Professional Documents
Culture Documents
Vehicular Communications
www.elsevier.com/locate/vehcom
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.
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 adopted by Mao et al. in [9] for multi-user mobile ad hoc networks
by the mobility of the vehicle and other channel-caused adverse- (MANETs) environment. To maximizing the network throughput of
ness. Our previous work in [4] also reviews the benefit of employ- the MANETs, they design the cross-layer architecture using ZF-
ing a MIMO system and cross-layer design in VANETs to improve BLAST by selecting the optimum transmitter’s antenna combina-
the overall network performances. By using antenna diversity and tion. The network throughput is calculated using two adopted
spatial multiplexing, we can create a multi-user MIMO (MU-MIMO) automatic-repeat-request (ARQ) retransmission protocol at the data
system for VANETs, which offers a possibility of higher network link layer, i.e., GBN and selective repeat (SR) protocol. The simula-
throughput. tion shows the improvement of network throughput by allowing
Moreover, the cross-layer approach that operates in multiple more than one pair of nodes to communicate simultaneously for
layers can be used to provide a more efficient and robust proto- both ARQ protocols. In [10], Bouraoui et al. extend the work of [9]
col for answering the distinctive characteristics of VANETs such as for VANETs-V2V communication by proposing a PHY/MAC cross-
high mobility of the vehicles, rapid change of topology, and the layer design based on transmit AS with a dedicated MAC protocol.
ability to predict paths. In [19], Khurana et al. compare the per- The proposed solution allows vehicles to transmit data simultane-
formance of MU-MIMO with single-user (SU-MIMO) in realistic ously without interfering with each other as long as a hypothesis
VANET urban and highway road scenarios. They proposed MU- called antenna quantity restriction (AQR), i.e., the total number of
MIMO in VANETs using combined user and transmitter antennas transmit antennas within the receiver’s detection range does not
selection considering linear precoding and Singular Value Decom- exceed the number of antennas of the receiver, is fulfilled. Indeed,
position (SVD). They show that MU-MIMO is a better choice in the receiver blocking problem is mitigated, and the simulation re-
commercial and safety applications of VANETs, as it doubles the sults show the improvement of the average network throughput.
throughput, increase packet delivery rate (PDR) significantly and However, we can observe that the AQR condition is very restrictive
reduces end-to-end delay to nearly half. The difference with our in the context of VANETs with high density. The detection area
proposed solution is as follows: we considered a ZF-VBLAST encod- likely contains a large number of transmitting vehicles.
ing schema with greedy QR decomposition and ordered successive Recent work in [21,22], Belmekki et al., study the cooperative
interference cancellation (OSIC), and we are also implementing the vehicular communication at a road intersection in the presence of
transmit power adaptation (TPA) algorithm. interference. They show that, in line of sight (LOS) scenario, di-
The current MAC standard of IEEE 802.11p dedicated to VANETs rect transmission is better for high densities of vehicles. In this
is using CSMA/CA with DCF that only allow one transmission at paper, we extend the work in [10] to mitigate the receiver block-
a time and prohibits simultaneous transmissions [5–7]. A receiver ing problem in dense VANETs by proposing a PHY/MAC cross-layer
can detect data from all transmitters within its detection range. All design based on TAS and TPA. We consider spatial multiplexing
neighbors are forbidden to transmit data during an ongoing trans- (ZF-VBLAST) over MIMO time-varying flat fading channel to be
mission to avoid potential interference and collision. This mech- implemented in V2V communication. The cross-layer approach is
anism raises a problem called the receiver blocking problem, i.e., implemented to get the maximal network throughput on the MAC
temporary blocking of reception at the nearby vehicles that conse- layer by using the symbol-error-rate (SER) information obtained
quently results in degradation of the average network throughput. from the PHY layer, while the MIMO spatial multiplexing tech-
This MAC protocol underutilizes the capability of the MU-MIMO nique is used to increase the spectral efficiency. The critical idea
VANETs system, in a way that even though the throughput for added to the work in [10], is to add a TPA mechanism in order to
one V2V link improves (due to the use of MIMO system), it still prevent high interference from nearby vehicles. The use of a con-
suffers from the overall network throughput degradation because stant transmit power could lead to unnecessary large transmission
the receiver blocking problem still exists. However, wireless ad-hoc ranges, and consequently to the point of saturation that violate the
communication preferably uses a randomly based access protocol AQR condition. We propose to enhance the cross-layer architecture
like CSMA/CA due to its simplicity for allowing users to access using transmit AS by adding the TPA algorithm so that the number
the medium. Therefore, over the past few years, some research ef- of simultaneous transmission in the same detection zone always
fort has been made to mitigate this blocking effect by proposing satisfies the AQR condition event in dense VANETs.
