Content uploaded by Cao Dun
Author content
All content in this area was uploaded by Cao Dun on Nov 05, 2018
Content may be subject to copyright.
IEICE TRANS. FUNDAMENTALS, VOL.E99–A, NO.11 NOVEMBER 2016
1
PAPER
Exponent-Based Partitioning Broadcast Protocol for Emergency
Message Dissemination in Vehicular Networks
Dun CAO†∗a) , Zhengbao LEI†, Baofeng JI††,Nonmembers,and Chunguo LI††† ,Member
SUMMARY We propose an exponent-based partitioning broadcast pro-
tocol (EPBP) to promise the prompt dissemination of emergency message
(EM) in vehicular networks. EPBP divides the communication range into
segments with different widths iteratively. The width varies corresponding
to the exponential curve. The design makes the farther no-empty segment
thinner, as a result of which the collision rate of candidates’ contention for
the relay node decreases and the one-hop message progress increases effi-
ciently. In addition, we adjust the interval of back-offtimers to avoid the
spurious forwarding problem, and develop more accurate analytical mod-
els for the performance. Our simulation verifies these models and show a
significant increase of EPBP compared with the state-of-the-art protocols.
EM dissemination speed can be improved as 55.94% faster in dense vehicle
networks, and packet delivery ratio has risen to higher than 99.99%.
key words: vehicular network, broadcast protocol, emergency message
(EM), exponent-based partitioning
1. Introduction
The development in occupant safety and the expectation of
new cars equipped with connected car solutions have led to
a renewed interest in the vehicular network. The vehicular
network can increase occupant safety by sharing the traffic
information between vehicles through the dissemination of
emergency message (EM) [1]. However, different from the
traditional Mobile Ad Hoc Network, the vehicular network
has these characteristics: the frequent network-topology
change induced by the high mobility, a wide range of node
densities, and the strict delay requirement of safety applica-
tions [2]–[4]. So it is challenging to deliver EM in the ve-
hicular network timely and reliably [5]–[8]. Currently, the
topic of reliable, robust, especially low-latency multi-hop
broadcast protocols in vehicular networks has been widely
researched [9]–[13]. In the low-latency multi-hop broadcast
protocol, it is critical to choose the relay node for next hop.
One of the key approaches for the relay-node selection
Manuscript received February 26, 2016.
Manuscript revised May 30, 2016.
†The authors are with the School of Automotive and Mechani-
cal Engineering, Changsha University of Science and Technology,
Changsha 410114, China.
††The author is with the College of Information Engineering,
Henan University of Science and Technology, Luoyang 471000,
China.
†††The author is with the School of Information Science and En-
gineering, Southeast University, Nanjing 210096, China.
∗Presently, with Hunan Provincial Key Laboratory of Intelli-
gent Processing of Big Data on Transportation and the School of
Computer and Communication Engineering, Changsha University
of Science and Technology, Changsha 410114, China.
a) E-mail: caodun@csust.edu.cn
DOI: 10.1587/transfun.E99.A.1
is the distance-based approach, which selects the farthest
node as the relay node [10]–[13]. The farthest node offers
a maximum one-hop coverage, resulting in the reductions
of the multi-hop latency. Moreover, because of the selec-
tion of a single relay node, the broadcast storm [14] can be
avoided. So the distance-based approach achieves more fa-
vor over non-distance-based ones [15].
The urban multi-hop broadcast protocol (UMB) [10]
is a typical example of the distance-based approach. UMB
achieves a low delay with the selection of the farthest node
by using the black burst [16] to jam the channel. The length
of the black burst is proportional to the distance from the
sender. Thus, the latency is exaggerated in high node den-
sities due to the farthest node close to the border. The po-
sition based multi-hop broadcast protocol (PMBP) [11] out-
performs UMB, because the black burst of the farthest nodes
in PMBP is the shortest. However, in the low node den-
sity, it is more likely that the relay node exists in the area
near the sender, which leads to a high waiting time. The
binary-partition-assisted broadcast protocol (BPAB) [12]
and the trinary partitioned black-burst-based broadcast pro-
tocol (3P3B) [13] repeatedly partition the communication
range into a certain number of segments. And the farthest
no-empty segment is chosen within a constant duration. So
BPAB and 3P3B show an improved performance when com-
pared with PMBP and UMB. However, in BPAB and 3P3B,
the segment widths are equal in a iteration. Under heavy
traffic conditions, there are many nodes existing in the final
selected segment, resulting in more collisions of candidates’
contention. Thus, the performance in terms of the multi-
hop delay would deteriorate seriously. Moreover, in high
vehicle densities, the approximations of analysis models in
[12],[13] can result in the deviation from the true perfor-
mances.
