Vehicular Communications 24 (2020) 100233

Contents lists available at ScienceDirect

Vehicular Communications
www.elsevier.com/locate/vehcom

A PHY/MAC cross-layer design with transmit antenna selection and

power adaptation for receiver blocking problem in dense VANETs
Andy Triwinarko a,b,∗ , Iyad Dayoub a,∗ , Marie Zwingelstein-Colin a , Mohamed Gharbi a ,
Basma Bouraoui a
a
Univ. Polytechnique Hauts-de-France, CNRS, Univ. Lille, ISEN, Centrale Lille, UMR 8520, IEMN, Institut d’Électronique de Microélectronique et de
Nanotechnologie, DOAE, Département d’Opto-Acousto-Électronique, F-59313, Valenciennes, France
b
Network and Hardware Technology Laboratory, Politeknik Negeri Batam, Batam, 29461, Indonesia

a r t i c l e i n f o a b s t r a c t

Article history: The medium access control (MAC) protocol of IEEE 802.11p dedicated to vehicular ad hoc networks
Received 28 July 2019 (VANETs) employs carrier sense multiple access with collision avoidance (CSMA/CA) with distributed
Received in revised form 12 December 2019 coordination function (DCF) which prohibits simultaneous transmissions in the same detection area in
Accepted 17 January 2020
order to avoid possible interference and collision between neighboring vehicles. This prohibition results
Available online 28 January 2020
in temporary blocking of data reception, which reduces the average network throughput. To solve this
Keywords: problem, we propose a physical (PHY)/MAC cross-layer design based on transmit antenna selection (TAS)
V2V communication and transmit power adaptation (TPA). We consider spatial multiplexing zero-forcing Bell-labs layered
ZF-VBLAST space-time (ZF-VBLAST) over multiple-input and multiple-output (MIMO) time-varying flat fading channel
MIMO to be implemented in vehicle-to-vehicle (V2V) communication. The cross-layer approach is implemented
Cross-layer architecture to get the maximal network throughput on the MAC layer by using the channel state information
TAS (CSI) obtained from the PHY layer, while the MIMO spatial multiplexing technique is used to increase
TPA
the spectral efficiency. This design helps transmitters to select the best combination of transmitting
antennas to maximize throughput and also to choose the adequate transmit power level to minimize
neighbors’ interference and collision. Also, this solution comes with a multi-user interference cancellation
method that allows simultaneous transmissions as long as the sum of transmit antennas within the
same radio range does not exceed the number of receive antennas. This paper evaluates the proposed
cross-layer architecture by calculating the average network throughput per V2V links concerning different
network parameters such as the number of vehicles and antennas. The simulation results show that this
solution allows more vehicles to communicate simultaneously and thus significantly improves the average
network throughput compared to 802.11p MAC standard, in particular for VANETs with high density.
© 2020 Elsevier Inc. All rights reserved.

1. Introduction objective of setting up these networks is the improvement of road

Connected vehicles are now becoming a closer reality to pro-
mote the emergence of intelligent transport systems (ITS), which is
expected to be able to overcome various problems in the modern
transportation system. Through the use of wireless communica-
tion technologies, deployment of VANETs systems allow vehicles to
communicate with each other (V2V) and also with the infrastruc-
ture on the road (vehicles-to-infrastructure or V2I). The primary
* Corresponding authors at: Univ. Polytechnique Hauts-de-France, CNRS, Univ.
Lille, ISEN, Centrale Lille, UMR 8520, IEMN, Institut d’Électronique de Microélec-
tronique et de Nanotechnologie, DOAE, Département d’Opto-Acousto-Électronique,
F-59313, Valenciennes, France.
E-mail addresses: andy@polibatam.ac.id (A. Triwinarko), iyad.dayoub@uphf.fr
(I. Dayoub).
safety through communication and also the development of use-
ful applications such as driving assistance, traffic management, and
comfort or entertainment applications [1].
On the other hand, the recent development of wireless com-
munication, such as the use of the MIMO system in the PHY layer,
offers several advantages that can be adapted to VANETs. In [2],
Atia et al. provide a review of research challenges, opportunities,
and also the benefits of employing MIMO signal processing tech-
niques in VANETs. In [3], Moser et al. explain that although MIMO
systems are already standard in both infrastructural wireless LAN
and the mobile cellular network, the IEEE 802.11p standard for
V2V communication still comes without any multi-antenna ap-
proach. Therefore they implemented MIMO-enabling PHY layer en-
hancement for VANETs using Orthogonal Space-Time Block Code
(OSTBC). The simulation study shows that the MIMO-enabling ap-

