© istockphoto.

com/zonadearte

Vehicular Communications
Survey and Challenges of
Channel and Propagation Models

Wantanee Viriyasitavat, Mate Boban,

Hsin-Mu Tsai, and Athanasios V. Vasilakos

V
ehicular communication is characterized by a dynamic environment,
high mobility, and comparatively low antenna heights on the communi-
cating entities (vehicles and roadside units). These characteristics make
vehicular propagation and channel modeling particularly challenging. In
this article, we classify and describe the most relevant vehicular propagation and
channel models, with a particular focus on the usability of the models for the
evaluation of protocols and applications. We first classify the models based on
the propagation mechanisms they employ and their implementation approach.
We also classify the models based on the channel properties they implement and
pay special attention to the usability of the models, including the complexity of
implementation, scalability, and the input requirements (e.g., geographical data
input). We also discuss the less-explored aspects in vehicular channel modeling,
including modeling specific environments (e.g., tunnels, overpasses, and parking

Digital Object Identifier 10.1109/MVT.2015.2410341

Date of publication: 19 May 2015

june 2015 | IEEE vehicular technology magazine 1556-6072/15©2015ieee ||| 55

lots) and types of communicating vehicles (e.g., scooters instance, differences in the relative height of the transmit-
and public transportation vehicles). We conclude by iden- ter and receiver antennas could lead to significantly differ-
tifying the underresearched aspects of vehicular propaga- ent signal propagation behavior. The operating frequency
tion and channel modeling that require further modeling and communication distance in vehicular communications
and measurement studies. also differ from those in cellular systems. Vehicular com-
munication systems are envisioned to operate at 5.9 GHz
Introduction and over short distances (10–500 m), whereas currently de-
The most important characteristics that separate vehicular ployed cellular systems operate at 700–2,100 MHz over long
communications, and therefore the vehicular channel mod- distances (up to tens of kilometers) [1]. Several surveys on
eling, from other types of wireless communications are V2V channel models exist. For example, Molisch et al. [1]
■■ the diverse environments where the communication describe key issues in V2V channels and summarize the
happens V2V channel measurement studies in various scenarios.
■■ the combinations of different communication types, We classify V2V channels based on their implementation
e.g., vehicle-to-vehicle (V2V), vehicle-to-infrastructure approach and discuss the advantages and disadvantages
(V2I), and vehicle-to-pedestrian (V2P) communication of each approach. Mecklenbrauker et al. [2] review both
■■ the objects, both static and mobile, that affect the the V2V and V2I propagation channels, focusing on the
vehicular communication. impact of different vehicular channel characteristics on
Together, these characteristics result in complex prop- the design of a vehicular wireless system. Wang et al. [3]
agation environments that are a challenge to model. survey V2V channel measurement and models, including
Figure 1 shows how the small- and large-scale signal the model classification based on the implementation ap-
statistics vary rapidly in a typical urban setting due to proach and guidelines for setting up a V2V measurement
the dynamic environment, low antenna heights, and high and developing realistic V2V channel models.
mobility of the vehicles. Looking into the propagation We survey the state of the art in vehicular channel mod-
characteristics, Figure 2 shows that the built-up nature eling with a particular focus on 1) the usability of the mod-
of the environment causes the signal traversing from els for simulation at different scales (e.g., link level versus
the transmitter to the receiver to interact with a large system level) and considering the amount of geographic
number of surrounding objects. Even for single-bounce information available, 2) the specific issues that need to be
(e.g., first-order) reflections and diffractions in urban en- considered for the actual deployment of vehicular commu-
vironments, the number of resulting rays at the receiver nication systems, and 3) providing guidelines for choosing
is large. The high density of objects, combined with the a suitable channel model. Because of the rollout of vehicu-
high mobility of the communicating vehicles and their lar communication systems that is planned for the com-
surroundings, shows that capturing the characteristics ing years in the European Union, the United States, Japan,
of vehicular channels is far from trivial. and other countries, large-scale evaluation and fine-tuning
While a number of existing mobile channel models have
been extensively used for cellular systems, they are often
not well suited for vehicular systems, due to the unique
features of vehicular channels mentioned previously. For

-50
Received Power [dBm]

-60

-70

-80

-90

500 600 700 800 Figure 2  A simulation of propagation mechanisms in an urban

Time (s) area. Reflections and diffractions are shown for randomly selected
communication pairs. The objects in the scene are represented by
different colors: the buildings are shown with black lines, the
Figure 1 The received power measurements at 5.9 GHz for V2V vehicles with blue lines, the reflected rays with green dashed lines,
communication in an urban environment. and the diffracted rays with magenta dash-dotted lines.