In this paper, we propose an exponent-based partition-
ing broadcast protocol (EPBP) to quicken the EM dissemi-
nation. Similar to BPAB and 3P3B, EPBP picks a relay node
in the farthest no-empty segment by dividing the communi-
cation range into segments iteratively. The segment number
in each partition and the iteration number are fixed. Dif-
fering from BPAB and 3P3B, the segment width is differ-
ent, which decreases in the EM broadcasting direction cor-
responding to the exponential curve. The design makes the
farthest no-empty segment is the thinnest one. Therefore,
EPEB attains a smaller contention latency at cost of the
same partition latency with BPAB and 3P3B. Meanwhile,
the one-hop message progress increases. Finally, the one-
Copyright c
2016 The Institute of Electronics, Information and Communication Engineers
Please confirm each of your
IEICE memberships.
2
IEICE TRANS. FUNDAMENTALS, VOL.E99–A, NO.11 NOVEMBER 2016
hop delay and the hop count are lowered in an effort to re-
duce the multi-hop delay. Additionally, the interval of back-
offtimers in CTB contention is adjusted to avoid the spuri-
ous forwarding problem [17], which has come to the atten-
tion of researchers recently. Finally, more accurate models
for the performance analysis of EPBP, compared with those
in [12],[13], are developed.
The rest of the paper is organized as follows. In Sect. 2,
the exponent-based partitioning broadcast protocol is devel-
oped. Section 3 analyzes the performances of the proposed
protocol. The analytical models are validated and the perfor-
mances are evaluated in Sect. 4. The final section concludes
the paper.
2. Exponent-Based Partitioning Broadcast Protocol
(EPBP)
We consider a vehicular network without infrastructures,
which is more flexible and economical than that with in-
frastructures. Vehicles move on a straight way, and directly
communicate among themselves via IEEE 802.11 network.
Moreover, it is assumed that GPS is available, so that ve-
hicles can synchronize and know their own positions [10]–
[13]. It is further assumed that the communication range of
any vehicle is fixed as R. In the considered scenario, the
network along a straight way is a typical vehicular network,
referred as 1-D network [9]. And the 1-D network can pro-
vide insight to complex 2-D networks, e.g. the network at
the intersection.
As shown in Fig. 1, EPBP consists of four phases. 1)
Mini-distribute inter-frame space (mini-DIFS) [9]: when
multiple EMs are broadcasted at the same time, each ve-
hicle sending EM will randomly pick the i-th mini-slot from
the [0, w −1] interval to access to the channel. The mini-
slot is a fraction of DIFS. So the mini-DIFS can mitigate the
synchronization of multiple EMs and reduce waiting time.
After accessing the channel, the sender transmits an request-
to-broadcast (RTB) packet and waits for the corresponding
clear-to-broadcast(CTB) packet. 2) Exponent-based parti-
tioning: the exponent-based partitioning method makes the
final selected segment thinner. So in 3) the CTB contention
phase, only one relay node is selected with fewer collisions.
The relay node responds CTB and completes the RTB/CTB
handshake to avoid the hidden node problem. Finally, 4) the
data transmission for next hop follows. The whole proce-
dure is repeated until the EM arrives at the destination or
expires. More details about the exponent-based partition-
ing and the CTB-contention are described in the following
subsections.
1) Exponent-based partitioning: as illustrated in Fig. 1,
after receiving the RTB packet, all receivers in the commu-
nication range broadcast the black burst BAat the same time
of t1with the support from GPS. Then the sender senses the
existence of the potential relay node(s).
The exponent-based partitioning method divides the
communication range into Npart segments for Niter iterations.
In each of the first (Niter −1) iterations, the farthest nonempty
segment is selected as the input of next iteration. And in the
Niter-th iteration, the farthest nonempty segment is selected
as the final segment.
To make the final segment is the thinnest one, we as-
sume that the segment width ∆xin each partition varies di-
rectly with the distance xfrom the border. ∆xis given as
∆x∝x(1)
Therefore, (1) can be expressed as
dx
dy=Gx (2)
where Gis a positive coefficient, and yis an independent
variable. The solution to (2) can be derived as
ln x=Gy+C(3)
where Cis a constant. We assume that the variables of x
and yare normalized. Thus, (3) should meet the boundary
condition that when x=0 or 1, y=0 or 1. To satisfy the
condition of (x, y)=(1,1), (3) can be transformed as
ln x=Gy−G(4)
Meanwhile, to meet the condition of (x, y)=(0,0), (4) can
be revised as
y=ln (1+Ax)
ln (1+A)(5)
where A,A=(eG−1) >0, is a compression coefficient.