https://doi.org/10.1016/j.vehcom.2020.100233
2214-2096/© 2020 Elsevier Inc. All rights reserved.
2 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
proach offers higher robustness against short-term fading caused
by the mobility of the vehicle and other channel-caused adverse-
ness. Our previous work in [4] also reviews the benefit of employ-
ing a MIMO system and cross-layer design in VANETs to improve
the overall network performances. By using antenna diversity and
spatial multiplexing, we can create a multi-user MIMO (MU-MIMO)
system for VANETs, which offers a possibility of higher network
throughput.
Moreover, the cross-layer approach that operates in multiple
layers can be used to provide a more efficient and robust proto-
col for answering the distinctive characteristics of VANETs such as
high mobility of the vehicles, rapid change of topology, and the
ability to predict paths. In [19], Khurana et al. compare the per-
formance of MU-MIMO with single-user (SU-MIMO) in realistic
VANET urban and highway road scenarios. They proposed MU-
MIMO in VANETs using combined user and transmitter antennas
selection considering linear precoding and Singular Value Decom-
position (SVD). They show that MU-MIMO is a better choice in
commercial and safety applications of VANETs, as it doubles the
throughput, increase packet delivery rate (PDR) significantly and
reduces end-to-end delay to nearly half. The difference with our
proposed solution is as follows: we considered a ZF-VBLAST encod-
ing schema with greedy QR decomposition and ordered successive
interference cancellation (OSIC), and we are also implementing the
transmit power adaptation (TPA) algorithm.
The current MAC standard of IEEE 802.11p dedicated to VANETs
is using CSMA/CA with DCF that only allow one transmission at
a time and prohibits simultaneous transmissions [5–7]. A receiver
can detect data from all transmitters within its detection range. All
neighbors are forbidden to transmit data during an ongoing trans-
mission to avoid potential interference and collision. This mech-
anism raises a problem called the receiver blocking problem, i.e.,
temporary blocking of reception at the nearby vehicles that conse-
quently results in degradation of the average network throughput.
This MAC protocol underutilizes the capability of the MU-MIMO
VANETs system, in a way that even though the throughput for
one V2V link improves (due to the use of MIMO system), it still
suffers from the overall network throughput degradation because
the receiver blocking problem still exists. However, wireless ad-hoc
communication preferably uses a randomly based access protocol
like CSMA/CA due to its simplicity for allowing users to access
the medium. Therefore, over the past few years, some research ef-
fort has been made to mitigate this blocking effect by proposing
a cross-layer (PHY/MAC) design using MIMO capability in the PHY
layer and also enabling multiple clients doing simultaneous trans-
mission, while keeping the random access property in the MAC
layer.
In [18], Bazzi et al. proposed an OFDMA-based MAC protocol
for VANETs (OBV) to overcome the heavy traffic condition. The
simulation results show that OBV outperforms all reference MAC
protocols, even doubling their throughput under heavy-load net-
work conditions. Nevertheless, it is not mentioned in the article
whether this solution can be implemented on the MU-MIMO ar-
chitecture. In [8], Saleh et al. proposed cross-layer TAS for decision-
feedback detection (DFD) in correlated Ricean MIMO channels. The
proposed cross-layer architecture is a point-to-point single-user
MIMO wireless system with N t transmit antennas and N r receive
antennas, where N t ≤ N r . The transmitter implements some 1 : K
(K ≤ N t ) spatial multiplexer and the receiver employs a DFD us-
ing the ZF detector to cancel interference and improve the detec-
tion of the transmitted packets. The network throughput is cal-
culated using an adopted go-back-n (GBN) protocol at the link
level. The simulation result reveals that although the capacity-
based antenna selection (AS) is more robust to imperfect channel
estimation, the proposed cross-layer AS delivers higher throughput
gains than the capacity-based AS. This transmission scheme is then
adopted by Mao et al. in [9] for multi-user mobile ad hoc networks
(MANETs) environment. To maximizing the network throughput of
the MANETs, they design the cross-layer architecture using ZF-
BLAST by selecting the optimum transmitter’s antenna combina-
tion. The network throughput is calculated using two adopted
automatic-repeat-request (ARQ) retransmission protocol at the data
link layer, i.e., GBN and selective repeat (SR) protocol. The simula-
tion shows the improvement of network throughput by allowing
more than one pair of nodes to communicate simultaneously for
both ARQ protocols. In [10], Bouraoui et al. extend the work of [9]
for VANETs-V2V communication by proposing a PHY/MAC cross-
layer design based on transmit AS with a dedicated MAC protocol.
The proposed solution allows vehicles to transmit data simultane-
ously without interfering with each other as long as a hypothesis
called antenna quantity restriction (AQR), i.e., the total number of
transmit antennas within the receiver’s detection range does not
exceed the number of antennas of the receiver, is fulfilled. Indeed,
the receiver blocking problem is mitigated, and the simulation re-
sults show the improvement of the average network throughput.
However, we can observe that the AQR condition is very restrictive
in the context of VANETs with high density. The detection area
likely contains a large number of transmitting vehicles.
Recent work in [21,22], Belmekki et al., study the cooperative
vehicular communication at a road intersection in the presence of
interference. They show that, in line of sight (LOS) scenario, di-
rect transmission is better for high densities of vehicles. In this
paper, we extend the work in [10] to mitigate the receiver block-
ing problem in dense VANETs by proposing a PHY/MAC cross-layer
design based on TAS and TPA. We consider spatial multiplexing
(ZF-VBLAST) over MIMO time-varying flat fading channel to be
implemented in V2V communication. The cross-layer approach is
implemented to get the maximal network throughput on the MAC
layer by using the symbol-error-rate (SER) information obtained
from the PHY layer, while the MIMO spatial multiplexing tech-
nique is used to increase the spectral efficiency. The critical idea
added to the work in [10], is to add a TPA mechanism in order to
prevent high interference from nearby vehicles. The use of a con-
stant transmit power could lead to unnecessary large transmission
ranges, and consequently to the point of saturation that violate the
AQR condition. We propose to enhance the cross-layer architecture
using transmit AS by adding the TPA algorithm so that the number
of simultaneous transmission in the same detection zone always
satisfies the AQR condition event in dense VANETs.
We organize the remainder of this paper as follows. Section 2
will explain the system model, the receiver blocking problem in
dense VANETs, and also describe the proposed cross-layer design.
Section 3 presents the performance analysis, which explains the
simulation environment and also the obtained results. Finally, sec-
tion 4 concludes this paper. The notations adopted in this paper
are as follows: matrices are represented as bold uppercase letters,
vectors as bold lowercase letters, and complex scalars as italic low-
ercase letters. (.) T and (.) H are the transposes and the conjugate
transpose (Hermitian); r i , j is the (i , j )th entry of the upper trian-
gular matrix R, xi and ni are the ith entry of the vector [x1 x2 ] T
and n respectively; Real and Complex sets are denoted as R and
˜ and (.) re-
C ; symbols multiplied by Q H or
are denoted by (.)
spectively.
2. Cross-layer design based on TAS and TPA
In this section, we provide an overview of the system model
and the PHY/MAC cross-layer design based on transmit AS and TPA
algorithm. First, we describe the system modeling and then explain
the receiver blocking problem in dense VANETs environment. Then
we summarize the work in [10] by describing the cross-layer de-
sign using transmit AS and the interference-free symbol detection
 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
Fig. 1. A communication range between vehicles.
in case of simultaneous transmission for V2V communication. Fur-
thermore, we describe the proposed PHY/MAC cross-layer design.
2.1. System modeling
The radio propagation modeling for V2V communication in this
paper is based on the empirical adaptation of free-space propa-
gation for non-ideal channel condition by considering additional
environment-dependent path loss exponent α [20] given by
(4π )2 dα
PL = (1)
λ2
where d is the distance between the transmitter and receiver, α
is the path loss exponent, and λ = c / f is the wavelength where c
is the speed of light and f is the carrier frequency. And then the
average received power P r is calculated as
P r = P t + G t + G r − 10 log10 ( P L ) (2)
where P t is the transmission power, G t and G r are the antenna
gains of the transmitting and receiving antennas respectively, and
P L is the path loss. The communication between vehicles depends
on the transmission power ( P t ) of the transmitter and the receiver
sensitivity (R s ) of the receiver. We define the radio communication
range R as the distance where P r = R s . So, the radio communica-
tion range is calculated as
 1/α
P t G t G r λ2
R= . (3)
(4π )2 R s
Let us refer to Fig. 1, where d1 is the distance between vehicles
V 1 and V 2 , d2 is the distance between V 3 and V 2 , and R is the
distance for a radio communication range of V 2 calculated using
equation (3). When d1 is less than R, it means that the vehicle V 1
is in the communication zone with V 2 ; thus the data transmission
from V 1 is possible but with a probability of transmission error
due to path loss, fading and channel noise. On the contrary, d2 is
greater than R, which means the vehicles V 3 is outside the com-
munication range with V 2 ; Thus, the data transmission will not
occur.
The mobility of the vehicles in our system is considered using
the time-varying nature of the channel that characterized by the
maximum Doppler shift f D [16], which is proportional to the rel-
ative velocity between the transmitter and the receiver. It is given
by f D = v /λ, where v is the relative velocity between vehicles,
and λ is the wavelength. Therefore, the channel is assumed sta-
tionary for a coherence time T c which is inversely proportional to
3
Fig. 2. Receiver blocking problem in dense VANETs.
f D , i.e., T c ≈ 1/ f D . In our simulation model, we define the max-
imum value of the relative velocity between the transmitter and
receiver as v max and calculate the value of T c based on the v max .
Although each vehicle has different speeds, the relative speed be-
tween vehicles never exceeds the value of v max . So the value of T c
of each vehicle will never be smaller than the value of T c based
on the v max .
2.2. Receiver blocking problem in dense VANETs
The IEEE 802.11p MAC protocol standard using CSMA/CA can
cause the hidden node and exposed node problem. The use of
optional request-to-send/clear-to-send (RTS/CTS) mechanism can
solve these problems but also can cause another problem called
receiver blocking problem. To provide more understanding of the
receiver blocking problem in dense VANETs scenario, let us refer
to the vehicle mobility scenario in Fig. 2. When V 1 establishes
communication with V 2 by sending the RTS/CTS acknowledgments,
the MAC protocol forbids the simultaneous transmissions within
the same communication range, in this case from V 3 and V 4 , to
prevent the potential interference and collision. V 3 becomes an
exposed node, and V 4 becomes a hidden node; thus, both vehicles
become defer nodes that will be blocked during the transmission
from V 1 to V 2 . When V 5 wants to make a communication with
V 3 , it will send RTS to V 3 . However, V 3 cannot respond or send
CTS to V 5 because he is still a defer node. As long as V 5 does not
receive the CTS respond from V 3 , it will continuously send RTS
packets to the VANETs network. The same receiver blocking prob-
lem also happens when V 6 also wants to make a communication
to V 4 (the hidden nodes). V 6 will keep sending the RTS, while V 4
cannot respond because it will be blocked during the transmission
from V 1 to V 2 . This problem will get worse in high-density VANET
because many vehicles will be blocked. In Fig. 2, we can see that
V 7 and V 8 will also become defer nodes because both vehicles
know that both V 5 and V 6 sent RTS to V 3 and V 4 , respectively.
Over the past years, some research effort has been made to
mitigate the receiver blocking effect in ad-hoc networks. For exam-
4 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233