56 ||| IEEE vehicular technology magazine | june 2015

of standardized protocols and applications before their density of vehicular traffic significantly impact the signal
deployment has become the primary focus of simulation propagation in these environments. Therefore, the classifi-
campaigns. With the finalization of the Release 1 standard- cation of environments should be taken with a grain of salt
ization package by the European Telecommunications because it is not uncommon to have an urban area that
Standards Institute (ETSI) and European Committee for has open spaces akin to highways or neighborhoods with
Standardization/International Organization for Standard- low-rise buildings that could be arranged similarly to a
ization (CEN/ISO) following the European Commission typical suburban setting. This is confirmed by numerous
mandate M/453 [4] and the recent announcement by the measurement studies that have reported highly variable
U.S. Department of Transportation to move forward with and often contradicting path-loss exponents for the same
V2V communication [5], it is clear that the implementation environment: 1) 1.6–2.9 on highways [6]–[8], 2) 2.3–3.5 in a
of V2V communication is imminent. For this reason, we pay suburban environment [8], [9], and 3) 1.8–3.4 in an urban
particular attention to the usability aspects of vehicular environment [8], [10]. Similarly, mean delay spreads in the
channel models. In other words, we investigate whether range of 140–400 ns, 80–104 ns, and 150–370 ns have been
or not the state-of-the-art models can be used for efficient reported for highway, suburban, and urban scenarios,
simulations of vehicular communication systems on a large respectively [11]–[13]. Therefore, designing the propaga-
scale. Using an appropriate channel model is critical for ac- tion models with a specific environment in mind cannot
curately evaluating vehicular protocols and applications ensure that the model will accurately apply to a different
before the actual deployment. To that end, we provide environment of the same class. For this reason, the prefer-
guidelines for choosing a suitable channel model depend- able method is to design propagation models that consid-
ing on the type of protocol or application under evaluation, er the specific objects in the environment along with their
available geographical information, and time constraints accurate dimensions and locations.
with respect to the simulation execution.
Link Types
Specific Considerations for Vehicular Channels In addition to the nature of the propagation environment,
it is also important to distinguish between different link
Environments types as they exhibit vastly different propagation proper-
The radio propagation is strongly influenced by the type ties. In V2V channels, the transmitter and receiver anten-
of environment where the communication occurs. In nas are usually mounted on the vehicle rooftop and both
vehicular communications, the most important objects vehicles are mobile, whereas in V2I channels, the base sta-
that influence the propagation are buildings, vehicles tion (or access point) is stationary and may be elevated.
(both static and mobile), and vegetation. A combination of V2P communication links are envisioned to support Vul-
different object types, as well as their number, size, and nerable Road User safety applications [14]. Differences in
density, has a profound impact on the radio propagation. mobility, shadowing, and relative height of the transmitter
While identifying different object types is not difficult, the and receiver antennas create significant differences in
classification of vehicular environments that they create is reflections, diffractions, and scattering patterns of the
not a trivial task. Therefore, the environments where transmitted waves [15].
vehicular communication occurs, are most often qualita-
tively classified as highways, suburban areas, and rural Vehicle Types
areas. Figure 3 shows the most often analyzed propaga- Different types of vehicles (e.g., personal vehicles, com-
tion environments. The varying presence, locations, and mercial vans, trucks, scooters, and public transportation
density of roadside objects as well as the velocity and vehicles) have distinct dimensions and mobility dynamics.

(a) (b) (c)

Figure 3 The qualitative classification of typical vehicular communication environments and dedicated propagation obstacles: (a) An
urban area comprises high-rise buildings, moving vehicles, parked vehicles, and occasional foliage. (b) A suburban area comprises low-
rise buildings, moving vehicles, and frequent foliage. (c) A highway comprises moving vehicles. (Images courtesy of Google Earth.)

june 2015 | IEEE vehicular technology magazine ||| 57

Therefore, the models for the propagation characteristics high Doppler shift, and frequency-selective fading caused
of one vehicle type are not readily applicable to other by both mobile and static objects. Because modeling all of
types. The distinct features of vehicle types have an these aspects is a complex task, the most common
impact on the propagation modeling even if the vehicle approach thus far has been piecemeal modeling, wherein
itself is not the transmitter or the receiver. For example, the problem is split into manageable parts and modeling is
the additional attenuation caused by a large truck blocking performed on one or more parts.
the line of sight (LOS) between the transmitter and the
receiver can be more than 20 dB higher than the attenua- Large-Scale Propagation
tion caused by personal vehicles [16], [17]. The most commonly used large-scale propagation model
for vehicular channels is the log-distance path loss model
Objects [15], with the associated path loss exponent being estimat-
Regardless of the link types, vehicular propagation envi- ed based on empirical measurements. Cheng et al. [9] fit
ronments also consist of a number of different types of the dual-slope log-distance model with suburban channel
objects that impact the signal propagation. The level of measurements. Similar approaches are used in various sce-
impact varies depending on the object type, the link type, narios, including highways [6], rural and highway scenarios
and the environment. For instance, mobile objects (i.e., [7], urban intersection scenarios [10], and garage scenarios
vehicles on the road) are more important for modeling [19]. In addition, other large-scale models are used. The
vehicular channels in highway environments because the geometry-based efficient propagation model for V2V com-
communication between the transmitting and receiving munication (GEMV2) proposed by Boban et al. [16] uses dif-
vehicles on highways usually happens over the road sur- ferent types of path loss models for LOS and non-LOS
face. On the other hand, in urban environments with two- conditions (the two-ray ground reflection model [15] and
dimensional topology, the communicating vehicles are log-distance path loss, respectively), whereas the model
likely to be on different streets. In this case, along with proposed by Maurer et al. [18] uses ray-tracing techniques
mobile objects, accounting for static objects is critical for [20] to model the large-scale propagation effects.
modeling vehicular channels since both types of objects
are sources of shadowing, reflections, and diffractions [18]. Small-Scale Fading
A number of measurement campaigns have also indi- In addition to large-scale propagations, a number of mod-
cated that the LOS condition is a key factor in modeling els have been proposed to account for the small-scale
V2V propagation channels. For example, measurements signal variations caused by multipath propagations and
performed by Tan et al. [13] have shown that regardless Doppler effects due to the mobility of vehicles and
of the propagation environment (e.g., highway or urban objects in their surroundings. Similar to the large-scale
scenarios), non-LOS channels have noticeably larger propagation modeling, the small-scale fading is usually
root-mean-square delay spreads than that of LOS chan- modeled using well-known distributions, such as the
nels. This is due to the stronger signal attenuation and Weibull [21], Nakagami [9], and Gaussian [16] distribu-
multipath effects caused by an increasing number of re- tions, with parameters estimated from the measurement
flections and diffractions. data. For instance, in the GEMV2 model [16], the small-
scale fading is modeled using the Gaussian distribution
Classification and Description of with varying standard deviation depending on the num-
Vehicular Channel Models ber of vehicles and density of objects in the area. Ray-
In this section, we give an overview and recent advances tracing techniques have also been used to estimate the
in vehicular propagation and channel modeling. The mod- small-scale fading in various environments [7], [18].
els described in this section are chosen mainly based on It is worth noting that the propagation characteristics
their usability (e.g., scalability, database input require- of vehicular communications are highly dependent on
ments, and extensibility to different environments) and the existence of the LOS path, as indicated by empirical
ability to realistically model a wide range of environ- measurements (e.g., [6] and [16]). As a result, the large-
ments. The models presented in this section have been and small-scale propagation characteristics are usually
validated against measurements. We classify the models modeled separately for LOS and non-LOS links. Mecha-
based on the propagation mechanism they model, the nisms to differentiate the link types (e.g., LOS, non-LOS
implementation approach they employ, and the channel due to vehicles, non-LOS due to buildings) are included
properties they implement. in recent models [6], [16], [19].