The bigger is A, the faster the width of segments decreases
in the EM broadcasting direction. Therefore, in high vehicle
densities, the selection of a big Acan result in a low CTB-
contention latency and a high one-hop coverage. However,
in low vehicle densities, a big Acan result in a high waiting
latency in the partition phase and a low one-hop coverage.
By the uniform partition of ywith Npart (i.e. yi=
i/Npart,i=0,1,· · · ,Npart ), we obtain the widths of Npart seg-
ments ∆xi,(i=1,2,· · · ,Npart) in each iteration. ∆xican
derived by
∆xi=xi−xi−1
=1
A(1+A)
i−1
Npart (1+A)
1
Npart −1(6)
Shown from (6), ∆xiincreases with the growth of ibe-
cause (1 +A)>1. This implies that the segment width
decreases in the EM broadcasting direction corresponding
to a exponential curve.
The process of Fig. 2 exemplifies the operation of the
exponent-based partitioning with Npart =3 and Niter =3. To
make it easier for each node to calculate the segment width,
we assume A=2.3. In this case, the widths of the 1st, 2nd
and 3rd segments in each partition are about 1/4, 1/4 and
1/2 of their partition range, respectively.
In an iteration, all potential relay nodes determine
which segment they belong to with the position information
of the sender from RTB and own from GPS, then broadcast
a black burst B only in the corresponding time slot.
CAO et al.: EPBP FOR EMERGENCY MESSAGE DISSEMINATION IN VEHICULAR NETWORKS
3
Fig. 1 Procedure of EPBP.
Fig. 2 Example of the exponent-based partitioning with three partitions and three iterations.
In the first iteration of the example, there are nodes in
the 1st segment. Thus, these nodes will broadcast a B si-
multaneously in the first time slot. Then the nodes in the
other segments sense the B, and leave the partitioning pro-
cess. Finally, the 1st segment is made as the input segment
for the second iteration. The process will continue in the
next Niter −1 iterations.
In the second iteration, because there is no node in the
1st segment, the 2nd segment is selected with the width of
R/16 and the cost of two time-slots.
In the last iteration, no B is broadcasted during both
first two time slots. Thus, the nodes in the 3rd segment de-
duce they are the potential relay nodes, and don’t broadcast
any B. At the end, the method yields the final segment of
width R/32 with total five time-slots.
The thinner the final segment is, the larger the progress
range per hop covers and the more rarely the potential re-
lay nodes contend. When Npart =3 and Niter =2, EPBP
can yield the thinnest final segment of width R
2Npart−1Niter =
R
/
16, whereas R
NNiter
part =R
/
9for 3P3B.
Only the nodes in the final segment participate in the
CTB-contention as the candidates for the relay node.
2) CTB-contention: The candidates randomly choose
their back-offperiods from the contention windows. The
choice of a back-offtimer conforms to the the carrier sense
multiple access with collision avoidance (CSMA/CA) pol-
icy of IEEE 802.11p. The candidate, whose back-offtimer
expires firstly, broadcasts a CTB packet. However, a CTB
collision will happen when more than one candidate chooses
the same back-offperiod. In such case, after a certain period,
the sender recalls the process of CTB-contention by broad-
casting a black burst BHfor only half of a time slot, shown in
Fig. 1. The certain period has the duration of the maximum
back-offtimer plus the transmission time, TCTB , of CTB.
Upon the detection of BH, the candidates’ contend with each
other by randomly selecting a new back-offvalue. Before
the contentions re-attempt over the preassigned threshold
Nre con, the process is repeated until the CTB is transmitted
successfully.
In the CTB-contention phase, the spurious forwarding
[17] will occur if the back-offtimer of the second candidate
expires before the CTB-receiving from the first one selected
for forwarding. To avoid the spurious forwarding problem,
the interval of any two back-offperiods must be larger than
the time required for the lower protocol layers to complete
the delivery of CTB. Therefore, the interval of two adjacent
back-offtimers, Tinter, must satisfy the following inequality.
Tinter >TCTB (7)
On the other hand, with the increase of the back-offpe-
riod, the contention latency will rise. Thus, in our proposal
protocol, Tinter =1
2Tslot with the choice of the system pa-
rameters from IEEE802.11p, where Tslot is the duration of
one time slot.