ple, Zhai et al. [12] propose a MAC protocol using a dual-channel Table 1
(DUCHA); it uses one channel for data and the other for the pack- An example of the AS list in the 4 × 4 MIMO system.
et’s control. This solution improves the throughput comparing to Antenna Selected Associated
the conventional MAC protocol. However, it did not consider as re- quantity antennas ID’s throughput
alistic, due to hardware limitation and cost. In [13], Feng et al. 4 1, 2, 3, 4 η4
present a multiple receiver transmission (MRT), a fast NAV (Net- 2 2, 4 η2
work Allocation Vector) truncation (FNT), and an adaptive receiver 3 1, 2, 4 η3
1 2 η1
transmission (ART) scheme to mitigate this blocking problem with-
out the adoption of additional control channels. Each node trans-
mits to multiple receivers in the MRT scheme, and the NAV du-
ration in the RTS packet reduces in FNT protocol. They observe a ỹ = Rx + ñ (7)
throughput improvement considering the drawbacks of the MRT
where ỹ = Q H y, x =
x, and ñ = Q H n. Knowing that R is an upper
and FNT schemes. However, their proposal did not describe for use
triangular matrix so that r i , j = 0 if i > j, then the ith element of
in the vehicular environment. Considering MIMO and cross-layer
ỹ is given by
design, Bouraoui et al. in [10] propose a PHY/MAC cross-layer de-
sign based on transmit AS for VANETs V2V communication. This
solution comes with a dedicated MAC protocol that allows vehi- 
Nt
ỹ i = r i ,i xi + r i , j x j + ñi , i = 1, 2, . . . , N t . (8)
cles to transmit data simultaneously without interfering with each
j = i +1
other by implementing ZF linear detection. Hence, the receiver
blocking problem is mitigated because the vehicles can transmit The signal-to-noise ratio (SNR) of each channel is determined by
simultaneously and resulting in a significant improvement of the
the diagonal elements of the matrix R. At the ith antenna; it is
network throughput when compared to the existing IEEE 802.11.p
obtained by
MAC protocol standard. However, this solution still suffers from the
receiver blocking problem in high-density VANETs environment.
r i2,i E b
SNRi = , i = 1, 2, . . . , N t . (9)
2.3. TAS algorithm Nt No
Using Binary Phase-Shift Keying (BPSK) modulation, the symbol er-
We consider spatial multiplexing over MIMO time-varying flat
ror rate (SER) at the ith antenna given by [16] is
fading channel to be implemented in V2V communication. Each
vehicle equipped with multiple antennas with N t transmit anten-
 