a cross-layer (PHY/MAC) design using MIMO capability in the PHY We organize the remainder of this paper as follows. Section 2
layer and also enabling multiple clients doing simultaneous trans- will explain the system model, the receiver blocking problem in
mission, while keeping the random access property in the MAC dense VANETs, and also describe the proposed cross-layer design.
layer. Section 3 presents the performance analysis, which explains the
In [18], Bazzi et al. proposed an OFDMA-based MAC protocol simulation environment and also the obtained results. Finally, sec-
for VANETs (OBV) to overcome the heavy traffic condition. The tion 4 concludes this paper. The notations adopted in this paper
simulation results show that OBV outperforms all reference MAC are as follows: matrices are represented as bold uppercase letters,
protocols, even doubling their throughput under heavy-load net- vectors as bold lowercase letters, and complex scalars as italic low-
work conditions. Nevertheless, it is not mentioned in the article ercase letters. (.) T and (.) H are the transposes and the conjugate
whether this solution can be implemented on the MU-MIMO ar- transpose (Hermitian); r i , j is the (i , j )th entry of the upper trian-
chitecture. In [8], Saleh et al. proposed cross-layer TAS for decision- gular matrix R, xi and ni are the ith entry of the vector [x1 x2 ] T
feedback detection (DFD) in correlated Ricean MIMO channels. The and n respectively; Real and Complex sets are denoted as R and
proposed cross-layer architecture is a point-to-point single-user ˜ and (.) re-
C ; symbols multiplied by Q H or are denoted by (.)
MIMO wireless system with N t transmit antennas and N r receive spectively.
antennas, where N t ≤ N r . The transmitter implements some 1 : K
(K ≤ N t ) spatial multiplexer and the receiver employs a DFD us- 2. Cross-layer design based on TAS and TPA
ing the ZF detector to cancel interference and improve the detec-
tion of the transmitted packets. The network throughput is cal- In this section, we provide an overview of the system model
culated using an adopted go-back-n (GBN) protocol at the link and the PHY/MAC cross-layer design based on transmit AS and TPA
level. The simulation result reveals that although the capacity- algorithm. First, we describe the system modeling and then explain
based antenna selection (AS) is more robust to imperfect channel the receiver blocking problem in dense VANETs environment. Then
estimation, the proposed cross-layer AS delivers higher throughput we summarize the work in [10] by describing the cross-layer de-
gains than the capacity-based AS. This transmission scheme is then sign using transmit AS and the interference-free symbol detection
A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233 3
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 5
2.4. Interference free symbol detection we can do the interference nulling and cancelation process from
both V 3 and V 4 as the interference vehicles. Indeed, this mech-
Our proposed solution comes with a multi-user interference anism will unblock many vehicles and allows them to communi-
cancellation method that allows simultaneous transmissions to cate simultaneously. For this purpose, vehicles negotiate with each
overcome the receiver blocking problem in VANETs V2V commu- other the number of selected transmit antennas so that the AQR is
nication. Let us consider the typical VANET scenario depicted in satisfied and maximize the network throughput.
Fig. 2, where V 1 and V 3 act as the transmit vehicles, while V 2
2.5. Transmission power adaptation
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
The AQR condition requires that the number of transmitting
and V 3 send data simultaneously, the received signal at V 2 is rep-
antennas in a receiver detection range does not exceed the num-
resented by
ber of its receiving antennas (i.e., K x ≤ N r ). Hence, AQR limits
the maximum number of transmitters within the receiver detec-
x1
y = [H1 H2 ] +n (14) tion range to N r transmitters if each one uses one transmitting
x2
antenna. However, in dense VANETs, it is common for a receiver to
where H1 ∈ C N r × K 1 is the channel matrix between V 1 and V 2 , have more than N r nearby transmitters. Therefore, some of them
H2 ∈ C N r × K 2 is the channel matrix between V 3 and V 2 , x1 ∈ may remain blocked during an ongoing transmission, despite the
C K 1 ×1 is the desired symbol vector from V 1 , x2 ∈ C K 2 ×1 is the cross-layer using the transmit AS mechanism described in the pre-
interferences symbol vector from V 3 . vious section.