If receiving CTB from the relay node, the sender broad-
casts EM after a standard inter-frame space (SIFS) period.
Moreover, the sender confirms the relay success of next hop
by receiving the same EM from the relay node. If not re-
ceiving the EM after a certain period, the sender will restart
the procedure of EPBP until the repetition number over the
preassigned threshold Nre EPBP to ensure the reliability.
3. Analytical Models
To verify the effectiveness for efficient dissemination, we
analyze the EPBP performance in terms of the message dis-
semination speed. In our analysis, wireless channel error is
not considered. So the message dissemination speed is equal
to that in one hop, which is defined as
4
IEICE TRANS. FUNDAMENTALS, VOL.E99–A, NO.11 NOVEMBER 2016
v=Rβ
Td
(8)
where βis the one-hop message progress, and Tdis the one
hop delay.
As shown in Fig. 1, Tdconsists of the initial latency
Tinit, the partitioning latency Tpart , the contention latency
Tcont and the data transmission latency Tdata, (i.e. the EM
transmission time). Tdis given as
Td=Tinit +Tpart +Tcont +Tdata (9)
Because of the difference in the segment width de-
sign, the proposed exponent-based partitioning method can
change the performances of Tpart,Tcont , and β, compared
with those in [12],[13]. So in the following subsections,
we only develop the analytical models for the three perfor-
mances.
3.1 Partitioning Latency Tpart
The assumption is widely accepted that the number of vehi-
cles, X, in a segment follows Poisson distribution [10]–[13].
The probability of Xis given as
Pµ(X=n)=e−µµn
n!(10)
where µ,µ=λWseg, is the average vehicle number in a
segment with a width of Wseg.Wseg is the relative width
normalized by R, and λis the average vehicle number in the
communication range.
As shown in Fig. 2, in the jth iteration, there are
Nseg(j)=Nj
part possible segments in the whole communi-
cation range. The kth segment is selected when there is no
vehicle in other k −1segments in the EM broadcasting di-
rection and there is at least one vehicle in the kth segment.
(The definitions of the parameter kand the concept of other
segments in the EM broadcasting direction are illustrated in
Fig. 2.) The probability of the selection of the kth segment
in the jth iteration can be given as
pseg sel(j,k)=Pµseg bro(j,k)(X=0) 1−Pµseg(j,k)(X=0)
=e−µseg bro(j,k)(1 −e−µseg(j,k)) (11)
where µseg bro(j,k) and µseg(j,k) are the average vehicle
numbers in other segments in the EM broadcasting direc-
tion and in the kth segment, respectively.
Therefore, the average duration spent during the parti-
tioning phase can be obtained as
Tpart =
Niter
j=1
Nseg(j)
k=1Np slot(k)pseg sel (j,k)
+1
Tslot (12)
where Np slot(k) is the number of time slots spent when the
kth segment is selected. Np slot (k) can be given as
Np slot(k)=kmod Npart if (kmod Npart )<Npart
Npart −1 if (kmod Npart)=Npart
(13)
3.2 Contention Latency Tcont
There are three possible cases of the final segment during
the contention period.
•Idle: No CTB is broadcasted, and the channel is quiet
for the whole time slot.
•Success: Only one vehicle’s back-offtimer expires
firstly, and a CTB packet is broadcasted successfully.
•Collision: More than one vehicle’s back-offtimer ex-
pires firstly, and broadcast CTB packets simultane-
ously, resulting in a collision.
Thus, the durations spent in each case can be derived
as
Tidle =Tslot (14)
Tsuc =TCTB con +TCTB (15)
Tcol =TCTB col +TCTB +1
2Tslot (16)
where TCTB con is the average back-offtime in each CTB-
contention, equal to half of (CW−1)Tinter since the prob-
ability of the selection of each back-offtimer is equal to
p=1/CW.CWis the number of available back-offtimers
and (CW−1)Tinter represents the maximum back-offdura-
tion. TCTB col is the collision time, equal to (CW−1)Tinter.
And 1
2Tslot represents the duration of a BH.
According to the binomial probability, the probability
of mvehicles selecting the same back-offtimer in nvehicles
is given as
P(mselection in nvehicles) =n
mpm(1 −p)n−m(17)
Therefore, the probabilities of one back-offtimer being se-
lected by only one node or being selected by non-node are
given as
pby one =
∞
n=0
P(1 selection in nvehicles)Pµk(X=n)
=
∞
n=0
np(1 −p)n−1e−µkµkn
n!
pby none =
∞
n=0
P(0 selection in nvehicles)Pµk(X=n)
=
∞
n=0
(1 −p)ne−µkµkn
n!
where µk=µseg(Niter ,k) for the simple presentation.