Propagation Mechanisms Channel Model Implementation Approaches

The key distinguishing aspects of vehicular channels are Depending on the implementation approach and the avail-
varying path loss across space (e.g., different environ- ability of geographical information, the models can be clas-
ments) and time (e.g., different time of day), potentially sified based on their implementation approach as follows.

58 ||| IEEE vehicular technology magazine | june 2015

Geometry-Based Models
■■ Ray-tracing models [20] are the most commonly used We pay special attention to the usability of
geometry-based (GB) deterministic (GBD) models for the models, including the complexity of
vehicular channel modeling. Ray-tracing methods implementation, scalability, and data input
require a detailed description of the propagation envi- requirements.
ronment to produce the actual physical propagation
process for a given environment to accurately calculate Nongeometry-Based Models
the channel statistics. The model proposed by Maurer Most nongeometry-based (NG) models conform to a specific
et al. [18] is an example of a model that is based on ray procedure: they measure the channel characteristics in a
tracing. It calculates the channel statistics by analyzing specific environment and adjust the parameters of the path
the 50 strongest propagation paths between the trans- loss, shadowing, and the small-scale fading accordingly. One
mitter and the receiver. A more scalable ray-tracing data of the most widely used NG models is the tapped-delay line
interpolation and interfacing model (RADII) is proposed (TDL) model. Each tap in this model represents signals
by Pilosu et al. [22]. RADII uses a combination of pre- received from several propagation paths; each with a differ-
processing ray-tracing techniques to compute the aver- ent delay and different type of Doppler spectrum. Based on
age attenuation of each region of interest (ROI) and an extensive measurement campaign performed in urban,
uses an interpolation technique to compute attenuation suburban, and highway environments with two levels of traf-
between connected ROIs offline so that the simulations fic density (high and low), Sen and Matolak [21] proposed a
can use a lookup table without the need to recalculate TDL model for each region. The Markov chain technique is
the channel statistics. used to model the multipath component, whereby the non-
■■ Simplified GB models consider the geometric proper- stationarity property of the model is incorporated by adding
ties of the surroundings while simplifying geometric the persistence process that accounts for the finite lifetime
calculations by extracting some of the channel statis- of the propagation paths. Similarly, Wang et al. [24] pro-
tics, either from measurements or simulations. Exam- posed a TDL-based channel model with birth/death process-
ples of these models have been described by Cheng et es to account for a sudden appearance of a LOS component.
al. [9] and Sun et al. [19]. In these models, channel
parameters are estimated separately for a given mea- Properties of the Model
surement scenario. For instance, two sets of model Since the focus of this article is the usability of the model
parameters are estimated for two suburban environ- for protocol and application evaluations, we identify the
ments in Cheng et al. [9] and several sets of parame- most important properties that enable the usability of
ters are estimated in a parking garage environment by the model. Based on these properties, Table  1 qualita-
Sun et al. [19], depending on the LOS/non-LOS condi- tively summarizes the state-of-the-art propagation and
tion and the locations of the transmitter and the channel models.
receiver. Karedal et al. [7] propose a more complex
model that considers four-distinct signal components: Spatial and Temporal Dependency
1) LOS, 2) discrete components from mobile objects, While the small-scale fading models account for the time-
3) discrete components from static objects, and 4) dif- varying signal attenuation due to propagation effects
fuse scattering. Model parameters were extracted from (e.g., reflections and scattering), measurements have
measurements in highway and suburban environ- demonstrated that the variation in signal attenuation is
ments. Abbas et al. [6] designed a model that differen- strongly correlated over both time and space. This spatial
tiates LOS and non-LOS conditions of a link based on a and temporal dependency arises from the static and
Markov chain probabilistic model. The transition prob- dynamic physical world features, respectively. In other
abilities between conditions are estimated from the words, different communication links in an area are affect-
probability distributions of the LOS and non-LOS com- ed by the same effects (generated by, e.g., obstructing
ponents measured in different environments. Mangel objects, ambient noise, and interference). These links
et al. [10] developed a channel model that incorpo- exhibit similar characteristics due to spatial correlation.
rates relevant information about street intersections On the other hand, mobility of vehicles and varying traf-
(e.g., street width, existence of buildings on intersec- fic density lead to the signal attenuation that is correlated
tion corners, and so on.). The model is fitted to the over time (i.e., temporal dependency). The ability of a
measurements that we performed at representative channel model to include the spatial-temporal dependen-
intersections. Boban et al. [16] developed a GEMV2 cy is shown in Table 1.
model that uses outlines of vehicles, buildings, and
foliage to distinguish three types of links LOS, non-LOS Temporal Variance and Nonstationarity
due to other vehicles [23], and non-LOS due to build- In addition to the spatial and temporal correlation, mea-
ings or foliage. surements have revealed that vehicular channels exhibit

june 2015 | IEEE vehicular technology magazine ||| 59

 Table 1 The classification of propagation and channel models.
Propagation Scale Properties of the Channel Models

Spatial-
Implementation Temporal Antenna Scalability and
Model Large Small Environment Approach Dependency Nonstationarity Extensibility Applicability Configuration Link Complexity
Sun Log distance — Parking GBD — — — — — Large, O ^1 h
et al. [19] garage
Fayziyev Measurement- — Tunnel GBD — — — — — Large, O ^1 h
et al. [33] fitted impulse
response
Abbas Log distance —* Highway GBS — — — — — Large, O ^1 h
et al. [6]
Sen and — Weibull Urban, NG # # — — — Large, O ^1 h
Matolak highway
[21]
Wang — Rician All GBS # # # # — Large, O ^1 h
et al. [24]

Cheng Dual-slope Nakagami Suburban NG — — — — —

60 |||
Large, O ^1 h
et al. [9] log distance

Mangel Log distance Nakagami Intersections GBD # — # — — Large, O ^1 h

et al. [10] (LOS) Nor-
mal (NLOS)

Karedal Log distance Simplified Rural, GBS # # # — # Medium,

et al. [7] ray tracing highway O ^R + V h

Boban Two-ray (LOS) Normal+ All GBD # # # # — Large, O ^V h

et al. [16] log distance
(NLOS)
Pilosu Preprocessed Prepro- All GBD # # # # # Small
et al. [22] ray tracing cessed ray 2 O ^^R + V h2h
tracing
Maurer Ray tracing Ray tracing All GBD # # # # # Small,
et al. [18] 2 O ^^R + V h2h

R and V denote the number of roadside objects and vehicles, respectively.