nas and N r receive antennas for transmission and reception, where SERi = Q 2SNRi , i = 1, 2, . . . , N t (10)
N t ≤ N r . The received MIMO symbol at the receiver is expressed by
where Q (.) is the complementary error function under Gaussian
y = Hx + n (4) statistic. Considering the packet’s length of L MIMO symbols, the
packet error rate (PER) is calculated as
where y ∈ C N r ×1 is the received signal vector, H ∈ C N r × Nt is the
channel matrix, x ∈ C Nt × 1   L /Nt
 His the transmitted signal vector with 
Nt
transmit power P t = E b x x , and n ∼ CN (0, No INr ) is the cir- PER = 1 − (1 − SERi ) . (11)
cularly symmetric complex Gaussian noise vector. We assume a
i =1
channel state information at the receiver side (CSI-R), which means
that the channel matrix is perfectly known at the receiver, and the Employing GBN with a window size W and SR ARQ retransmission
receiver uses the ZF-VBLAST detection technique with OSIC. The protocol, the throughput is calculated as
MIMO channel defined by the H matrix can be broken down into
several parallel SISO channels using the QR-decomposition method
K × (1 − PER)
η (GBN) = (12)
as [1 + ( W − 1) × PER]
η (SR) = K × (1 − PER) (13)
H = QR
(5)
where η (GBN) and η (SR) are throughput for GBN protocol and SR
where Q ∈ C N r × Nt is a unitary matrix, R ∈ C Nt × Nt is an upper
protocol respectively, K is the number of transmit antennas used
triangular matrix, and
∈ {0, 1}Nt ×Nt is the permutation matrix
for transmission, W is window size for GBN protocol, and PER is
that corresponds to the optimal order of detection
the packet error rate for the transmitted packet.
⎡ ⎤
q11 · · · q1Nt The PHY/MAC cross-layer techniques are implemented by find-
⎢ . .. .. ⎥ ; ing the maximum throughput in the MAC layer based on the SER
Q = ⎣ .. . . ⎦ information obtained from the PHY layer. The receiver looks among
q Nr 1 · · · q Nr Nt all the possible combinations that give the best network through-
⎡ ⎤ put based on the SER information to find the best combination
r11 · · · r1Nt
⎢ . .. .. ⎥ ; of transmitting antennas, t. Then, the receiver feeds back to the
R = ⎣ .. . . ⎦ transmitter a sorted list of the best transmit antennas subsets of
0 · · · r Nt Nt 1, 2, . . . , N t antennas. This list called the AS list, which includes
⎡ ⎤ the number of transmit antennas, the corresponding antenna’s ID,
π11 · · · π1Nt and the associated throughput. Furthermore, it is assumed that
⎢ .. ⎥ .

= ⎣ ... ..
. . ⎦ (6) the receiver sends the list to the transmitter over an error-free
π Nt 1 · · · π Nt Nt feedback channel. Table 1 shows an example of the AS list sends
from receiver to transmitter in the 4 × 4 MIMO system, where
By multiplying the two sides of equation (4) by Q H , the Hermitian η4 > η2 > η3 > η1 . So now, the transmitter can select the best
matrix of Q, we obtain: transmit antenna that has the maximum throughput.
 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
2.4. Interference free symbol detection
Our proposed solution comes with a multi-user interference
cancellation method that allows simultaneous transmissions to
overcome the receiver blocking problem in VANETs V2V commu-
nication. Let us consider the typical VANET scenario depicted in
Fig. 2, where V 1 and V 3 act as the transmit vehicles, while V 2
acts as a receiver vehicle. We assume that V 1 and V 3 have se-
lected K 1 and K 2 transmit antennas, respectively. Thus, when V 1
and V 3 send data simultaneously, the received signal at V 2 is rep-
resented by
 
x1
y = [H1 H2 ] +n
x2
where H1 ∈ C N r × K 1 is the channel matrix between V 1 and V 2 ,
H2 ∈ C N r × K 2 is the channel matrix between V 3 and V 2 , x1 ∈
C K 1 ×1 is the desired symbol vector from V 1 , x2 ∈ C K 2 ×1 is the
interferences symbol vector from V 3 .
We considered a ZF-VBLAST encoding schema with greedy QR
decomposition and OSIC as in [14]. Like equation (5), we can write
the channel matrix as [H1 H2 ] = QR
, where Q ∈ C N r × N r is a uni-
tary matrix, R ∈ C N r ×( K 1 + K 2 ) is an upper triangular matrix, and

∈ R( K 1 + K 2 )×( K 1 + K 2 ) is the permutation matrix corresponding to
the optimal detection order. Now equation (7) can be written as
 
x1
ỹ = R + ñ
x2
furthermore, we can express the ith element of ỹ in the equation
(8) as:
1+K2)
5
we can do the interference nulling and cancelation process from
both V 3 and V 4 as the interference vehicles. Indeed, this mech-
anism will unblock many vehicles and allows them to communi-
cate simultaneously. For this purpose, vehicles negotiate with each
other the number of selected transmit antennas so that the AQR is
satisfied and maximize the network throughput.
2.5. Transmission power adaptation
The AQR condition requires that the number of transmitting
antennas in a receiver detection range  does not exceed the num-
ber of its receiving antennas (i.e., K x ≤ N r ). Hence, AQR limits
the maximum number of transmitters within the receiver detec-
(14) tion range to N r transmitters if each one uses one transmitting
antenna. However, in dense VANETs, it is common for a receiver to
have more than N r nearby transmitters. Therefore, some of them
may remain blocked during an ongoing transmission, despite the
cross-layer using the transmit AS mechanism described in the pre-
vious section.
In what follows, we address this problem in terms of transmit
power adaption. Researchers have investigated power adaptation in
VANETs. Caizzone et al. [11] propose a transmit power algorithm
to solve the problem of nodes isolation in rural zones and also to
keep the connection time longer. The criteria used for power adap-
tion is the neighbors’ density. The algorithm increases or decreases
transmission power so that the number of neighbors of each node
is always within a minimum and maximum threshold. We adapt
(15) this TPA algorithm to our MU-MIMO VANETs using transmit AS so
that the number of simultaneous transmission in the same detec-
tion zone will satisfy the AQR condition.
Let us refer to Fig. 3 for describing the scenario of using the
TPA algorithm in dense VANETs. We deploy VANETs using 4 × 4
( K MIMO systems, which means each vehicle will be equipped with
ỹ i = r i ,i xi + r i , j x j + ñi ,
j = i +1
i = 1, 2, . . . , ( K 1 + K 2 ).
The multiuser interference cancelation is done as follows: first,
the receiver will detect the symbols of x one after the other, start-
ing from the symbol of x K 1 + K 2 . This estimated symbol is symbol-
ized as x̂ and represented as
 