According to ∞
n=0(xn/n!) =ex,pby one and pby none
can be computed as
pby one =µkpe−µk
∞
n=0(1 −p)µkn−1
(n−1)! =µkpe−µkp(18)
pby none =e−µk
∞
n=0(1 −p)µkn
n!=e−µkp(19)
CAO et al.: EPBP FOR EMERGENCY MESSAGE DISSEMINATION IN VEHICULAR NETWORKS
5
The qth back-offtimer is selected successfully when
the forefront q−1 back-offtimers aren’t selected by any
node and the qth back-offtimer is selected by only one node.
The probability of this case is given as
pone timer selected suc(q)=pq−1
by none pby one
=µkpe−µkpq (20)
The success case in the kth possible final segment
happens when any back-offtimer is selected successfully.
Therefore, the probability of the success case is the sum of
the probabilities of the success selection of every back-off
timer. The probabilities of the three cases can be determined
as follows:
pidle(k)=pcw
by none =e−µk(21)
psuc(k)=
cw
q=1
pone timer selected suc(q)
=µkpe−µkpe−µk−1
e−µkp−1(22)
pcol(k)=1−psuc (k)−pidle(k)
=1−e−µk−µkpe−µkpe−µk−1
e−µkp−1(23)
According to the geometric distribution, the expected
value of the fail attempt number f(k) before the successful
CTB transmission can be derived by
f(k)=1−psuc(k)
psuc(k)(24)
Note that the contention happens in the nonempty seg-
ment, so the idle case doesn’t exist. Therefore, f(k) can be
computed as
f(k)=pcol(k)
psuc(k)=1−e−µkp
µkpe−µkp−1 (25)
When the kth segment is the final segment, the average
contention latency can be obtained as
Tcont seg(k)=f(k)Tcol +Tsuc (26)
Finally, the average contention latency is derived by
Tcont =
Nseg(Niter )
k=1
pseg sel(Niter ,k)Tcont seg(k) (27)
3.3 One-Hop Message Progress β
One-hop message progress βis the average distance relative
to Rin one hop, which can be calculated as
β=
Nseg(Niter )
k=1
Mkpseg sel(Niter ,k) (28)
where Mkis the average message progress if the kth segment
is the final segment. Because the final segment has a small
width, it is fair to assume that the vehicles in the final seg-
ment are distributed uniformly. Moreover, every candidate
is selected randomly with the same probability, thus Mkcan
be given by
Mk=Wseg(k)/2+
Nseg(Niter )
n=k+1
Wseg(n) (29)
where Wseg(♡) is the the width of the ♡th segment (♡repre-
sents kor n).
4. Validations of Analytical Model and Evaluations of
Performance
The scenarios are composed of VANETs in a straight 4-lane
dual driveway. The rectangular area in the communication
range, as shown in Fig. 2, is called unit area. The major
simulation parameters are given in Table 1. The arrival rate
of EMs is set to 2 EMs/s. To evaluate EPBP effective for
a large range of vehicle densities, the density of vehicles
λvaries from 20 to 160 vehicles/unit area. The simulation
area is divided into a number of thinnest segments of width
Wseg min =R/(2Npart −1)Niter . And the vehicle number of each
thinnest segment is selected randomly from the Poisson dis-
tribution with λWseg min/R. In each thinnest segment, the
vehicle is distributed uniformly. The maximum speed vmax
of vehicles complies with the rule related to the safe inter-
vehicle distance. Each vehicle chooses a random speed fol-
lowing a uniform distribution in [1
2vmax, vmax ] at the begin-
ning of the simulation, and keeps the chosen speed during
the simulation. Vehicle move in the different lanes accord-
ing their speed. The rule is that the vehicles with the speed
above 3
4vmax move on the fast lane and the vehicles with the
speed below 3
4vmax move on the slow lane. Lane change
and overtaking are not modeled for vehicle movement. The
preassigned thresholds for the contention re-attempt Nre con
and the EPBP repetition Nre EPBP are set to 20 and 1, respec-
tively.
By the ergodic theorem [18], the stationary probabil-
ity distribution is approximated by the empirical measures
of the random states of the Monte Carlo sampler. So in
the following simulation, we performed Monte Carlo sim-
ulation [19],[20] with 100000 realizations in the MATLAB
link level simulation.
Table 1 Major simulation parameters.