)
Only spatial correlation of shadow fading is considered.
+
Signal deviation depends on the number of vehicles and static objects in the area.

IEEE vehicular technology magazine | june 2015

strong nonstationarity; i.e., in addition to a change in
the channel state, the channel statistics may also We identify the less-explored aspects of
change, especially if the channel involves vehicles that vehicular channel modeling, including
travel at high speeds. The nonstationarity of the model under-explored environments, modeling
also arises from static and mobile objects that could links other than V2V, and the interaction
cause a sudden appearance/disappearance of the LOS
with future 5G cellular systems.
component. Table 1 classifies the channel models based
on their ability to simulate the nonstationarity property
of vehicular channels. Note that all of the models in ability to support different antenna configurations (e.g.,
Table 1 can simulate the temporal variance property of multiple-input, multiple-output antenna configuration).
the channel.
Scalability
Extensibility to Different Environments In addition to the properties of the channel itself, we
With regard to the applicability of a model to different also classify the models based on their efficiency, which,
environments, we distinguish between the channel mod- in turn, determines the model’s scalability. Given the
els that were calibrated by extracting the pertinent increase in demand for efficient evaluations of vehicular
parameters from measurements at a specific set of loca- applications, it is necessary for channel models to be
tions and those that have the ability to model effects able to support large-scale simulations. The efficiency of
beyond those captured at particular locations. Since the the model depends largely on the complexity of the
former category depends on measurements, these mod- mechanisms employed for calculating the channel sta-
els can give no accuracy guarantees for locations with tistics. In general, models that utilize the ray-tracing
considerably different characteristics. On the other techniques can provide good accuracy but do not scale
hand, models that consider geometry-specific informa- well. The scalability properties of the channel models
tion of the simulated area can give some insight for envi- are assessed qualitatively and shown in Table 1.
ronments beyond those characterized by measurements.
For this reason, we indicate the extensibility of the Comparison of Selected Channel Models
model in Table 1 to describe whether or not the model Figure 4 shows the comparison of the received power
can be generalized to other propagation environments results obtained for a V2V measurement campaign per-
beyond those that were used to generate the model. formed in Porto, Portugal, with four models:
2
■■ GEMV model [16]
Applicability ■■ two models proposed by Cheng et al. [9]: the single-
Since the primary purpose of vehicular channel models is slope model, with a path loss exponent of 2.75 and
to support the realistic development of vehicular and standard deviation for fading of 5.5; and the dual-slope
intelligent transportation system (ITS)-related applica- model, with a path loss exponent of 2.1 and fading
tions, we analyze the ability of models to consider appli- standard deviation of 2.6 dB for distances below 100 m
cation-specific scenarios. For example, instead of and path loss exponent of 3.8 and fading standard
analyzing general highway scenarios, Bernado et al. [25] deviation of 4.4 dB for distance above 100 m
performed measurements and subsequently developed ■■ log-distance path loss model with log-normal shadow
channel models for different applications on highways, fading; we used a path loss exponent of 2.5 and fading
such as merging lane scenarios, traffic congestion scenar- deviation of 5 dB.
ios, scenarios in which a car approaches a traffic jam, and The parameters for the log-distance path loss model
so on. While the classification by propagation environ- have been set to approximate the values extracted from
ments can be used to identify some practical applica- the measurement data. Note that the actual locations of
tions, some applications require dedicated channel vehicles surrounding the communicating vehicles dur-
characterization (e.g., precrash and postcrash warning ing the measurements are unknown. In case of the
[25]). In Table 1, we identify the channel models that can GEMV2 model, this implies that their locations cannot be
be applied to other use cases in addition to the ones for used in the model itself. Instead, simulated locations
which they are originally calibrated. were used, thus reducing the estimation accuracy of
non-LOS links.
Antenna Configuration The main takeaway from the comparison is that if the
Related to the channel model’s ability to incorporate measurements for a specific environment are not avail-
small-scale fading is the ability to support different able, then the NG models provide inconsistent results.
types of antenna configurations that exploit the positive For example, the path loss exponent for distances above
and counter the negative effects of small-scale fading. 100 m in the dual-slope Cheng model is clearly too high,
Therefore, we include information about the model’s thus resulting in unrealistically low received power

june 2015 | IEEE vehicular technology magazine ||| 61

 -20 Experiments
GEMV2
Cheng Single Slope
-40 Cheng Dual Slope
Log−Distance

-60
Tx Power (dBm)

-80

-100

-120

-140
0 100 200 300 400 500 600 700
Time (s)

Figure 4  A comparison of the received power results estimated by four models against the results obtained from V2V measurements
performed in Porto, Portugal. The mean absolute error of each model (i.e., absolute difference for each measured data point): 6.7 dB for the
GEMV2 model, 11.1 dB for the Cheng single-slope model, 14.4 dB for the Cheng dual-slope model, and 7.7 dB for the log-distance model.

