2.5–9.2 High [106]
NA Straight 60 5 0.05–1.89 Low [42]
Köln Straight,Turn NA 10.275 0–10 Low [142]
Manhattan,USA Straight,Turn 50.4 25 20 High [124]
Manhattan,USA Straight,Turn NA 2x2 30 Low [129]
Istanbul,Turkey Straight,Turn Max 72 NA 2.07–3.54 Average [70]
NA Straight 110 1.8 0.18 High [69]
Rome,Italy Straight,Turn NA 1.8–6 0.74–6.83 High [117]
NUIM, Ireland Straight,Turn NA NA 4.06 High [108]
Manhattan,USA Straight,Turn 40-80 1 0.5–2.0 High [61]
NA Straight,Turn 40–100 1.4 1 High [38]
NA Straight,Turn 54 5x5 1-4 Average [77]
Bologna,Italy Straight,Turn 50 1.5 0.2 Average [43]

Table 8  Non-cooperative Accuracy (m) Network density Landmark density Kind of localization References
localization
NA High High GeoLV [39]
0.1–0.7 High High MCL [40]
35 Average Average MCL [133]
0.1–9 High High IEEE 802.11p [134]
2–3 Low Low ZigBee [138]
1–5 High Average ReCoP [38]
10 Average Average Neural networks [142]
0–100 Low Low Cellular localization [146]

Localization, Lee Song [144] proposed the Automatic Vehi- or for blind spot identification (the areas out of the coverage
cle Location (AVL) method using signal transmission zone). These systems may also be used for truck/vehicle fleet
to a mobile phone present on a vehicle. In this study, the control, besides the rationale for this method first became
localization error rate was found in the range of 0 to 575 m, significant in the issue of autonomous control of today’s
based on the simulation activities in the lab environment. fleet vehicles.
This value was observed when the base stations are 5 km In [146] VANET localization or tracking is based on
far from one another and under the assumption of stand- GPS-free methods. The major contribution of the study is
ard deviation (3 dB) [145]. Using fuzzy logic, a suggestion to use RSS measurements for the current GPS in localiza-
was made to reduce this rate to an interval of 0 to 200 m. tion in VANET when GPS services are inactive. The pro-
An AVL is integrated onto particular vehicles, and signal posed localization technique samples the received signals
transmitters onto traffic signs to share the traffic and road from three base stations surrounding the mobile unit in
information through intra-object communication. Further- a very short time. Each of the RSS samples will later be
more, the road conditions of the vehicles with AVL devices matched with a distance estimate based on the shading radio
are recorded into a database to which they are connected in propagation model. The median values of the established
order to keep records for the traffic congestion. Afterward, distance clusters are used to make a final location estimate
they can be utilized to determine the shortest route to travel afterward. Short sampling time eliminates the influence of

13
F. B. Günay et al.

the movement of mobile vehicle, and the median choice the vehicles interested in them can be analyzed. According
neutralizes deviations from the time-varying radio channel to the simulation results, the SafeAnon scheme has better
characteristics. The algorithm is totally dependent on the performance than Fixed Silent Period (FSP) and Random
current cellular network signals and thus removes an extra Silent Period (RSP) schemes in terms of location privacy.
infrastructure necessity. Furthermore, the updated position is Moreover, it was concluded that the proposed SafeAnon
only known by the mobile unit and thereby location privacy scheme can improve not only location privacy but also traf-
is also provided. Therefore, a location privacy scheme based fic safety and broadcasting efficiency in VANETs [148]. The
on a safe distance, supporting DSRC, V2V, and V2I wire- techniques pertaining to the aforementioned studies are sum-
less communication technologies, called SafeAnon has been marized in Table 9.
proposed to enhance location privacy and security as well
as traffic safety. For driving safety, VANET vehicles should
periodically broadcast the updated locations, speeds, and 8 Future Work
directions [147]. Chen explains that “the safety applications
known as Vehicle Safety Communication (VSC) include the There are various study areas to further the localization
ways of Wrong-Way Driver Warning, Lane Change Warn- issue. These days, however, security and privacy, reliable
ing, Highway Merge Assistant, and so on” [148]. All of them applications, challenges of mobility specifically when the
have common needs such as periodic broadcasting and low vehicles are mobile, particularly in intra-vehicular ad hoc
delay communication. According to another study [149], networks become current issues. For sustainable intra-
exchange of safety messages in VANETs requires merely vehicular communication, all VANET vehicles should be
vehicular id identification. Nevertheless, there is no require- open to communicate with the others, which in turn causes
ment for encryption to prevent access of foreigners to the them to be exposed to system threats and cyberattacks. It
messages. In other words, security messages are transmit- can be required to investigate protocol congestion struc-
ted as plain text and signed with the sender’s special elec- tures for middle and large scale networks and if necessary
tronic key. This allows for the irrelevant others to block from include new protocols. Another important issue is that data
access to the security messages and the vehicles to automati- transmission speed should be higher in the Ad-hoc networks
cally track using the location or id-data. That’s why safety having many mobile vehicles. Either bandwidth or data
messages are broadcasted in order to improve driving safety classification processes may be required. Furthermore, the
and traffic efficiency of the vehicles in VANETs. Addition- authenticity of this data is another issue in terms of safety
ally, the position of the attackers and the information of and privacy. One of the most important problems for the