Parameters Default Values
Bit Rate 18Mbps
Communication Range 900m
EM Packet Size 500Bytes
RTB Packet Size 20 Bytes
CTB Packet Size 14 Bytes
Slot Time 13 µs
DIFS 58 µs
SIFS 32 µs
Transceiver’s Switching Time 1 µs
6
IEICE TRANS. FUNDAMENTALS, VOL.E99–A, NO.11 NOVEMBER 2016
Fig. 3 Validation of the analytical models.
4.1 Validation of Analytical Models
The comparisons between the analytical and simulation re-
sults in three scenarios are shown in Fig. 3. One parameter
of Niter,Npart and Cwis varied while the others are fixed. As
shown, the simulation results of all vehicle densities in the
three scenarios closely match the analytical results. There-
fore, the analytical models are valid.
In addition, Fig. 3 shows that the proposed EPBP
presents a stable dissemination speed even at the high ve-
hicle density. It benefits from the design of the thinnest fi-
nal segment, which reduces the collision rate of the CTB-
contention. And a little inferior performance of EPBP in
low traffic scenarios can’t be avoided because of no avail-
able vehicle near the border.
Figure 3 also shows the effect of Niter,Npart and Cwon
EPBP’s performance in terms of the dissemination speed.
It is shown that with the increase of Niter or Npart, the dis-
semination speed decreases due to the increase of the parti-
tion latency. And the speed increases with the decrease of
Cw, owing to the decrease of the average back-offduration
in a CTB-contention. However, in the case of small Niter,
Npart and Cw, the frequent CTB-collision at the high vehi-
cle density will result in the decline of the dissemination
speed. As an example, it is shown in Fig. 3(b) that the curve
corresponding to the case of (Niter,Npart,Cw)=(3,3,4)
descends when the vehicle density higher than 120 vehi-
cle/unite area. In the aforementioned scenarios, EPBP with
Niter,Npart ,Cw=(2,4,4) achieves the fastest average mes-
sage dissemination speed, which is selected in the following
performance evaluations.
4.2 Performance Evaluations of EPBP
The performances of EPBP are compared against several
state-of-the-art broadcast protocols, 3P3B [13], BPAB [12]
and PMBP [11]. To evaluate the impact of the exponent-
based partitioning method, mini-DIFS is employed in BPAB
and PMBP. And to alleviate the collision in CTB-contention
at high vehicle densities, the exponential back-offtimer [16]
is applied in 3P3B, BPAB and PMBP.
Figure 4 compares the performance in terms of the
Fig. 4 Performance comparison in terms of the CTB-contention latency.
contention latency. As shown, among all protocols, EPBP
achieves the lowest contention latency, which is similar to
the performance of PMBP. The greater performances ben-
efit from the endeavor of EPBP to narrow the farthest seg-
ment, and the endeavor of PMBP to increase the iteration
number, respectively. However, for the same reason, PMBP
presents a high partition latency, resulting in the highest one
hop delay as shown in Fig. 5. Moreover, Fig. 4 shows that
the contention latency of 3P3B and BPAB deteriorate se-
riously with the increasing of the vehicle density, whereas
EPBP presents a little growth. The deterioration of 3P3B
and BPAB owes to the existence of many candidates in the
final segment. For example, at the vehicle density of 120
vehicles/unit area, the average numbers of vehicles in the
final segment of BPAB and 3P3B are 15 and 14 respec-
tively, whereas only 2 vehicles in EPBP. As we know, with
the exponential back-offtimer applied, the collision can be
lessened when the width of back-offtimers doubles after a
collision. However, the average back-offdelay in each con-
tention rises with the growth of the back-offtimer width.
So, at the vehicle density of 160 vehicles/unit area, the con-
tention latency of EPBP is only 9.50% and 10.84% of those
of BPAB and 3P3B, respectively.
Figure 5 shows the comparisons of the message
progress (corresponding to the curves included in the top
CAO et al.: EPBP FOR EMERGENCY MESSAGE DISSEMINATION IN VEHICULAR NETWORKS
7
Fig. 5 Performance comparison in terms of one hop delay and message
progress.
ellipse), and the comparisons of the one hop delay (corre-
sponding to the curves included in the bottom ellipse). As
demonstrated in Fig. 5, EPBP achieves the lowest and most
stable one-hop delay among all the examined protocols. The
improvement is attributable to the lowest contention latency
and the limited partition latency. On the other hand, EPBP
achieves the highest message progress. At the same time,
its message progress is only slightly higher than that of the
second best performing protocol PMBP, with a marginal dif-
ference of no greater than 0.31%. The performance gap be-
tween EPBP and PMBP for one hop delay is well above
29%. Moreover, EPBP significantly outperforms BPAB and
3P3B in both performance measures in all the examined
cases.