values above 100 m. If the model’s parameters are ex- Ultimately, choosing the right model should depend
tracted from the measurement data for a given location, on the type of application and/or protocol that needs to
then the estimate is better, as shown by the log-distance be evaluated, constrained by processing power and avail-
model in Figure 4. However, if the geographical infor- ability of the required data (either geographical or mea-
mation is available, then the GBD models, such as the surements). To that end, the flowchart shown in Figure 5
GEMV2 model, are a better choice. provides a guideline in choosing a suitable channel model.
For example, if only system-wide performance analysis is
Guidelines for Choosing a Suitable Channel Model required (e.g., overall packet delivery ratio, average end-to-
The models listed in Table 1 differ in many ways and offer end delay, and so on), any type of model [NG, GB stochastic
different tradeoffs between accuracy and complexity/scal- (GBS), and GBD models] might be suitable. However, if an
ability. Stochastic models that do not require any informa- application requires network topology statistics (e.g., the
tion about the environment are simple and highly scalable, number of neighboring vehicles) or location-dependent
at the expense of lower accuracy. GB models trade off scal- statistics (e.g., the packet delivery rate or end-to-end delay
ability for accuracy, where the trade off can differ quite sig- in an area with rapid channel fluctuations), GB models that
nificantly from one model to another. Ray-tracing models can model dynamic link transitions and small-scale varia-
(e.g., [18]) require detailed information about the propaga- tions should be used. For safety-critical applications that
tion environment (which can be hard to collect) and high- disseminate time-sensitive information about a specific
er computational power. On the other hand, the model safety event, GBD models are the best choice.
proposed by Abbas et al. [6] only requires information on Once the channel model category for a specific applica-
the type of the environment to estimate the channel statis- tion is identified, the suitable channel model should be cho-
tics. Thus, it is highly scalable and can provide environ- sen based on the availability of geographic/measurement
ment-specific but not location-agnostic channel data and processing power. If the complete geographic
information. Simplified GB models that consider the actual information (e.g., location, dimensions, and material prop-
locations of objects (e.g., the GEMV2 model [16]) can erties of vehicles, buildings, and foliage) is available and
achieve good accuracy/scalability tradeoff, offering a large processing speed is not an issue, then ray-tracing-based
gain on scalability compared to the ray-tracing models, GBD models (e.g., the model proposed by Maurer et al. [18])
while providing sufficient accuracy and ease of use. could be used for maximum accuracy. If limited information

62 ||| IEEE vehicular technology magazine | june 2015

 System-Wide Type of Application Safety-Critical
Analysis or Study? Application

Network Topology or
Location-Dependent
Statistics

No Geographic Yes Yes Geographic No No Detailed Yes

Information Information Geographic Information
Available? Available? Available?

Yes Is Speed No
an Issue? Yes Is Speed
an Issue?
NG Models GBS/GBD GBS/GBD
— Sen and Matolak [21] — Sun et al. [19] — Boban et al. [16] No Suitable No
(Urban Highway)z — Abbas et al. [6] — Karedal et al. [7] Models
— Cheng [9] (Suburban) — Wang et al. [24] — Pilosu et al. [22]
— Mangel et al. [10] — Maurer et al. [18] GBD Models
GBD Model — Pilosu et al. [22]
— Boban et al. [18] — Maurer et al. [18]

Figure 5  The guidelines for choosing a suitable channel model.

about the propagation environment is available (e.g., densi- in Figure 4, these models cannot capture the characteris-
ty of vehicles and surrounding objects) and the processing tics of vehicular channels, i.e., rapid transitions between
speed is important, then simplified GB models can be used LOS and non-LOS conditions and changes in delay and
(e.g., [7] and [16]). Otherwise, other GB models, such as [9] Doppler spreads. Consequently, simple models were
and [24], that require only a qualitative type of simulated shown to exhibit poor performance in terms of link-level
environment may be used. modeling, particularly in complex environments [26]. A
way forward in this respect would be to combine GB scal-
Toward Realistic and Efficient able propagation models (e.g., [6] and [16]), which are able
Vehicular Channel Modeling to distinguish between different LOS conditions and envi-
In this section, we discuss the recent trends in the vehicular ronments, with small-scale channel models that are able to
channel modeling, including the need for models that are provide appropriate delay and Doppler statistics for each
usable in large-scale vehicular network simulators. We also representative environment (e.g., [9] and [25]). Finally,
discuss vehicular channel emulation as an alternative attempts should be made to implement such realistic mod-
approach for realistic protocol and application evaluations. els in large-scale network simulators to enable realistic
Finally, we point out open problems in the area of propaga- evaluations of protocols and applications.
tion and channel modeling that require further attention. The GEMV2 model is an example of a computationally
efficient channel model that can model the signal propa-
Efficient Models for Realistic Large-Scale Simulation gation in a large set of environments (e.g., highway, rural,
As the deployment phase in main ITS markets is getting urban, and complex intersections) and is able to simulate
closer, realistic channel models for large-scale simulations city-wide vehicular networks with thousands of communi-
are necessary for the efficient evaluation of applications cating vehicles. It allows importing realistic mobility data
before they are deployed in the real world. However, chan- from the simulation of urban mobility and building/foli-
nel and propagation models currently used to simulate age outlines from OpenStreepMap [27]. The source code
V2V and V2I communication links in vehicular ad hoc net- of the GEMV2 model is available at http://vehicle2x.net/.
work (VANET) simulators [e.g., NS-3 (www.nsnam.org)] are Apart from the propagation-related statistics, the GEMV2
based on simple statistical models (e.g., free space and log- model allows for the analysis of networking related met-
distance path loss [15]) that are used indiscriminately for rics, such as packet delivery rates, effective transmission
ail environments where communication occurs. As shown range, and neighborhood size (Figure 6).