Table 9  General comparison Accuracy Network density Landmark density Type of localization References
table
Low Average Small Cellular Loc. [144]
Low Low Small Cellular Loc. [146]
Average Average Average GPS+V2V [97]
Low High Small RSU+DR [19]
High Low Small GPS+V2V [48]
High High High V2V(IEEE 802.15.4) [40]
High High High GPS+V2V+EKF [57]
High High High GPS+V2V+V2R+PF [64]
High Average Average GPS+V2V+EKF+INS [58]
High Low Low GPS+V2V(AOA)+V2I(AOA) [85]
Average Average High GPS+CMM [72]
High Low Low GPS+AOA+TOA+V2V+V2I [60]
Average Low Average GPS+RSSI [116]
High Low Low GPS+T-CP+UWB [86]
Low High High MDS-MAP, DV-HOP, V2V(RSSU) [130]
Low Low Low WL,WLD,V2V(RSSI) [129]
Low High High OWL,V2V (RSSI) [37]
Low High High GeoLV [39]
High High High ReCoP [38]
High High High Monte Carlo [133]

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

localization process is the ability to estimate the location 3. Boukerche A, Oliveira HABF, Nakamura EF, Loureiro AAF
accurately. Therefore, a general study can be conducted for (2008) Vehicular ad hoc networks: a new challenge for locali-
zation-based systems. Comput Commun 31(12):2838–2849
enhancement of localization accuracy using the right method 4. Savarese C, Rabaey JM, Beutel J (2001) Location in distrib-
in the right zone (e.g. tunnel, inner-city, dense traffic, etc.). uted ad-hoc wireless sensor networks. In: International confer-
A connection can be made between these techniques and the ence on acoustics, speech, and signal processing. Proceedings.
areas where they will be used. In addition to these problems, (ICASSP’01), volume 4, pp. 2037–2040. IEEE
5. Raut SB, Malik LG (2014) Survey on vehicle collision pre-
the broadcast/multicast/unicast and Quality of Service (QoS) diction in vanet. In: 2014 IEEE International conference on
criteria should be reviewed. These have to be re-designed so computational intelligence and computing research, pp 1–5
that the needs will be satisfied. Finally, power consumption 6. Kuutti S, Fallah S, Katsaros K, Dianati M, Mccullough F, Mou-
became another most significant problem, derived from the zakitis A (2018) A survey of the state-of-the-art localization
techniques and their potentials for autonomous vehicle applica-
vehicular transmitters and all the objects on roadsides to tions. IEEE Internet Things J 5(2):829–846
assist in localization 7. White CE, Bernstein D, Kornhauser AL (2000) Some map
matching algorithms for personal navigation assistants. Transp
Res Part C: Emerg Technol 8(1):91–108
8. Kihei B, Copeland JA, Chang Y (2014) Doppler domain
9 Conclusions localization for collision avoidance in vanets by using omni-
directional antennas. In: 2014 International conference on con-
In this study, the current localization techniques applied nected vehicles and Expo (ICCVE), pp 331–337
9. Zhu W, Gao D, Foh CH, Zhao W, Zhang H (2016) A collision
in VANET have been analyzed. VANET localization algo- avoidance mechanism for emergency message broadcast in
rithms are primarily divided into three groups: (1) GPS urban vanet. In: 2016 IEEE 83rd vehicular technology confer-
Based Cooperative, (2) Non-GPS Based Cooperative, and ence (VTC Spring), pp. 1–5
(3) Non-Cooperative. The GPS based techniques are also 10. Ansari AR, Saeed N, Ul Haq MI, Cho S (2018) Accurate 3d
localization method for public safety applications in vehicular
grouped under Filter Usage and Non-Filter Usage situations. ad-hoc networks. IEEE Access 6:20756–20763
It is observed that the used tables and the applied techniques 11. Chen R, Jin W, Regan A (2010) Broadcasting safety informa-
in the study have distinctive characteristics and can produce tion in vehicular networks: issues and approaches. IEEE Netw
no perfect solution. It is shown that methods using filter 24(1):20–25
12. Deshmuk M, Dinesh D (2014) Challenges in vehicle ad hoc
have higher landmark density, in particular, they are suc- network (vanet). Int J Eng Technol Manag Appl Sci 2(7):76–88
cessful in terms of accuracy and can manage to get submeter 13. Parkinson BW, Enge P, Axelrad P, Spilker Jr JJ (1996) Global
localization error. Cooperative Localizations made with GPS positioning system: theory and applications, Volume II. Ameri-
generally have better results than those without GPS. Moreo- can Institute of Aeronautics and Astronautics
14. Misra P, Enge P (2006) Global positioning system: signals,
ver, the RSU structures do not achieve to enhance localiza- measurements and performance, 2nd edn. Ganga-Jamuna
tion accuracy in the study, in fact, they decrease the number Press, Massachusetts
of mobile anchor nodes that are required in VANETs. 15. Alkan RM, Karaman H, Sahin M (2005) Gps, galileo and
For the obtained results, VANET operation was tested for glonass satellite navigation systems amp;amp; gps moderni-
zation. In: Proceedings of 2nd international conference on
both realistic settlements and the speeds approximating to recent advances in space technologies, 2005. RAST 2005., pp
the real world. However, good and bad results were achieved 390–394
for accuracy. More consistent accuracy results were deter- 16. Li X, Ge M, Dai X, Ren X, Fritsche M, Wickert J, Schuh H
mined in the smaller number of tests. In the experiments, (2015) Accuracy and reliability of multi-gnss real-time pre-
cise positioning: gps, glonass, beidou, and galileo. J Geodesy
the zones having no GPS and cellular communication or 89(6):607–635
tunnel situations were also considered, and relatively sat- 17. Pereira FC, Costa H, Pereira NM (2009) An off-line map-match-
isfactory outcomes were achieved. Along with traditional ing algorithm for incomplete map databases. Eur Transp Res Rev
methods, such as TOA and TDOA, more innovative tech- 1(3):107–124
18. Nascimento PPLD, Kimura BYL, Guidoni DL, Villas LA (2018)
niques including GeoLV, ReCop, and OWL were used for An integrated dead reckoning with cooperative positioning solu-
VANET localization. tion to assist gps nlos using vehicular communications. Sensors
18(9):2895
19. Sun L, Wu Y, Xu J, Xu Y (2012) An rsu-assisted localization
method in non-gps highway traffic with dead reckoning and v2r
References communications. In: 2nd International conference on consumer
electronics, communications and networks (CECNet), 2012 , pp
1. Amundson I, Koutsoukos XD (2009) A survey on localization 149–152. IEEE
for mobile wireless sensor networks. In: Fuller R, Koutsoukos 20. Andrews MR, Mitra PP et al (2001) Tripling the capacity of wire-
XD (eds) Mobile entity localization and tracking in GPS-less less communications using electromagnetic polarization. Nature
environnments. Springer, Berlin, pp 235–254 409(6818):316
2. Al-Sultan S, Al-Doori MM, Al-Bayatti AH, Zedan H (2014) A 21. Parker R, Valaee S (2007) Vehicular node localization using
comprehensive survey on vehicular ad hoc network. J Netw Com- received-signal-strength indicator. IEEE Trans Veh Technol
put Appl 37:380–392 56(6):3371–3380