Therefore, as shown in Fig. 6, EPBP achieves the best
and most stable performance in terms of the message dis-
semination speed in a wide range of vehicle densities. Spe-
cially, at a vehicle density of 160vehicles/unit area, EPBP
attains 65.90% faster message dissemination speed com-
pared with 3P3B, 55.94% faster compared with the second
best performing PMBP. It is because that the contention la-
tency is the major contributor of one hop delay in high vehi-
cle densities.
To evaluate the reliability, the performances in terms
of the packet delivery ratio (PDR) at the vehicle density of
120 vehicles/unit area are compared in Fig. 7. The PDR
is the ratio of the number of vehicles that receive the EM
successfully to the total number of vehicles, both at the des-
tination. As shown, the PDR decreases with the increases
of the dissemination distance. However, among all exam-
ined protocols, EPBP achieves the highest and stable PDR,
which can be maintained to more than 99.99% even at the
dissemination distance of 20 km. The good performance of
EPBP profits from few collision between a few candidates
in the thinnest final segment. By using the exponential back-
offtimer, 3P3B and BPAB also achieve good performances
in terms of PDR, which are more than 99.97% and 99.81%
respectively. Although a higher PDR can be got by increas-
Fig. 6 Performance comparison in terms of message dissemination
speed.
Fig. 7 Performance comparison in terms of PDR.
ing Nre EPBP, the multi-hop delay will increase. So, EPBP
presents a good performance in terms of PDR at the cost of
the low delay.
5. Conclusion
In this paper, we have proposed an exponent-based parti-
tioning broadcast protocol, EPBP, for the EM dissemination
in vehicular networks. EPBP employs the exponent-based
partitioning method to identify the farthest possible relay.
The analytical and simulation results demonstrate that the
proposed EPBP presents a stable and improved performance
in terms of the message dissemination speed and PDR, es-
pecially at high vehicle densities. In addition, the interval
of the back-offtimer is determined to provide robustness
against the spurious forwarding problem.
Our ongoing work is focused on relaxing the assump-
tions of the presented model (especially the road condition,
e.g. intersection) for more realistic scenarios.
8
IEICE TRANS. FUNDAMENTALS, VOL.E99–A, NO.11 NOVEMBER 2016
Acknowledgments
We thank the reviewers for their helpful comments and
the Joint Funds of the National Natural Science Foun-
dation of China (Grant No.U1404615), Open Funds
of State Key Laboratory of Millimeter Waves (Grant
No.K201504), China Postdoctoral Science Foundation
(Grant No.2015M571637), and Science and Technology
Project of Hunan Province (Grant No.2015JC3057) for their
support.
References
[1] R. Chen, W.-L. Jin, and A. Regan, “Broadcasting safety information
in vehicular networks: Issues and approaches,” IEEE Netw., vol.24,
no.1, pp.20–25, Jan.-Feb. 2010.
[2] S. Ucar, S. Coleri Ergen, and O. Ozkasap, “Multi-hop cluster
based IEEE 802.11p and LTE hybrid architecture for VANET safety
message dissemination,” IEEE Trans. Veh. Technol., 2015. DOI:
10.1109/TVT.2015.2421277
[3] G. Karagiannis, O. Altintas, E. Ekici, G. Heijenk, B. Jarupan, K.
Lin, and T. Weil, “Vehicular networking: A survey and tutorial on
requirements, architectures, challenges, standards and solutions,”
IEEE Commun. Surv. Tutorials, vol.13, no.4, pp.584–616, Oct.
2011.
[4] M.R. Jabbarpour, A. Jalooli, E. Shaghaghi, A. Marefata, R.M. Noor,
and J.J. Jung, “Analyzing the impacts of velocity and density on
intelligent position-based routing protocols,” J. Comput. Sci., 2014.
DOI: 10.1016/j.jocs.2014.11.003
[5] S. Bitam, A. Mellouk, and S. Zeadally, “Bio-inspired routing algo-
rithms survey for vehicular ad hoc networks,” IEEE Commun. Surv.
Tutorials, vol.17, no.2, pp.843–867, April 2015.
[6] H. Yoo and D. Kim, “ROFF: RObust and fast forwarding in vehic-
ular ad-hoc networks,” IEEE Trans. Mobile Comput., vol.14, no.7,
pp.1490–1502, July 2015.