We considered a ZF-VBLAST encoding schema with greedy QR In what follows, we address this problem in terms of transmit
decomposition and OSIC as in [14]. Like equation (5), we can write power adaption. Researchers have investigated power adaptation in
VANETs. Caizzone et al. [11] propose a transmit power algorithm
the channel matrix as [H1 H2 ] = QR, where Q ∈ C N r × N r is a uni-
to solve the problem of nodes isolation in rural zones and also to
tary matrix, R ∈ C N r ×( K 1 + K 2 ) is an upper triangular matrix, and keep the connection time longer. The criteria used for power adap-
∈ R( K 1 + K 2 )×( K 1 + K 2 ) is the permutation matrix corresponding to tion is the neighbors’ density. The algorithm increases or decreases
the optimal detection order. Now equation (7) can be written as transmission power so that the number of neighbors of each node
is always within a minimum and maximum threshold. We adapt
x1
ỹ = R + ñ (15) this TPA algorithm to our MU-MIMO VANETs using transmit AS so
x2 that the number of simultaneous transmission in the same detec-
tion zone will satisfy the AQR condition.
furthermore, we can express the ith element of ỹ in the equation
Let us refer to Fig. 3 for describing the scenario of using the
(8) as:
TPA algorithm in dense VANETs. We deploy VANETs using 4 × 4
1+K2)
( K MIMO systems, which means each vehicle will be equipped with
ỹ i = r i ,i xi + r i , j x j + ñi , four transmit antennas and four receive antennas. In Fig. 3(a), the
j = i +1 transmitter T uses a maximum transmit power level. As seen, this
creates an unnecessary broad transmission range and thus a large
i = 1, 2, . . . , ( K 1 + K 2 ). (16)
number of interfered vehicles (four vehicles in this case). If all
The multiuser interference cancelation is done as follows: first, transmitter (T, I1 , I2 , I3 , and I4 ) want to transmit simultaneously,
the receiver will detect the symbols of x one after the other, start- the total number of transmit antennas in one of the receiver de-
ing from the symbol of x K 1 + K 2 . This estimated symbol is symbol- tection zones may be more than the number of receive antennas;
hence the violation of the AQR condition. However, in Fig. 3(b), T
ized as x̂ and represented as
reduces the transmit power, and therefore the number of the in-
ỹ K 1 + K 2 terfered vehicles is also reduced (to one vehicle in this case). Thus
x̂ K 1 + K 2 = q (17) these two vehicles are allowed to communicate simultaneously be-
r K1+K2,K1+K2
cause it respects the AQR condition. It is clear that an adapted
where q (.) is the quantification function. Then, the detected sym- transmit power level decreases the number of blocked neighbors
bol x̂ K 1 + K 2 is removed in the next step to detect x̂( K 1 + K 2 )−1 . This so it can mitigate the receiver blocking problem.
operation is repeated until all components are detected. So we will For this purpose, we propose to add a PHY/MAC cross-layer
get design where the MAC layer collects the receiver’s neighborhood
⎛ ( 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 mission power adaptation. This adaptation is made progressively
x̂i = q ⎝ ⎠, with a step off to keep the connectivity between the vehicles.
r i ,i Each transmitter adapts the transmit power P t according to its cor-
responding receiver’s neighborhood density, noted N ( R ). It starts
i = (( K 1 + K 2 ) − 1) , (( K 1 + K 2 ) − 2) , . . . , 1 (18)
with an initial power P ini and then it increases or decreases the
By assuming no error propagation, symbol detection can be power level by without exceeding the maximal transmission
correctly carried out sequentially in K 1 + K 2 steps [15], from power P max , as described by Algorithm 1, where P t (t . T c ) is the
the last symbol ỹ K 1 + K 2 to ỹ 1 . The interference-free symbol de- transmit power level during the tth period of T c .