13
F. B. Günay et al.

22. Kul G, Özyer T, Tavli B (2014) IEEE 802.11 wlan based real time localization using previous path detection. In: Advance com-
indoor positioning: literature survey and experimental investiga- puting conference (IACC), 2015 IEEE International, pp 82–87.
tions. Procedia Comput Sci 34:157–164 The 9th International IEEE
Conference on Future Networks and Communications (FNC’14) 43. Soatti G, Nicoli M, Garcia N, Denis B, Raulefs R, Wymeersch
23. Duymaz E, Oguz AE, Temeltas H (2017) Eş zamanlı konum H (2017) Implicit cooperative positioning in vehicular net-
belirleme ve haritalama probleminde yeni bir durum tah- works. arXiv preprint arXiv​:1709.01282​
min yöntemi olarak parçacık akış filtresi. Gazi Üniversitesi 44. Rezaei S, Sengupta R (2007) Kalman filter-based integration of
Mühendislik-Mimarlık Fakültesi Dergisi, 32(4) dgps and vehicle sensors for localization. IEEE Trans Control
24. Akhlaghi S, Zhou N, Huang Z (2017) Adaptive adjustment of Syst Technol 15(6):1080–1088
noise covariance in kalman filter for dynamic state estimation. 45. Drawil N, Basir O (2008) Vehicular collaborative technique for
CoRR, abs/1702.00884 location estimate correction. In: Vehicular technology confer-
25. Čapkun S, Hamdi M, Hubaux J-P (2002) Gps-free positioning in ence, 2008. VTC 2008-Fall. IEEE 68th, pp 1–5. IEEE
mobile ad hoc networks. Clust Comput 5(2):157–167 46. Drawil N, Basir O (2009) Toward increasing the localization
26. Venkatraman S, Caffery J, You HR (2002) Location using los accuracy of vehicles in vanet. In: International conference on
range estimation in nlos environments. In: 55th vehicular tech- vehicular electronics and safety (ICVES), pp 13–18. IEEE
nology conference. VTC Spring, volume 2, pp 856–860. IEEE 47. Mo Y, Dexin Y, Song J, Zheng K, Guo Y (2016) Vehicle
27. Le BL, Ahmed K, Tsuji H (2003) Mobile location estimator with position updating strategy based on kalman filter predic-
nlos mitigation using kalman filtering. In: Wireless communica- tion in vanet environment. Discret Dyn Nat Soc. https​://doi.
tions and networking. WCNC, volume 3, pp 1969–1973. IEEE org/10.1155/2016/14043​96
28. Wylie MP, Holtzman J (1996) The non-line of sight problem in 48. Rohani M, Gingras D, Vigneron V, Gruyer D (2015) A new
mobile location estimation. In: 5th IEEE international conference decentralized Bayesian approach for cooperative vehicle locali-
on universal personal communications. Record., volume 2, pp zation based on fusion of gps and vanet based inter-vehicle dis-
827–831 tance measurement. IEEE Intell Transp Syst Mag 7(2):85–95
29. Caffery JJ, Stuber GL (1998) Overview of radiolocation in cdma 49. Howard A, Mataric MJ, Sukhatme GS (2003) Putting
cellular systems. IEEE Commun Mag 36(4):38–45 the’i’in’team’: an ego-centric approach to cooperative locali-
30. Fox V, Hightower J, Liao L, Schulz D, Borriello G (2003) Bayes- zation. In: IEEE International conference on robotics and auto-
ian filtering for location estimation. IEEE Pervasive Comput mation. Proceedings. ICRA’03., volume 1, pp 868–874
2(3):24–33 50. Fox D, Burgard W, Thrun S (1999) Markov localization for
31. Golestan K, Seifzadeh S, Kamel M, Karray F, Sattar F (2012) mobile robots in dynamic environments. J Artif Intell Res
Vehicle localization in vanets using data fusion and v2v com- 11:391–427
munication. In: Proceedings of the second ACM international 51. Howard A, Matark MJ, Sukhatme GS (2002) Localization for
symposium on Design and analysis of intelligent vehicular net- mobile robot teams using maximum likelihood estimation. In:
works and applications, pp 123–130. ACM IEEE/RSJ International conference on intelligent robots and
32. Gentile C, Alsindi N, Raulefs R, Teolis C (2013) Cellular locali- systems., volume 1, pp 434–439
zation, pp 137–159. 11 52. Ghaleb FA, Zainal A, Rassam MA, Abraham A (2017)
33. Malik Y, Khaliq KA, Abdulrazak B, Tariq MU (2011) Mobile Improved vehicle positioning algorithm using enhanced inno-
node localization in cellular networks. CoRR, abs/1201.2102 vation-based adaptive kalman filter. Pervasive Mob Comput
34. Chausse F, Laneurit J, Chapuis R (2005) Vehicle localization 40:139–155
on a digital map using particles filtering. In: IEEE Proceedings. 53. Kulkarni PS, Labade RP (2017) Vehicle positioning using
Intelligent vehicles symposium, pp 243–248 kalman filter for dedicated short range communication. Int J Eng
35. Fogue M, Sanguesa JA, Martinez FJ, Marquez-Barja JM (2018) Technol Sci Res 4(5):608–613
Improving roadside unit deployment in vehicular networks by 54. Najah Abu Ali and Mervat Abu-Elkheir (2015) Improving locali-
exploiting genetic algorithms. Appl Sci 8(1):86 zation accuracy: successive measurements error modeling. Sen-
36. Nasr A, Mohamed SAE (2012) Accurate distance estimation for sors 15(7):15540–15561
vanet using nanointegrated devices. Opt Photonics J 2(02):113 55. Nabil Mohamed Drawil and Otman Basir (2010) Intervehicle-
37. Altoaimy Lina, Mahgoub Imad (2014) Owl: optimized weighted communication-assisted localization. IEEE Trans Intell Transp
localization for vehicular ad hoc networks. In: 2014 International Syst 11(3):678–691
conference on connected vehicles and expo (ICCVE), pp 699– 56. Jaiswal RK, Jaidhar CD (2017) Location prediction algorithm for
704. IEEE a nonlinear vehicular movement in vanet using extended kalman
38. Alotaibi MM, Boukerche A, Mouftah H (2014) Distributed rela- filter. Wireless Netw 23:2021–2036
tive cooperative positioning in vehicular ad-hoc networks. In: 57. Jaiswal RK, Jaidhar CD (2017) Location prediction algorithm for
Global information infrastructure and networking symposium a nonlinear vehicular movement in vanet using extended kalman
(GIIS), 2014, pp 1–8. IEEE filter. Wireless Netw 23(7):2021–2036
39. Kaiwartya O, Cao Y, Lloret J, Kumar S, Aslam N, Kharel R, 58. Elazab M, Noureldin A, Hassanein HS (2017) Integrated coop-
Abdullah AH, Shah RR (2018) Geometry-based localization for erative localization for vehicular networks with partial gps access
gps outage in vehicular cyber physical systems. IEEE Trans Veh in urban canyons. Veh Commun 9:242–253
Technol 67:3800–3812 59. Hoang GM, Denis B, Härri J, Slock DTM (2017) Mitigating
40. Kumar S, Kislay K, Singh MK, Hegde RM (2014) A range-free unbalanced gdop effects in range-based vehicular cooperative
tracking algorithm in vehicular ad-hoc networks. In: 2014 Twen- localization. In: International conference on communications
tieth national conference on communications (NCC), , pp 1–6. workshops (ICC Workshops), pp 659–664. IEEE
IEEE 60. Golestan K, Sattar F, Karray F, Kamel M, Seifzadeh S (2015)
41. Patwari N, Ash JN, Kyperountas S, Hero AO, Moses RL, Correal Localization in vehicular ad hoc networks using data fusion and
NS (2005) Locating the nodes: cooperative localization in wire- v2v communication. Comput Commun 71:61–72
less sensor networks. IEEE Signal Process Mag 22(4):54–69 61. Mohammadabadi PH, Valaee S (2014) Cooperative node posi-
42. Suryawanshi S, Gupta D, Gupta S, Jain S (2015) On the hybrid tioning in vehicular networks using inter-node distance measure-
augmentation of inter-vehicular communication assisted ments. In: 2014 IEEE 25th annual international symposium on