ỹ K 1 + K 2
x̂ K 1 + K 2 = q
r K1+K2,K1+K2
where q (.) is the quantification function. Then, the detected sym-
bol x̂ K 1 + K 2 is removed in the next step to detect x̂( K 1 + K 2 )−1 . This
operation is repeated until all components are detected. So we will
get
⎛ ( K 1 + K 2 ) ⎞ information, which is then used by the PHY layer for the trans-
ỹ i − j = i +1
r i , j xˆj
x̂i = q ⎝ ⎠,
r i ,i
i = (( K 1 + K 2 ) − 1) , (( K 1 + K 2 ) − 2) , . . . , 1
By assuming no error propagation, symbol detection can be
correctly carried out sequentially in K 1 + K 2 steps [15], from
the last symbol ỹ K 1 + K 2 to ỹ 1 . The interference-free symbol de-
tection is only possible if the sum of transmit antennas within
the same communication range K 1 + K 2 does not exceed N r (i.e.,
K 1 + K 2 ≤ N r ). Otherwise, the system is under-determined. This
condition is referred to as the AQR. Then finally, the receiver V 2
retains the desired data from V 1 and ignores the rest.
From Fig. 2, it is also possible for the V 4 , as the hidden node in
the communication range of V 2 , doing the simultaneous transmis-
sion. Using the same ZF-VBLAST architecture as mentioned above,
four transmit antennas and four receive antennas. In Fig. 3(a), the
transmitter T uses a maximum transmit power level. As seen, this
creates an unnecessary broad transmission range and thus a large
(16)
number of interfered vehicles (four vehicles in this case). If all
transmitter (T, I1 , I2 , I3 , and I4 ) want to transmit simultaneously,
the total number of transmit antennas in one of the receiver de-
tection zones may be more than the number of receive antennas;
hence the violation of the AQR condition. However, in Fig. 3(b), T
reduces the transmit power, and therefore the number of the in-
terfered vehicles is also reduced (to one vehicle in this case). Thus
(17) these two vehicles are allowed to communicate simultaneously be-
cause it respects the AQR condition. It is clear that an adapted
transmit power level decreases the number of blocked neighbors
so it can mitigate the receiver blocking problem.
For this purpose, we propose to add a PHY/MAC cross-layer
design where the MAC layer collects the receiver’s neighborhood
mission power adaptation. This adaptation is made progressively
with a step off  to keep the connectivity between the vehicles.
Each transmitter adapts the transmit power P t according to its cor-
responding receiver’s neighborhood density, noted N ( R ). It starts
(18)
with an initial power P ini and then it increases or decreases the
power level by  without exceeding the maximal transmission
power P max , as described by Algorithm 1, where P t (t . T c ) is the
transmit power level during the tth period of T c .
Indeed, if N ( R ) exceeds N r , the transmitter decreases the
power with a step of  to decrease the communication range and
minimize the number of interfering vehicles so that the neighbor-
ing vehicles still allowed to communicate simultaneously because
it respects the AQR condition. If N ( R ) is less than N r , it increases
the transmit power with a step off  to expand the communica-
tion range so it can improve the number of transmission links in
order to improve network throughput. Otherwise, it maintains the
6 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
Fig. 3. Transmission power adaptation (TPA). (a) is the scenario without TPA where there are four interfered vehicles. (b) is a scenario using TPA and only has one interfered
vehicles.
same power level. This process is repeated after each round of the
T c period and until the end of the transmission.
Algorithm 1 Transmit power adaptation.
Input: N(R), , Tc , Pini , Pmax , t;
Output: Pt ;
if t = 1 then
Pt (1.Tc ) = Pini ;
else if N(R) > Nr then
Pt (t.Tc ) = max{Pt ((t − 1).Tc ) − , Pmin };
else if N(R) = Nr then
Pt (t.Tc ) = Pt ((t − 1).Tc );
else
Pt (t.Tc ) = min{Pt ((t − 1).Tc ) + , Pmax };
end if
2.6. PHY/MAC cross-layer design based on TAS and TPA
In this section, we propose a MAC protocol for VANETs adapted
from [9,10], to support the cross-layer design based on transmit AS
and TPA. The use of the AS algorithm proposed above requires an
exchange of antenna list between the vehicles. The RTS/CTS control
packets used in the IEEE 802.11p MAC protocol are not sufficient.
We proposed a new MAC protocol using new control packets, in
addition to RTS/CTS packets, to be implemented in the VANETs
networks. This protocol is broken down into four phases: a phase
to exchange the AS list, a phase for the negotiation of the number
of antennas used by each vehicle between neighboring vehicles, a
phase for simultaneous data transmission, and a phase for trans-
mitting power adaptation. To describe the protocol, let us consider
again the V2V communication scenario presented in Fig. 2, where
V 2 acts as a receiver and detects three other vehicles, i.e., V 1 , V 3
(an exposed node), and V 4 (a hidden node) that want to communi-
cate simultaneously. We focus on the exposed node problem of V 3
and describe how it can transmit simultaneously with V 1 to V 2 ,
and at the same time, it still can communicate with V 5 ; hence,
it will not be a deferred node. The protocol advances through the
following phases:
AS list exchange phase, this phase is also defined as the hand-
shake step, where vehicles share their AS list with the neighbors.
Assuming that V 1 is the first to have the channel access, it sends
a transmission request of request-of-antenna-selection-list (RASL)
to V 2 , and V 2 responds with request-for-antenna-selection-list-
acknowledgment (RASL-ACK) to accept the connection. Then, V 1
broadcasts a training sequence to all vehicles in its transmission
area, to allow them to estimate the channel between them and V 1 .
The estimation of the channel by the neighbors will allow them to
cancel any interference during the communication between V 1 and
V 2 . Based on the channel estimation, V 2 also looks for the best-
transmitting antenna subsets of 1, 2... N t at V 1 in terms of maxi-
mum throughput calculated using equations (12) and (13). Then,
it creates a list of sorted AS with the number of antennas, the
identifiers of the antennas, and the corresponding throughput, as
described in Fig. 3. V 2 retransmits this list via antenna-selection-
list (ASL) to V 1 . V 1 then broadcasts this list to the neighbors via
broadcast-AS-RTS (B-RTS). Note that whenever a transmitter re-
ceives AS list from the neighborhood, it forwards the AS list to its
receiver. This phase is repeated successively until all vehicles ex-
change their AS lists (in this example between V 3 and V 5 ). Fig. 4
details the packet exchange and also the examples of control pack-
ets of this phase.
Negotiation phase, the objective of this phase is to negotiate the
number of transmitting antennas used by each vehicle to allow
simultaneous transmission and ensure the AQR condition. Assum-
ing that V 1 has first access to the channel, it verifies whether it
reaches the AQR. At first, the AQR is not violated, so V 1 sends
broadcast-CTS (B-CTS) to all neighboring vehicles that it can use
all of its antennas, in this case, four antennas. When the channel
is released, V 5 accesses in turn to the channel for the negotia-
tion phase. V 5 knows that AQR is violated, so V 5 selects another
combination from the AS list so that the AQR is satisfied, and it
maximizes the network throughput, based on previous B-RTS. It
chooses two antennas for V 1 (η2 ) and one antenna for V 3 (η1 ).
Once it is done, it broadcasts the selected list to V 3 and the neigh-
bors (V 2 and V 1 ) via B-CTS. Then, V 3 and V 1 update the AS list
accordingly and informs the neighborhood with RTS. Fig. 5 de-
tails the AS list negotiation phase and also the examples of control
packets of this phase.
Data transmissions phase, in this phase, V 1 and V 3 , send a CTS
respectively to V 2 and V 5 and start to send data simultaneously
through the final subsets of selected antennas. Then V 2 and V 5
extract the data using the ZFVBLAST detector as described in sec-
tion 2.4. The lost and damaged packets are retransmitted using the
 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233 7