june 2015 | IEEE vehicular technology magazine ||| 63

 vehicles) are rare, despite their
considerably different dimen-
sions and road dynamics. For
example, the mobility of scoot-
ers and motorcycles is notably
different from that of personal
cars [30]. Combined with their
smaller dimensions and lack of
roof for antenna placement, the
mobility of scooters indicates
Figure 6 The visualization of neighborhood size generated by the GEMV2 model. For each vehicle,
the colored bar represents the number of vehicles with which it can directly communicate (i.e., its
that the propagation characteris-
neighbors). Warmer and taller bar colors indicate more neighbors. (Image courtesy of Google Earth.) tics for scooters can be signifi-
cantly different from that of
Vehicular Channel Emulations personal cars. Similarly, recent studies have shown that,
Performing experiments with real hardware in realistic en- in the same environment, commercial vans and trucks
vironments is inherently the most realistic approach to experience different channel propagation characteristics
characterize wireless vehicular channels. However, cost from personal cars. This resulted in different reliable
and repeatability issues make this approach infeasible for communication ranges and packet error rates [31]. There-
large-scale evaluations (e.g., involving tens or hundreds of fore, further studies are needed to investigate channel
vehicles). At the other end of the spectrum is channel sim- characteristics for vehicles other than personal cars.
ulation, which can ensure high repeatability, configurabili-
ty, and manageability. However, designing a realistic Underexplored Environments
channel simulator is a challenge, since the simulation envi- While vehicular communications can take place in any
ronment must either be highly detailed to account for all scenario, signal propagation measurements are usually
aspects of the real system or be able to make certain as- performed in common environments (e.g., those in Fig-
sumptions about the real world. Between the channel ex- ure 3) and measurements in other environments, such as
perimentation and simulation lies the channel emulation, multilevel highways, tunnels, parking garages, bridges,
where parts of the real communication systems are used in and roundabouts, are quite rare. For example, V2V signal
conjunction with the simulated ones, with the main goal of propagation measurements in a parking garage have been
maintaining the repeatability and configurability of simula- performed in one study to date [19]. Similarly, there are
tion environments, combined with a high level of realism of only a few measurements performed in tunnels [32], [33]
a testbed. One example is the Carnegie Mellon University and on-bridge environments [25]. Further measurements
(CMU) wireless emulator [28], where the emulator inputs and modeling studies are particularly necessary for envi-
the signals generated by real devices, subjects them to ronments with distinct application use cases that can
simulated realistic signal propagation and feeds the signals occur in them (e.g., service updates in parking garages,
back into the real devices. The emulator was shown to be- cooperative awareness functionalities without global
have realistically by comparing its output to the real-world positioning systems in tunnels, and so on).
measurements. In terms of the vehicular channel emula-
tion, a detailed model for small-scale statistics of vehicular V2X Channels
channels has been developed and implemented in NS-3 by Despite significant differences between V2V, V2I, and V2P
Mittag et al. [29]. We compared the results from the simula- communications, the propagation characteristics of V2I and
tor with those generated by the CMU wireless emulator V2P channels are not as well researched as V2V channels.
[28] and found a good match in terms of frame reception Part of the reason is that V2I systems resemble existing cel-
rate results. Therefore, provided that the size of the sys- lular systems, where one of the communicating entities
tem is limited to a few dozen vehicles, the channel emula- (base station) is stationary, while the other (user equip-
tion is a feasible approach for a realistic and reproducible ment) is mobile. However, typical positioning of static
vehicular channel evaluation. (infrastructure) nodes in V2I communications is unique for
vehicular communications: on highways, roadside units will
Open Research Issues be placed close to the road at heights considerably lower
than that of cellular base stations (see, e.g., current efforts
Channel Models for Different Vehicle Types within the Amsterdam Group:https://amsterdamgroup.
Vehicular channel measurements and modeling have pri- mett.nl). In urban areas, the most beneficial locations are
marily focused on personal cars (e.g., [6], [10], and [16]). near large intersections. In addition to V2I measurement
Studies dealing with other types of vehicles (e.g., com- campaigns [13], [34], there are only a few dedicated V2I
mercial vans, trucks, scooters, and public transportation channel models. Acosta-Marum and Ingram [12] developed

64 ||| IEEE vehicular technology magazine | june 2015

TDL models to capture the joint Doppler-delay characteris-
tics of V2I channels. The models are based on extensive A number of measurement campaigns have
measurements in urban, suburban, and highway environ- indicated that the LOS condition is a key
ments. Furthermore, a study performed by Gozalvez et al. factor in modeling V2V propagation
[34] showed that V2I communications in urban areas is channels.
highly variable, with both static and mobile objects creat-
ing a considerably changing channel over both space and Finally, we discuss the less-explored aspects of vehicu-
time. Therefore, there exists a need for further studies to lar channel modeling and point out the areas where fur-
investigate V2I channels. ther research efforts are required.
In terms of V2P communications, recent studies by Wu
et al. [35] and Anaya et al. [14] have explored the basic Author Information
channel properties of V2I links. Channel models for differ- Wantanee Viriyasitavat is a lecturer in the Faculty of
ent communication technologies that can enable V2P com- Information and Communication Technology at Mahidol
munications (e.g., dedicated short-range communication, University, Bangkok, Thailand. During 2012–2013, she was a
Wi-Fi, and cellular-based systems) need to be explored research scientist in the Department of Electrical and Com-
further. Therefore, there is much work to be done to fully puter Engineering at Carnegie Mellon University (CMU),
understand and model the V2P communication channels. Pittsburgh, Pennsylvania. She received her B.S./M.S., and
Ph.D. degrees in electrical and computer engineering from
V2X and Fifth Generation CMU in 2006 and 2012, respectively. From 2007 to 2012, she
As recent research efforts on future fifth-generation (5G) was a research assistant at CMU, where she was a member
cellular networks start to look more deeply into ITS-relat- of General Motors Collaborative Research Laboratory and
ed applications [36], it is reasonable to expect gradual was working on the design of a routing framework for safe-
convergence of the efforts on the channel modeling for ty and nonsafety applications of vehicular ad hoc wireless
vehicle-to-x (V2X) and 5G systems. For example, as the networks. Her research interests include traffic mobility
delay requirement becomes more stringent for many 5G modeling, network connectivity analysis, and protocol
scenarios, the proposed system overcomes the main design for wireless ad hoc networks.
obstacle for use in a vehicular setting (i.e., lack of low- Mate Boban is a research scientist at NEC Laborato-
latency guarantees). Initial steps needed for enabling V2X ries Europe. He earned his Ph.D. degree in electrical and
systems through 5G, along with the related requirements computer engineering from Carnegie Mellon University
for channel modeling, are discussed by Kyrolainen et al. and his diploma in Informatics from University of Zagreb.
[37]. Additionally, when applied to highly mobile termi- He is an alumnus of the Fulbright Scholar Program. His
nals, the device-to-device (D2D) concept in 5G systems current research is in the areas of cooperative intelligent
shares many similarities with V2X communications; transportation systems, wireless communications, and
therefore, efforts on modeling D2D channels (e.g., [38]) networking. He received the Best Paper Award at the IEEE
can benefit from the existing V2V channel modeling work VTC-Spring 2014 and at IEEE VNC 2014. More information
and vice versa. can be found on his Web site: http://mateboban.net.
Hsin-Mu Tsai is an assistant professor in the Depart-
Conclusions ment of Computer Science and Information Engineering
This article provides a survey of recent developments in and Graduate Institute of Networking and Multimedia at
the area of propagation and channel modeling for vehic- National Taiwan University. He received his B.S.E. degree
ular communications. We pay special attention to the in computer science and information engineering from
usability aspects of the models, including their suitabili- National Taiwan University in 2002, and his M.S. and Ph.D.
ty for large-scale evaluations of protocols and applica- degrees in electrical and computer engineering from
tions for future cooperative ITSs. We first discuss the Carnegie Mellon University in 2006 and 2010, respectively.
key channel characteristics that distinguish the vehicu- His recognitions include the 2014 Intel Labs Distinguished
lar communications from other types of wireless com- Collaborative Research Award, 2013 Intel Early Career
munications. Next, based on the distinguishing features, Faculty Award (the first to receive this honor outside of
we classify and summarize the state-of-the-art vehicular North America and Europe), and National Taiwan Univer-
channel and propagation models based on the propaga- sity’s Distinguished Teaching Award. He served as the
tion mechanisms they model and their implementation workshop cochair for the first ACM Visible Light Commu-
approach. In addition, we provide guidelines for choos- nication System Workshop in 2014, and TPC cochair for
ing a suitable channel model, depending on the type of ACM VANET 2013. His research interests include vehicu-
protocol or application under investigation and consid- lar networking and communications, wireless channel
ering the availability of geographical information and and link measurements, vehicle safety systems, and visi-
processing power available for simulation execution. ble light communications.