tection is only possible if the sum of transmit antennas within Indeed, if N ( R ) exceeds N r , the transmitter decreases the
the same communication range K 1 + K 2 does not exceed N r (i.e., power with a step of to decrease the communication range and
K 1 + K 2 ≤ N r ). Otherwise, the system is under-determined. This minimize the number of interfering vehicles so that the neighbor-
condition is referred to as the AQR. Then finally, the receiver V 2 ing vehicles still allowed to communicate simultaneously because
retains the desired data from V 1 and ignores the rest. it respects the AQR condition. If N ( R ) is less than N r , it increases
From Fig. 2, it is also possible for the V 4 , as the hidden node in the transmit power with a step off to expand the communica-
the communication range of V 2 , doing the simultaneous transmis- tion range so it can improve the number of transmission links in
sion. Using the same ZF-VBLAST architecture as mentioned above, 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 to V 2 , and V 2 responds with request-for-antenna-selection-list-
T c period and until the end of the transmission. acknowledgment (RASL-ACK) to accept the connection. Then, V 1
broadcasts a training sequence to all vehicles in its transmission
Algorithm 1 Transmit power adaptation. 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
Input: N(R), , Tc , Pini , Pmax , t;
Output: Pt ; cancel any interference during the communication between V 1 and
if t = 1 then V 2 . Based on the channel estimation, V 2 also looks for the best-
Pt (1.Tc ) = Pini ; transmitting antenna subsets of 1, 2... N t at V 1 in terms of maxi-
else if N(R) > Nr then mum throughput calculated using equations (12) and (13). Then,
Pt (t.Tc ) = max{Pt ((t − 1).Tc ) − , Pmin };
else if N(R) = Nr then it creates a list of sorted AS with the number of antennas, the
Pt (t.Tc ) = Pt ((t − 1).Tc ); identifiers of the antennas, and the corresponding throughput, as
else described in Fig. 3. V 2 retransmits this list via antenna-selection-
Pt (t.Tc ) = min{Pt ((t − 1).Tc ) + , Pmax }; list (ASL) to V 1 . V 1 then broadcasts this list to the neighbors via
end if
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-
2.6. PHY/MAC cross-layer design based on TAS and TPA 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-
In this section, we propose a MAC protocol for VANETs adapted ets of this phase.
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 Negotiation phase, the objective of this phase is to negotiate the
exchange of antenna list between the vehicles. The RTS/CTS control number of transmitting antennas used by each vehicle to allow
packets used in the IEEE 802.11p MAC protocol are not sufficient. simultaneous transmission and ensure the AQR condition. Assum-
We proposed a new MAC protocol using new control packets, in ing that V 1 has first access to the channel, it verifies whether it
addition to RTS/CTS packets, to be implemented in the VANETs reaches the AQR. At first, the AQR is not violated, so V 1 sends
networks. This protocol is broken down into four phases: a phase broadcast-CTS (B-CTS) to all neighboring vehicles that it can use
to exchange the AS list, a phase for the negotiation of the number all of its antennas, in this case, four antennas. When the channel
of antennas used by each vehicle between neighboring vehicles, a is released, V 5 accesses in turn to the channel for the negotia-
phase for simultaneous data transmission, and a phase for trans- tion phase. V 5 knows that AQR is violated, so V 5 selects another
mitting power adaptation. To describe the protocol, let us consider combination from the AS list so that the AQR is satisfied, and it
again the V2V communication scenario presented in Fig. 2, where maximizes the network throughput, based on previous B-RTS. It
V 2 acts as a receiver and detects three other vehicles, i.e., V 1 , V 3 chooses two antennas for V 1 (η2 ) and one antenna for V 3 (η1 ).
(an exposed node), and V 4 (a hidden node) that want to communi- Once it is done, it broadcasts the selected list to V 3 and the neigh-
cate simultaneously. We focus on the exposed node problem of V 3 bors (V 2 and V 1 ) via B-CTS. Then, V 3 and V 1 update the AS list
and describe how it can transmit simultaneously with V 1 to V 2 , accordingly and informs the neighborhood with RTS. Fig. 5 de-
and at the same time, it still can communicate with V 5 ; hence, tails the AS list negotiation phase and also the examples of control
it will not be a deferred node. The protocol advances through the packets of this phase.
following phases:
Data transmissions phase, in this phase, V 1 and V 3 , send a CTS
AS list exchange phase, this phase is also defined as the hand- respectively to V 2 and V 5 and start to send data simultaneously
shake step, where vehicles share their AS list with the neighbors. through the final subsets of selected antennas. Then V 2 and V 5
Assuming that V 1 is the first to have the channel access, it sends extract the data using the ZFVBLAST detector as described in sec-
a transmission request of request-of-antenna-selection-list (RASL) 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.