Fig. 4. The AS list exchange phase and its control packet examples. V 1 and V 3 act as the transmitters, while V 2 and V 5 act as the receiver.

Fig. 5. The negotiation phase and its control packet examples.

SR protocol, and received packets are acknowledged with Acknowl- 3. Performance analysis
edgment (ACK). The V2V between vehicles lasts for each T c round.
When connection is over, it waits until the channel is fully released 3.1. Simulation environment
before going on to the next step.

Transmit the power adaptation phase; at this phase, if V 1 and V 3
continue to transmit data, they account for the number of AS list
at their disposal. This stage will give them the neighbor’s density
of their respective receivers, and depending on this, they choose
the new transmit power level for the next T c period, as mentioned
in section 2.5. The four phases of the proposed MAC protocol are
summarized in Algorithm 2.
We use SUMO (Simulation of Urban MObility), MATLAB (MA-
Trix LABoratory), and TraCI4Matlab [17] as the simulation tools.
We choose SUMO to generate the mobility of the vehicles in
our scenario, MATLAB to modeling our MU-MIMO VANETs sys-
tem, and also to calculate the performance of our proposed cross-
layer design, and TRACI4Matlab as an interface between SUMO
and MATLAB. We design the V2V VANETs scenario where all ve-
8 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233

Algorithm 2 AS with TPA. detection range of R1 . We assume T1 , T2 , T3 , and T4 will transmit