june 2015 | IEEE vehicular technology magazine ||| 65

 Athanasios V. Vasilakos is currently a professor [15] T. S. Rappaport, Wireless Communications: Principles and Practice.
Englewood Cliffs, NJ: Prentice Hall, 1996.
with Lulea University of Technology, Sweden. He has [16] M. Boban, J. Barros, and O. Tonguz, “Geometry-based vehicle-to-
authored or coauthored over 200 technical papers in vehicle channel modeling for large-scale simulation,” IEEE Transac-
major international journals and conferences. He is the tions Veh. Technolo., vol. 63, no. 9, pp. 4146–4164, Nov. 2014.
[17] D. Vlastaras, T. Abbas, M. Nilsson, R. Whiton, M. Olback, and F.
author/coauthor of five books and 20 book chapters in Tufvesson, “Impact of a truck as an obstacle on vehicle-to-vehicle
the area of communications. He is a Senior Member of communications in rural and highway scenarios,” in Proc. IEEE 6th
Int. Symp. Wireless Vehicular Communications, 2014, pp. 1–6.
the IEEE. He served or is serving as an editor and/or [18] J. Maurer, T. Fugen, T. Schafer, and W. Wiesbeck, “A new inter-vehi-
guest editor for many technical journals, such as IEEE cle communications channel model,” in Proc. IEEE Vehicular Technol-
ogy Conf., Sept. 2004, pp. 9–13.
Transactions on Network and Services Management, IEEE [19] R. Sun, D. W. Matolak, and P. Liu, “Parking garage channel charac-
Transactions on Cloud Computing, IEEE Transactions on teristics at 5 GHz for V2V applications,” in Proc. IEEE 78th Vehicular
Technology Conf., 2013, pp. 1–5.
Information Forensics and Security, IEEE Transactions on [20] J. D. Parsons, The Mobile Radio Propagation Channel. New York:
Cybernetics, IEEE Transactions on Information Technology Wiley, 2000.
[21] I. Sen and D. W. Matolak, “Vehicle-vehicle channel models for the
in Biomedicine, ACM Transactions on Autonomous and 5-GHz band,” IEEE Trans. Intell. Transp. Syst., vol. 9, no. 2, pp. 235–245,
Adaptive Systems, and IEEE Journal on Selected Areas in June 2008.
[22] L. Pilosu, F. Fileppo, and R. Scopigno, “RADII: A computationally af-
Communications. He is also general chair of the Europe- fordable method to summarize urban ray-tracing data for VANETs,”
an Alliances for Innovation (www.eai.eu). in Proc. Int. Conf. Wireless Communications, Networking Mobile Com-
puting, 2011, pp. 1–6.
[23] M. Boban, T. T. V. Vinhoza, M. Ferreira, J. Barros, and O. K. Tonguz,
References “Impact of vehicles as obstacles in vehicular ad hoc networks,” IEEE
[1] A. Molisch, F. Tufvesson, J. Karedal, and C. Mecklenbräuker, “A sur- J. Select. Areas Commun., vol. 29, no. 1, pp. 15–28, Jan. 2011.
vey on vehicle-to-vehicle propagation channels,” IEEE Wireless Com- [24] X. Wang, E. Anderson, P. Steenkiste, and F. Bai, “Improving the ac-
mun., vol. 16, no. 6, pp. 12–22, 2009. curacy of environment-specific vehicular channel modeling,” in
[2] C. F. Mecklenbrauker, A. F. Molisch, J. Karedal, F. Tufvesson, A. Paier, Proc. 7th ACM Int. Workshop Wireless Network Testbeds, Experimental
L. Bernado, T. Zemen, O. Klemp, and N. Czink, “Vehicular channel Evaluation Characterization, New York, 2012, pp. 43–50.
characterization and its implications for wireless system design and [25] L. Bernadó, T. Zemen, F. Tufvesson, A. F. Molisch, and C. F. Meck-
performance,” Proc. IEEE, vol. 99, no. 7, pp. 1189–1212, 2011. lenbräuker, “Delay and Doppler spreads of non-stationary vehicular
[3] C.-X. Wang, X. Cheng, and D. I. Laurenson, “Vehicle-to-vehicle chan- channels for safety relevant scenarios,” CoRR, vol. abs/1305.3376,
nel modeling and measurements: Recent advances and future chal- 2013.
lenges,” IEEE Commun. Mag., vol. 47, no. 11, pp. 96–103, Nov. 2009. [26] D. Dhoutaut, A. Regis, and F. Spies, “Impact of radio propagation
[4] (2014, Feb.). New connected car standards put Europe models in vehicular ad hoc networks simulations,” in Proc. 3rd Int.
into top gear. [Online]. Available: http://europa.eu/rapid/ Workshop Vehicular Ad hoc Networks, 2006, pp. 69–78.
press-release_IP-14-141_en.htm [27] M. Haklay and P. Weber, “OpenStreetMap: User-generated street
[5] (2014, Feb.). U.S. Department of transportation announces deci- maps,” IEEE Pervasive Comput., vol. 7, no. 4, pp. 12–18, 2008.
sion to move forward with vehicle-to-vehicle communication [28] G. Judd and P. Steenkiste, “Repeatable and realistic wireless ex-
technology for light vehicles. [Online]. Available: http://www. perimentation through physical emulation,” ACM SIGCOMM Comput.