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

personal, indoor, and mobile radio communication (PIMRC), pp 81. Bauer S, Obst M, Streiter R, Wanielik G (2013) Evaluation of
1448–1452. IEEE shadow maps for non-line-of-sight detection in urban gnss vehi-
62. Shao Z, Li W, Wu Y, Shen L (2010) Multi-layer and multi- cle localization with vanets-the gain approach. In: IEEE 77th
dimensional information based cooperative vehicle localization vehicular technology conference (VTC Spring), pp 1–5
in highway scenarios. In: 2010 12th IEEE international confer- 82. Smaili C, El Najjar ME, Charpillet F (2008) A road matching
ence on communication technology (ICCT), pp 567–571. IEEE method for precise vehicle localization using hybrid Bayesian
63. Cruz SB, Abrudan TE, Xiao Z, Trigoni N, Barros J (2017) Neigh- network. J Intell Transp Syst 12(4):176–188
bor-aided localization in vehicular networks. IEEE Trans Intell 83. Taylor G, Blewitt G (2000) Road reduction filtering using gps.
Transp Syst 18(10):2693–2702 In: 3th AGILE conference on geographic information science,
64. Hoang GM, Denis B, Härri J, Slock DT (2017) Robust data pp 114–120. Citeseer
fusion for cooperative vehicular localization in tunnels. In: 84. Greenfeld JS (2002) Matching gps observations to locations
Intelligent vehicles symposium (IV), 2017 IEEE, pp 1372–1377. on a digital map. In: 81th annual meeting of the transportation
IEEE research board, volume 1, pp 164–173
65. Ali Ufuk Peker and Tankut Acarman (2017) Vanet-assisted coop- 85. Fascista A, Ciccarese G, Coluccia A, Ricci G (2017) Angle of
erative vehicle mutual positioning: feasibility study. IEICE Trans arrival-based cooperative positioning for smart vehicles. IEEE
Fundam Electron Commun Comput Sci 100(2):448–456 Trans Intell Transp Syst 19(9):2880–2892
66. Jensfelt P (2001) Approaches for mobile robot localization in 86. Shen F, Cheong JW, Dempster AG (2015) An ultra-wide band-
indoor environments. Ph. D. thesis, Royal Institute of Technol- width-based range/gps tight integration approach for relative
ogy, Stockholm, Sweden positioning in vehicular ad hoc networks. Meas Sci Technol
67. Rekleitis I (2003) Cooperative localization and multi-robot 26(4):045003
exploration. Ph. D. thesis, School of Computer Science, McGill 87. Kalman RE (1960) A new approach to linear filtering and predic-
University, Montreal, Quebec, Canada tion problems. J Basic Eng 82(1):35–45
68. Chausse F, Laneurit J, Chapuis R (2005) Vehicle localization 88. Martínez JL, Molina-Mesa R, Mandow A, Rodríguez-Serrano
on a digital map using particles filtering. In: Intelligent vehicles CA (2002) Continuous localization via wide-area dgps for out-
symposium, 2005. Proceedings. IEEE, pp 243–248. IEEE door navigation of mobile robots. Integr Comput-Aided Eng
69. Hoang GM, Denis B, Härri J , Slock DTM (2016) Cooperative 9:1–13
localization in gnss-aided vanets with accurate ir-uwb range 89. Fernandez-Madrigal JA, Cruz-Martin E, Gonzalez J, Galindo C,
measurements. In: Positioning, navigation and communications Blanco JL (2007) Application of uwb and gps technologies for
(WPNC), 2016 13th Workshop on, pp 1–6. IEEE vehicle localization in combined indoor-outdoor environments.
70. Peker AU, Acarman T, Yaman C, Yuksel E (2014) Vehicle locali- In: ISSPA, pp 1–4
zation enhancement with vanets. In: Intelligent vehicles sympo- 90. Zhang P, Zhang Z, Boukerche A (2012) Cooperative location
sium proceedings, 2014 IEEE, pp 661–666. IEEE verification for vehicular ad-hoc networks. In: IEEE International
71. Goli SA, Far BH, Fapojuwo AO (2015) Cooperative multi-sensor conference on communications (ICC), pp 37–41
multi-vehicle localization in vehicular adhoc networks. In: 2015 91. Agarwal Y, Jain K, Kumar S, Bhardwaj GN (2016) Tlst: time
IEEE International conference on information reuse and integra- of arrival based localization and smart tunnel concept in vanets.
tion (IRI), pp 142–149. IEEE In: 2016 3rd international conference on signal processing and
72. Rohani M, Gingras D, Gruyer D (2016) A novel approach for integrated networks (spin), pp 763–768. IEEE
improved vehicular positioning using cooperative map matching 92. Choudhary D, Yadav D, Joshi J, Deka M, Jha S (2016) A model
and dynamic base station dgps concept. IEEE Trans Intell Transp based connectivity and localization strategy for vehicular ad hoc
Syst 17(1):230–239 networks. In: 7th International conference on intelligent systems,
73. King T, Füßler H, Transier M, Effelsberg W (2006) Dead-reckon- modelling and simulation (ISMS), 2016 , pp 331–336. IEEE
ing for position-based forwarding on highways. In: International 93. Gunay FB, Cavdar T (2014) Mobile fleet localization model via
workshop on intelligent transportation (WIT), pp 199–204 rssi, toa and tdoa in wireless sensor networks. In: Signal process-
74. Füßler H, Mauve M, Hartenstein H, Käsemann M, Vollmer D ing and communications applications conference (SIU), 2014
(2002) A comparison of routing strategies for vehicular ad hoc 22nd, pp 1431–1434. IEEE
networks. Technical reports, 2 94. Altoaimy L, Mahgoub I, Rathod M (2014) Weighted localization
75. Abumansoor O, Boukerche A (2012) A secure cooperative in vehicular ad hoc networks using vehicle-to-vehicle communi-
approach for nonline-of-sight location verification in vanet. IEEE cation. In: 2014 Global information infrastructure and network-
Trans Veh Technol 61(1):275–285 ing symposium (GIIS), pp 1–5
76. Parker R, Valaee S (2006) Vehicle localization in vehicular net- 95. Altoaimy L (2016) Reliable vehicle-to-vehicle weighted localiza-
works. In: Vehicular technology conference, 2006. VTC-2006 tion in vehicular networks. Florida Atlantic University
Fall. 2006 IEEE 64th, pp 1–5. IEEE 96. Liu K, Lim HB, Frazzoli E, Ji H, Lee VC (2014) Improving
77. Salah Abdel Mageid (2016) Self-correcting localization scheme positioning accuracy using gps pseudorange measurements for
for vehicle to vehicle communication. Int J Comput Netw Appl cooperative vehicular localization. IEEE Trans Veh Technol
(IJCNA) 3(5):95–107 63(6):2544–2556
78. Barani H, Fathy M et al (2010) An algorithm for localization in 97. Yan T, Zhang W, Wang G, Zhang Y (2011) Got: grid-based on-
vehicular ad-hoc networks. J Comput Sci 6(2):168 road localization through inter-vehicle collaboration. In: 2011
79. Xiao B, Yu B, Gao C (2006) Detection and localization of Eighth IEEE international conference on mobile ad-hoc and sen-
sybil nodes in vanets. In: Proceedings of the 2006 workshop sor systems, pp 13–18. IEEE
on Dependability issues in wireless ad hoc networks and sensor 98. Savvides A, Park H, Srivastava MB (2002) The bits and flops of
networks, pp 1–8. ACM the n-hop multilateration primitive for node localization prob-
80. Li H, Chen X, Huang L, Yao D (2012) A gps/wi-fi integrated lems. In: Proceedings of the 1st ACM international workshop on
system for positioning in cooperative vehicle and infrastructure Wireless sensor networks and applications, pp 112–121. ACM
system. In: IEEE international conference on vehicular electron- 99. Benslimane A (2005) Localization in vehicular ad hoc networks.
ics and safety (ICVES), pp 285–289 In: In Proc. IEEE systems communications, pp 19–25