while Vehicles ask for data transmission do simultaneously, and each transmitter will use only one transmit
if distance ≤ communication range then antenna to satisfy the AQR condition. The total potential V2V links
AS list exchange phase; in this scenario are 7. We calculate the network performance in
if AQR is not satisfied then
Negotiation phase; terms of the normalized throughput per potential V2V-links based
else on equations (12) and (13).
repeat In Fig. 7, compared to without the AS algorithm, we can see
Simultaneous data transmission; that the AS algorithm is giving better network throughput per-
until the end of Tc
formance, and using SR protocol gives slightly better performance
Wait until all transmission are finished;
end if than using GBN. Also, the four vehicles scenario has a better
Receiver’s neighbors’ density estimation; throughput than eight vehicles scenario. Indeed, dense VANETs has
Transmit power adaptation; more V2V links which result in frequent violation of the AQR con-
end if dition. Some transmitters are blocked, and the receiver blocking
end while
problem reappears; hence, the network throughput performance
decrease.
Table 2 Now, we want to investigate our proposed cross-layer design
Simulation parameters. by adding the TPA algorithm. In Fig. 8, we provide the throughput
Simulation scenario Manhattan grid performance in the following cases: (i) without AS and TPA, (ii)
Simulation area 100 m × 100 m with AS and without TPA, (iii) with AS and TPA. Note that when
Simulation time T sim 50 s we do not use the TPA algorithm, the transmit power is by default
Number of simulations 30 P ini = P max , and we do not consider the transmit power adaptation
Carrier frequency f 5.9 GHz
phase of the MAC protocol. First, to see the gain brought by the
Number of vehicles 40
Relative velocity v 40 km/h
cross-layer based on AS, let us compare the throughput obtained in
Time coherence Tc 4.6 ms the cases (i) and (ii). In (ii), with AS, at lower E b / N o = 10 dB, the
Network configuration Single hop throughput grows up to more than 1.73 times of that given with-
Channel model Rayleigh flat fading out AS in (i). However, at better of E b / N o = 20 dB, the throughput
Frame length L 180 BPSK symbols
reaches until 6% of gain than that given without AS. In fact, at
ARQ protocol Selective Repeat (SR)
Throughput η K(1 − PER) [8] low of E b / N o , the AS algorithm tends to choose less number of
Symbol duration Ts 8 μs transmit antennas. However, at a high of E b / N o , the optimal sub-
Initial power P ini 30 dBm set of all available transmit antennas can be chosen as long as
Maximal transmission power P max 30 dBm
the AQR is satisfied. As the authors concluded in [9], the use of
Step of power  5 dBm
Receiver sensitivity Rs −70 dBm a cross-layer based on the TAS algorithm improves the throughput
Transmitting antennas gain Gt 0 dBi for each V2V link. Besides, it allows more nearby transmissions,
Receiving antennas gain Gr 0 dBi and therefore, it improves the average network throughput. Now,
Path loss exponent α 3 to see the gain brought by the cross-layer based on transmit power
adaptation, we compare the throughput achieved with (case (iii))
hicles distributed randomly in a predictable area of 100 m × and without TPA (case (ii)), both cases with AS. The throughput
100 m Manhattan grid, N t = N r = 4 antennas (or 4 × 4 MIMO sys- given by AS with TPA, is higher, especially for low noise power
tems) equipped in each vehicle, and the maximum relative velocity level. At of E b / N o = 20 dB, the gain can achieve 2%. However, at
between vehicles is 40 km/h. In this paper, using Monte Carlo of E b / N o = 10 dB (i.e., link quality is poor), we recorded a gain
simulation, we evaluate the performance of the proposed cross- of 26%. Indeed, decreasing the transmit power does not affect the
layer design in terms of normalized network throughput versus throughput per each V2V-link, but it increases the average net-
signal-to-noise ratio (SNR) in terms of E b / N o . To get the network work throughput. In sum, compared to the conventional case, at of
throughput value, we calculate the average throughput of each ve- E b / N o = 10 dB, the performance improvement is about 173% for
hicle over the number of simulations and the number of potential AS without TPA, whereas it is about 244% for AS with TPA. At of
V2V communication links. Then the network throughput value is E b / N o = 20 dB, the performance improvement is 6% for AS with-
normalized by dividing with the maximum value. Unless otherwise out TPA and 9% for AS with TPA.
specified, the numerical values obtained in the rest of this paper In Fig. 9, we evaluate the proposed approach with two dif-
are all based on network parameters listed in Table 2. ferent network density. The throughput improvement with TPA is
remarkable in both scenarios (i.e., 40 and 100 vehicles). The most
3.2. Simulation results significant improvement comes from the dense network where our
proposed solution mitigates the receiver blocking problem; hence,
First, we describe the receiver blocking problem in dense more vehicles are allowed to transmit data simultaneously.
VANETs. Let us refer to the vehicle mobility scenario in Fig. 6 to In Fig. 10, we study the impact of the total number of trans-
describe how the transmitter AS mitigates the receiver blocking mitting and receiving antennas on our proposed design. As illus-
problem, but it still has the limitation in dense VANETs environ- trated, the 8 × 8 MIMO system gives better network throughput
ment. Assuming that each vehicle is using four transmit antennas performance compared to 4 × 4 MIMO systems for both 40 and
and four receive antennas (4 × 4 MIMO system), in scenario as in 100 vehicles scenario. The AQR condition depends on the total
Fig. 6(a), we can see only two vehicles (i.e., T1 and T2 ) detected in number of the receive antennas; thus, the more receive antennas
the detection range of R1 . Using AS architecture with interference- means more transmit vehicles doing the simultaneous transmis-
free symbol detection and considering AQR condition, we assume sion. Hence, in terms of normalized average throughput per V2V
T1 and T2 are doing transmission simultaneously, and each trans- link, the 8 × 8 MIMO system in 100 vehicles scenario gives the
mitter will use two transmit antennas. The total potential V2V best performance, which proves our proposed cross-layer design
communication links in this scenario are 3. In the denser VANETs overcame the receiver blocking problem in dense VANETS. Indeed,
scenario, like the eight vehicles scenario in scenario as in Fig. 6(b), the more antennas mean more vehicles can transmit simultane-
we can see four vehicles (i.e., T1 , T2 , T3 , and T4 ) detected in the ously. However, it also means more time needed to synchronize
 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233 9

Fig. 6. Vehicle mobility scenario. (a) is the four vehicles scenario where R2 will have a receiver blocking problem. (b) is the eight vehicles scenario where the blocking
problem will get worse.

Fig. 7. The throughput is degrading in dense VANETs because there was still a re- Fig. 8. Throughput without antenna selection and power adaptation; with antenna
ceiver blocking problem even though we used the proposed AS selection. selection and without power adaptation; with antenna selection and power adapta-
tion.

between vehicles. We can see that the time needed to do the ne-
gotiation phase in our MAC protocol also increase. It is interesting
to add the delay parameter as one of the performance analysis in
our future research.