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 We use SUMO (Simulation of Urban MObility), MATLAB (MA-
continue to transmit data, they account for the number of AS list Trix LABoratory), and TraCI4Matlab [17] as the simulation tools.
at their disposal. This stage will give them the neighbor’s density We choose SUMO to generate the mobility of the vehicles in
of their respective receivers, and depending on this, they choose our scenario, MATLAB to modeling our MU-MIMO VANETs sys-
the new transmit power level for the next T c period, as mentioned tem, and also to calculate the performance of our proposed cross-
in section 2.5. The four phases of the proposed MAC protocol are layer design, and TRACI4Matlab as an interface between SUMO
summarized in Algorithm 2. and MATLAB. We design the V2V VANETs scenario where all ve-
8 A. Triwinarko et al. / Vehicular Communications 24 (2020) 100233
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
[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-
Fig. 10. Throughput comparison for 4 × 4 MIMO system and 8 × 8 MIMO system in
nication Technologies for Vehicles, vol. 9066, Springer International Publishing,
40 and 100 vehicles scenario.
Cham, 2015, pp. 24–34.
[11] G. Caizzone, P. Giacomazzi, L. Musumeci, G. Verticale, A power control algo-
crease the latency or the delay with the respect to the MAC stan- rithm with high channel availability for vehicular ad hoc networks, in: IEEE
dard of IEEE 802.11p that only use DATA/ACK approach. The future International Conference on Communications, ICC 2005, 2005, Seoul, South Ko-
rea, vol. 5, 2005, pp. 3171–3176.
work of this research will also consider the delay as the perfor-
[12] H. Zhai, J. Wang, Y. Fang, DUCHA: a new dual-channel MAC protocol for
mance analysis. multihop ad hoc networks, IEEE Trans. Wirel. Commun. 5 (10) (Nov. 2006)
3224–3233.
Declaration of competing interest [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.
The authors declare that they have no known competing finan- [14] Y. Jiang, M.K. Varanasi, The effect of ordered detection and antenna selection
cial interests or personal relationships that could have appeared to on diversity gain of decision feedback detector, in: 2007 IEEE International
influence the work reported in this paper. 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-
Acknowledgement ence on Communications (Cat. No. 99CH36311), Vancouver, BC, Canada, 1999,
pp. 436–440.
This research is supported by the Indonesia Endowment Fund [16] J. Proakis, M. Salehi, Digital Communications, 5th edition, McGraw-Hill, 2008.
for Education (LPDP), with grant number: PRJ-3752/LPDP.3/2016, [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-
and the Directorate General of Higher Education (DITJEN DIKTI) of
ware re-engineering process, in: Modeling Mobility With Open Data, 2015,
Indonesia. The funders had no role in study design, data collection, pp. 155–170.
and analysis, decision to publish, or preparation of the manuscript. [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-
References
MIMO in vehicular ad hoc network, Int. J. Appl. Eng. Res. 12 (16) (2017)
5872–5883.
[1] H. Hartenstein, K. Laberteaux, VANET Vehicular Applications and InterNetwork- [20] T. Rappaport, Wireless Communications: Principles and Practice, Prentice Hall
ing Technologies, John Wiley & Sons, 2009. PTR, Upper Saddle River, NJ, 2001.
[2] A. El-Keyi, T. ElBatt, F. Bai, C. Saraydar, MIMO VANETs: research challenges [21] B.E.Y. Belmekki, A. Hamza, B. Escrig, Cooperative vehicular communications
and opportunities, in: 2012 International Conference on Computing, Network- at intersections over Nakagami-m fading channels, Veh. Commun. 19 (2019)
ing and Communications (ICNC), Maui, HI, USA, 2012, pp. 670–676. 100165.
[3] S. Moser, L. Behrendt, F. Slomka, MIMO-enabling PHY layer enhancement [22] B.E.Y. Belmekki, A. Hamza, B. Escrig, On the outage probability of cooperative
for vehicular ad-hoc networks, in: 2015 IEEE Wireless Communications and 5G NOMA at intersections, in: 2019 IEEE 89th Vehicular Technology Conference
Networking Conference Workshops (WCNCW), New Orleans, LA, USA, 2015, (VTC2019-Spring), 2019, pp. 1–6.
pp. 142–147.