13
F. B. Günay et al.

100. Yan T, Zhang W, Wang G (2014) A grid-based on-road locali- improve positioning for vehicular networks. In: Vehicular tech-
zation system in vanet with linear error propagation. IEEE nology conference (VTC-Fall), 2016 IEEE 84th, pp 1–5. IEEE
Trans Wirel Commun 13(2):861–875 118. LeBlanc HJ, Hassan F, Gomez E, Alsbou N (2016) Poster:
101. Lee E-K, Oh SY, Gerla M (2012) Rfid assisted vehicle posi- a lightweight localization algorithm using v2v and v2i short
tioning in vanets. Pervasive Mob Comput 8(2):167–179 range communication in highly dense reference networks. In:
102. Alam N, Balaei AT, Dempster AG (2013) Relative position- Vehicular networking conference (VNC), 2016 IEEE, pp 1–2.
ing enhancement in vanets: a tight integration approach. IEEE IEEE
Trans Intell Transp Syst 14(1):47–55 119. Wahab AA, Khattab A, Fahmy YA (2013) Two-way toa with
103. Hao Y, Shen F (2015) A low-cost imu/gnss cooperative posi- limited dead reckoning for gps-free vehicle localization using
tioning method for vanets in the urban environments. Int J single rsu. In: 2013 13th International conference on ITS tel-
Smart Home 9(12):255–266 ecommunications (ITST), pp 244–249. Citeseer
104. Speth T, Riebl R, Brandmeier T, Facchi C, Al-Bayatti AH, 120. Correa A, Boquet G, Morell A, Lopez Vicario J (2017) Auton-
Jumar U (2016) Enhanced inter-vehicular relative position- omous car parking system through a cooperative vehicular
ing. In: 2016 IEEE 19th international conference on intelligent positioning network. Sensors 17(4):848
transportation systems (ITSC), pp 867–872. IEEE 121. Zhang R, Yan F, Xia W, Xing S, Wu Y, Shen L (2017) An opti-
105. Garip MT, Kim PH, Reiher P, Gerla M (2017) Interloc: an mal roadside unit placement method for vanet localization. In:
interference-aware rssi-based localization and sybil attack IEEE Global communications conference GLOBECOM 2017-
detection mechanism for vehicular ad hoc networks. In: 2017 2017 , pp 1–6. IEEE
14th IEEE Annual Consumer communications & networking 122. Al-Ezaly E, Abu-Elkeir M, Riad A (2017) Collaborative
conference (CCNC), pp 1–6. IEEE vehicle location management service for enhanced hybrid
106. Peng X, Shen Q, Li W, Ouyang J (2016) Positioning algorithm reactive and proactive multicast in vanets. Arab J Sci Eng
for vehicular ad hoc networks. In: 2016 12th International con- 42(2):691–704
ference on natural computation, fuzzy systems and knowledge 123. Derder A, Moussaoui S, Boualouache A (2015) Target tracking
discovery (ICNC-FSKD), pp 2283–2287. IEEE in vanets using v2v communication with packet load enhance-
107. Tsai M-F, Wang P-C, Shieh C-K, Hwang W-S, Chilam- ment. In: First international conference on new technologies of
kurti N, Rho S, Lee YS (2015) Improving positioning accu- information and communication (NTIC), 2015 , pp 1–6. IEEE
racy for vanet in real city environments. J Supercomput 124. Altoaimy L, Mahgoub I (2014) Fuzzy logic based localization
71(6):1975–1995 for vehicular ad hoc networks. In: 2014 IEEE symposium on
108. Ordóñez-Hurtado RH, Shorten RN (2014) Using stationary vehi- computational intelligence in vehicles and transportation systems
cles to enhance cooperative positioning in vehicular ad-hoc net- (CIVTS), pp 121–128. IEEE
works. In: 2014 International conference on connected vehicles 125. Elazab M (2015) Integrated cooperative localization in vanets
and expo (ICCVE), pp 867–868. IEEE for gps denied environments. Ph. D. thesis
109. Mahmoud A, Noureldin A, Hassanein HS (2016) Distributed 126. Altoaimy L, Mahgoub I (2016) Mobility data verification for