4. Conclusion

In this paper, we have proposed a solution to mitigate the re-

ceiver blocking problem in vehicular ad-hoc networks. This solu-
tion is based on TAS and TPA algorithms. This design lets the trans-
mitter choose the best combination of transmitting antennas that
maximize link throughput. Besides, it allows transmitters to send
data simultaneously, with its interference-free symbol detection,
which improves the network performance in terms of throughput.
The simulation results show a network throughput improvement
compared to the conventional case. Moreover, the improvement
is better in dense networks where more vehicles are allowed to
transmit data simultaneously.
Our proposed solution using RTS/CTS schema in order to over-
come blocking problem in the VANETs. This approach is based on
several control packet exchange between transmitter and receiver,
thus slowing the communication process. This solution would in- Fig. 9. Throughput in a network for 40 and 100 vehicles.
10 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
Fig. 10. Throughput comparison for 4 × 4 MIMO system and 8 × 8 MIMO system in
40 and 100 vehicles scenario.
crease the latency or the delay with the respect to the MAC stan-
dard of IEEE 802.11p that only use DATA/ACK approach. The future
work of this research will also consider the delay as the perfor-
mance analysis.
Declaration of competing interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Acknowledgement
This research is supported by the Indonesia Endowment Fund
for Education (LPDP), with grant number: PRJ-3752/LPDP.3/2016,
and the Directorate General of Higher Education (DITJEN DIKTI) of
Indonesia. The funders had no role in study design, data collection,
and analysis, decision to publish, or preparation of the manuscript.
References
[1] H. Hartenstein, K. Laberteaux, VANET Vehicular Applications and InterNetwork-
ing Technologies, John Wiley & Sons, 2009.
[2] A. El-Keyi, T. ElBatt, F. Bai, C. Saraydar, MIMO VANETs: research challenges
and opportunities, in: 2012 International Conference on Computing, Network-
ing and Communications (ICNC), Maui, HI, USA, 2012, pp. 670–676.
[3] S. Moser, L. Behrendt, F. Slomka, MIMO-enabling PHY layer enhancement
for vehicular ad-hoc networks, in: 2015 IEEE Wireless Communications and
Networking Conference Workshops (WCNCW), New Orleans, LA, USA, 2015,
pp. 142–147.
[4] A. Triwinarko, I. Dayoub, P. Wikanta, Using MIMO and cross layer design for
VANETs: a review, in: 2017 International Conference on Signals and Systems
(ICSigSys), Bali, Indonesia, 2017, pp. 13–18.
[5] IEEE Standard for Information Technology-Telecommunications and Information
Exchange Between Systems-Local and Metropolitan Area Networks-Specific Re-
quirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical
Layer (PHY) Specifications, IEEE Std. 802.11, 2007.
[6] IEEE Standard for Information Technology– Local and Metropolitan Area
Networks– Specific Requirements– Part 11: Wireless LAN Medium Access Con-
trol (MAC) and Physical Layer (PHY) Specifications Amendment 6: Wireless
Access in Vehicular Environments, IEEE, 2010.
[7] IEEE Standard for Information Technology–Telecommunications and Informa-
tion Exchange Between Systems Local and Metropolitan Area Networks–
Specific Requirements - Part 11: Wireless LAN Medium Access Control (MAC)
and Physical Layer (PHY) Specifications, IEEE, 2016.
[8] H.A. Abou Saleh, W. Hamouda, Cross-layer based transmit antenna selection for
decision-feedback detection in correlated Ricean MIMO channels, IEEE Trans.
Wirel. Commun. 8 (4) (Apr. 2009) 1677–1682.
[9] W.F. Mao, W. Hamouda, I. Dayoub, Performance of MIMO cross-layer MAC pro-
tocol based on antenna selection in ad hoc networks, Wirel. Commun. Mob.
Comput. 12 (7) (May 2012) 652–660.
[10] B. Bouraoui, M. Zwingelstein-Colin, M. Gharbi, I. Dayoub, R. Attia, Cross-layer
design based transmit antenna selection for vehicular ad-hoc networks, in: M.
Kassab, M. Berbineau, A. Vinel, M. Jonsson, F. Garcia, J. Soler (Eds.), Commu-
nication Technologies for Vehicles, vol. 9066, Springer International Publishing,
Cham, 2015, pp. 24–34.
[11] G. Caizzone, P. Giacomazzi, L. Musumeci, G. Verticale, A power control algo-
rithm with high channel availability for vehicular ad hoc networks, in: IEEE
International Conference on Communications, ICC 2005, 2005, Seoul, South Ko-
rea, vol. 5, 2005, pp. 3171–3176.
[12] H. Zhai, J. Wang, Y. Fang, DUCHA: a new dual-channel MAC protocol for
multihop ad hoc networks, IEEE Trans. Wirel. Commun. 5 (10) (Nov. 2006)
3224–3233.
[13] K.-T. Feng, J.-S. Lin, W.-N. Lei, Design and analysis of adaptive receiver trans-
mission protocols for receiver blocking problem in wireless ad hoc networks,
IEEE Trans. Mob. Comput. 12 (8) (Aug. 2013) 1651–1668.
[14] Y. Jiang, M.K. Varanasi, The effect of ordered detection and antenna selection
on diversity gain of decision feedback detector, in: 2007 IEEE International
Conference on Communications, Glasgow, Scotland, 2007, pp. 5383–5388.
[15] Da-Shan Shiu, M. Kahn, Layered space-time codes for wireless communica-
tions using multiple transmit antennas, in: 1999 IEEE International Confer-
ence on Communications (Cat. No. 99CH36311), Vancouver, BC, Canada, 1999,
pp. 436–440.
[16] J. Proakis, M. Salehi, Digital Communications, 5th edition, McGraw-Hill, 2008.
[17] A.F. Acosta, J.E. Espinosa, J. Espinosa, TraCI4Matlab: enabling the integra-
tion of the SUMO road traffic simulator and Matlab® through a soft-
ware re-engineering process, in: Modeling Mobility With Open Data, 2015,
pp. 155–170.
[18] A. Bazzi, A. Zanella, B.M. Masini, An OFDMA-based MAC protocol for next-
generation VANETs, IEEE Trans. Veh. Technol. 64 (9) (Sept. 2015) 4088–4100.
[19] M. Khurana, C. Ramakhrisna, S.N. Panda, Capacity enhancement using MU-
MIMO in vehicular ad hoc network, Int. J. Appl. Eng. Res. 12 (16) (2017)
5872–5883.
[20] T. Rappaport, Wireless Communications: Principles and Practice, Prentice Hall
PTR, Upper Saddle River, NJ, 2001.
[21] B.E.Y. Belmekki, A. Hamza, B. Escrig, Cooperative vehicular communications
at intersections over Nakagami-m fading channels, Veh. Commun. 19 (2019)
100165.
[22] B.E.Y. Belmekki, A. Hamza, B. Escrig, On the outage probability of cooperative
5G NOMA at intersections, in: 2019 IEEE 89th Vehicular Technology Conference
(VTC2019-Spring), 2019, pp. 1–6.