nhtsa.gov/About+NHTSA/Press+Releases/2014/USDOT+to+Move Commun. Rev., vol. 34, no. 1, pp. 63–68, 2004.
+Forward+with+Vehicle-to-Vehicle+Communication+Technology+ [29] J. Mittag, S. Papanastasiou, H. Hartenstein, and E. G. Ström, “En-
for+Light+Vehicles abling accurate cross-layer PHY/MAC/NET simulation studies of
[6] T. Abbas, K. Sjöberg, J. Karedal, and F. Tufvesson, “A measurement vehicular communication networks,” Proc. IEEE, vol. 99, no. 7, pp.
based shadow fading model for vehicle-to-vehicle network simula- 1311–1326, 2011.
tions,” arXiv preprint arXiv:1203.3370v2, 2012. [30] O. Shih, H. Tsai, H. Lin, and A. Pang, “A rule-based mixed mobility
[7] J. Karedal, F. Tufvesson, N. Czink, A. Paier, C. Dumard, T. Zemen, model for cars and scooters (poster),” in Proc. IEEE Vehicular Net-
C. Mecklenbräuker, and A. Molisch, “A geometry-based stochastic working Conf., 2011, pp. 198–205.
MIMO model for vehicle-to-vehicle communications,” IEEE Trans. [31] M. Boban, R. Meireles, J. Barros, P. A. Steenkiste, and O. K. Tonguz,
Wireless Commun., vol. 8, no. 7, pp. 3646–3657, July 2009. “TVR—Tall vehicle relaying in vehicular networks,” IEEE Trans. Mo-
[8] P. Paschalidis, K. Mahler, A. Kortke, M. Peter, and W. Keusgen, “Path- bile Comput., vol. 13, no. 5, pp. 1118–1131, May 2014.
loss and multipath power decay of the wideband car-to-car channel [32] G. Maier, A. Paier, and C. Mecklenbräuker, “Channel tracking for a
at 5.7 GHz,” in Proc. IEEE Vehicular Technology Conf., May 2011, pp. multi-antenna ITS system based on vehicle-to-vehicle tunnel mea-
1–5. surements,” in Proc. 19th IEEE Symp. Communications Vehicular Tech-
[9] L. Cheng, B. E. Henty, D. D. Stancil, F. Bai, and P. Mudalige, “Mobile nology Benelux, 2012, pp. 1–6.
vehicle-to-vehicle narrow-band channel measurement and char- [33] A. Fayziyev, M. PŁtzold, E. Masson, Y. Cocheril, and M. Berbineau,
acterization of the 5.9 GHz Dedicated Short Range Communication “A measurement-based channel model for vehicular communica-
(DSRC) frequency band,” IEEE J. Select. Areas Commun., vol. 25, no. tions in tunnels,” in Proc. IEEE Wireless Communications Network
8, pp. 1501–1516, Oct. 2007. Conf., 2014, pp. 128–133.
[10] T. Mangel, O. Klemp, and H. Hartenstein, “A validated 5.9 GHz non- [34] J. Gozalvez, M. Sepulcre, and R. Bauza, “IEEE 802.11p vehicle-to-
line-of-sight path-loss and fading model for inter-vehicle communi- infrastructure communications in urban environments,” IEEE Com-
cation,” in Proc. 11th Int. Conf. ITS Telecommunications, Aug. 2011, pp. mun. Mag., vol. 50, no. 5, pp. 176–183, May 2012.
75–80. [35] X. Wu, R. Miucic, S. Yang, S. Al-Stouhi, J. Misener, S. Bai, and W.-h.
[11] O. Renaudin, V. Kolmonen, P. Vainikainen, and C. Oestges, “Wide- Chan, “Cars talk to phones: A DSRC based vehicle-pedestrian safety
band MIMO car-to-car radio channel measurements at 5.3 GHz,” in system,” in Proc. IEEE Vehicular Technol. Conf., 2014, pp. 1–7.
Proc. IEEE 68th Vehicular Technology Conf., 2008, pp. 1–5. [36] A. Osseiran, F. Boccardi, V. Braun, K. Kusume, P. Marsch, M. Mater-
[12] G. Acosta-Marum and M. Ingram, “Six time- and frequency-selective nia, O. Queseth, M. Schellmann, H. Schotten, H. Taoka, H. Tullberg,
empirical channel models for vehicular wireless LANs,” IEEE Veh. M. A. Uusitalo, B. Timus, and M. Fallgren, “Scenarios for 5G mobile
Technol. Mag., vol. 2, no. 4, pp. 4–11, Dec. 2007. and wireless communications: The vision of the METIS project,” IEEE
[13] I. Tan, W. Tang, K. Laberteaux, and A. Bahai, “Measurement and Commun. Mag., vol. 52, no. 5, pp. 26–35, 2014.
analysis of wireless channel impairments in DSRC vehicular com- [37] J. Kyröläinen, P. Kyösti, J. Meinilä, V. Nurmela, L. Raschkowski, A.
munications,” in Proc. IEEE Int. Conf. Communications, May 2008, pp. Roivainen, and J. Ylitalo, “Channel modelling 5th generation mobile
4882–4888. communications,” in Proc. 8th European Conf. Antennas Propagation,
[14] J. J. Anaya, P. Merdrignac, O. Shagdar, F. Nashashibi, and J. E. Nara- 2014, pp. 219–223.
njo, “Vehicle-to-pedestrian communications for protection of vul- [38] V. Nurmela, T. Jämsä, P. Kyösti, V. Hovinen, and J. Medbo, “Chan-
nerable road users,” in Proc. IEEE Intelligent Vehicles Symp., 2014, pp. nel modelling for device-to-device scenarios,” COST IC, vol. 1004, pp.
1037–1042. 1–6, 2013.

66 ||| IEEE vehicular technology magazine | june 2015