vehicle selection for non-range based cooperative positioning in vehicle localization in vehicular ad hoc networks. In: Wireless
urban environments. In: 2016 IEEE international conference on communications and networking conference (WCNC), 2016
communications (ICC), pp 1–6. IEEE IEEE, pp 1–6. IEEE
110. Luo F, Wang S, Gong Y, Jing X, Zhang L (2018) Geographical 127. Kihei B, Copeland JA, Chang Y (2014) Doppler domain localiza-
information enhanced cooperative localization in vehicular ad- tion for collision avoidance in vanets by using omnidirectional
hoc networks. IEEE Signal Process Lett 25(4):556–560 antennas. In: 2014 International conference on connected vehi-
111. Fujii S, Fujita A, Umedu T, Kaneda S, Yamaguchi H, Higashino cles and expo (ICCVE), pp 331–337. IEEE
T, Takai M (2011) Cooperative vehicle positioning via v2v com- 128. Piran MJ, Murthy GR, Babu GP, Ahvar E (2011) Total gps-free
munications and onboard sensors. In: Vehicular technology con- localization protocol for vehicular ad hoc and sensor networks
ference (VTC Fall), 2011 IEEE, pp 1–5. IEEE (vasnet). In: Third international conference on computational
112. Chia-Ho O (2014) A roadside unit-based localization scheme for intelligence, modelling and simulation (CIMSiM), pp 388–393.
vehicular ad hoc networks. Int J Commun Syst 27(1):135–150 IEEE
113. Alimohammadi M, Pouyan AA (2014) Vehicular ad hoc net- 129. Altoaimy L, Mahgoub I, Rathod M (2014) Weighted localization
works: Introduction and a proposal for vehicle positioning. in vehicular ad hoc networks using vehicle-to-vehicle commu-
In: 13th International conference on traffic and transportation nication. In: Global information infrastructure and networking
engineering symposium (GIIS), 2014, pp 1–5. IEEE
114. Raut YC, Katkar VN, Sarode SM (2014) Early alert system using 130. Xiong W, Hu X, Wang B, Fang J (2014) Vehicle node localiza-
relative positioning in vehicular ad-hoc network. In: Annual tion without gps in vanet. In: 2014 International conference on
international conference on emerging research areas: magnet- connected vehicles and expo (ICCVE), pp 1068–1073. IEEE
ics, machines and drives (AICERA/iCMMD), pp 1–8. IEEE 131. Fascista A, Ciccarese G, Coluccia A, Ricci G (2017) A localiza-
115. Hashmi MT, Adnan A, Hadi F, Zubair M (2017) Localized tion algorithm based on v2i communications and aoa estimation.
data fusion model for vanets using gps and non-gps system. In: IEEE Signal Process Lett 24(1):126–130
2017 International conference on communication technologies 132. Fang M, Li L, Huang W (2010) Research of hybrid positioning
(ComTech), pp 180–185. IEEE based vehicle interactive navigation system. In: 2010 Interna-
116. Yokoyama RS, Kimura BYL, Villas LA, Moreira EDS (2015) tional conference on multimedia information networking and
Measuring distances with rssi from vehicular short-range com- security (MINES), pp 974–978. IEEE
munications. In: 2015 IEEE International conference on com- 133. Salah Abdel Mageid (2017) Connectivity based positioning sys-
puter and information technology; ubiquitous computing and tem for underground vehicular ad hoc networks. Int J Comput
communications; dependable, autonomic and secure comput- Netw Appl (IJCNA) 4(1):1–14
ing; pervasive intelligence and computing (CIT/IUCC/DASC/ 134. Boukerche A, Rezende C, Pazzi RW (2009) Improving neighbor
PICOM), 100–107. IEEE localization in vehicular ad hoc networks to avoid overhead from
117. Santos FA, Akabane AT, Yokoyama RS, Loureiro AAF, Vil- periodic messages. In: Global telecommunications conference,
las LA (2016) A roadside unit-based localization scheme to 2009. GLOBECOM 2009. IEEE, pp 1–6. IEEE

13
Vehicular Ad Hoc Network (VANET) Localization Techniques: A Survey

135. Montgomery J (2005) A real-time traffic and weather reporting for vehicular ad hoc networks. In: 2015 IEEE symposium on
system for motorists. In: Second IEEE consumer communica- computers and communication (ISCC), pp 359–364. IEEE
tions and networking conference. CCNC., pp 580–581 143. Abumansoor O, Boukerche A (2011) A cooperative multi-hop
136. Goel S, Imielinski T, Ozbay K (2004) Ascertaining viability of location verification for non line of sight (nlos) in vanet. In:
wifi based vehicle-to-vehicle network for traffic information dis- Wireless communications and networking conference (WCNC),
semination. In: The 7th international IEEE conference on intel- 2011 IEEE, pp 773–778. IEEE
ligent transportation systems. Proceedings., pp 1086–1091 144. Song H-L (1994) Automatic vehicle location in cellular com-
137. Thangavelu A, Bhuvaneswari K, Kumar K, SenthilKumar K, munications systems. IEEE Trans Veh Technol 43(4):902–908
Sivanandam SN (2007) Location identification and vehicle track- 145. Lee WC (1982) Mobile communications engineering. McGraw-
ing using vanet (vetrac). In: 2007 international conference on Hill Professional, New York
signal processing, communications and networking 146. Ashok VG, Gongjun Y, Olariu S, Gupta A (2010) Privacy aware
138. Jiangfeng W, Feng G, Fei Y, Shaoxuan S (2009) Design of wire- localization in vanet. In: 2010 IEEE 7th international conference
less positioning algorithm of intelligent vehicle based on vanet. on mobile adhoc and sensor systems (MASS), pp 739–744. IEEE
In: Intelligent vehicles symposium, 2009 IEEE, pp 1098–1102. 147. CAMP Vehicle Safety Communications Consortium et al (2005)
IEEE Vehicle safety communications project: task 3 final report:
139. Alfandi O, Bochem A, Gurabi MA, Díaz AR, Mehedi MI, identify intelligent vehicle safety applications enabled by dsrc.
Hogrefe D (2016) Calculating the speed of vehicles using wire- National Highway Traffic Safety Administration, US Department
less sensor networks. In: 2016 federated conference on computer of Transportation, Washington DC
science and information systems (FedCSIS), pp 1043–1047. 148. Chen Y-M, Wei Y-C (2012) Safeanon: a safe location privacy
IEEE scheme for vehicular networks. Telecommun Syst 50(4):339–354
140. Reza TA, Barbeau M, Alsubaihi B (2013) Tracking an on the run 149. Raya M, Hubaux J-P (2007) Securing vehicular ad hoc networks.
vehicle in a metropolitan vanet. In: Intelligent vehicles sympo- J Comput Secur 15(1):39–68
sium (IV), 2013 IEEE, pp 220–227. IEEE
141. Reza TA, Barbeau M, Lamothe G, Alsubaihi B (2013) Non- Publisher’s Note Springer Nature remains neutral with regard to
cooperating vehicle tracking in vanets using the conditional jurisdictional claims in published maps and institutional affiliations.
logit model. In: 16th International IEEE conference on intelligent
transportation systems-(ITSC), pp 626–633
142. Balico LN, Oliveira HABF, Souza EL, Pazzi RW, Nakamura EF
(2015) On the performance of localization prediction methods

13

You might also like