[7] J.J. Blum, A. Eskandarian, and L.J. Hoffman, “Challenges of inter-
vehicle ad hoc networks,” IEEE Trans. Intell. Transp. Syst., vol.5,
no.4, pp.347–351, Dec. 2004.
[8] M. Torrent-Moreno, M. Killat, and H. Hartenstein, “The challenges
of robust inter-vehicle communications,” Proc. 2005 IEEE 62nd Ve-
hicular Technology Conference, VTC-2005-Fall, pp.319–323, 2005.
[9] X. Ma, J. Zhang, X. Yin, and K.S. Trivedi, “Design and analysis of a
robust broadcast scheme for VANET safety-related services,” IEEE
Trans. Veh. Technol., vol.61, no.1, pp.46–61, Jan. 2012.
[10] G. Korkmaz, E. Ekici, and F. Ozguner, “Black-burst-based multi-
hop broadcast protocols for vehicular networks,” IEEE Trans. Veh.
Technol., vol.56, no.5, pp.3159–3167, Sept. 2007.
[11] Y. Bi, H. Zhao, and X. Shen, “A directional broadcast protocol
for emergency message exchange in inter-vehicle communications,”
Proc. 2009 IEEE International Conference on Communications,
pp.1–5, 2009.
[12] J. Sahoo, E.H.-K. Wu, P.K. Sahu, and M. Gerla, “Binary-partition-
assisted MAC-layer broadcast for emergency message dissemination
in VANETs,” IEEE Trans. Intell. Transp. Syst., vol.12, no.3, pp.757–
770, Sept. 2011.
[13] C. Suthaputchakun, M. Dianati, and Z. Sun, “Trinary partitioned
black-burst-based broadcast protocol for time-critical emergency
message dissemination in VANETs,” IEEE Trans. Veh. Technol.,
vol.63, no.6, pp.2926–2940, July 2014.
[14] S.-Y. Ni, Y.-C. Tseng, Y.-S. Chen, and J.-P. Sheu, “The broad-
cast storm problem in a mobile ad hoc network,” Proc. 5th An-
nual ACM/IEEE International Conference on Mobile Computing
and Networking, MobiCom’99, pp.151–162, 1999.
[15] S.M. Bilal, C.J. Bernardos, and C. Guerrero, “Position-based routing
in vehicular networks: A survey,” J. Netw. Comput. Appl., vol.36,
no.2, pp.685–697, March 2013.
[16] J.L. Sobrinho and A.S. Krishnakumar, “Distributed multiple ac-
cess procedures to provide voice communications over IEEE 802.11
wireless networks,” Proc. 1996 IEEE Global Telecommunications
Conference, GLOBECOM’96, pp.1689–1694, 1996.
[17] A. Baiocchi, P. Salvo, F. Cuomo, and I. Rubin, “Understanding spu-
rious message forwarding in VANET beacon-less dissemination pro-
tocols: An analytical approach,” IEEE Trans. Veh. Technol., 2015.
DOI: 10.1109/TVT.2015.2422753
[18] P. Del Moral, A. Doucet, and A. Jasra, “Sequential Monte Carlo
samplers,” J. Royal Statistical Soc. B, vol.68, no.3, pp.411–436,
June 2006.
[19] A.B. Reis, S. Sargento, F. Neves, and O.K. Tonguz, “Deploying
roadside units in sparse vehicular networks: What really works and
what does not,” IEEE Trans. Veh. Technol., vol.63, no.6, pp.2794–
2806, July 2014.
[20] V. Shah, N. Mehta, and R. Yim, “Optimal timer based selection
schemes,” IEEE Trans. Commun., vol.58, no.6, pp.1814–1823, June
2010.
Dun Cao received the B.S. degree in Com-
munication Engineering from Central South
University, China, in 2001, and the M.S. degree
in Information Systems and Communications
from Hunan University, China, in 2006. She is
currently a Ph.D student at the School of Auto-
motive and Mechanical Engineering, Changsha
University of Science and Technology, China,
and is also the faculty member of the School of
Computer and Communication Engineering of
the same university. She was a visiting scholar
at National Mobile Communications Research Laboratory, Southeast Uni-
versity, China from September 2012 to June 2013. Her current research
interests include Vehicular Networks and MIMO wireless communications.
Zhengbao Lei received the B.Sc. degree
in Mechanical Engineering from Sanxia Univer-
sity, China, in 1985, the M.Sc. degree in Vehicle
Engineering from Southwest Jiaotong Univer-
sity, China, in 1987, and Ph.D. degrees in Vehi-
cle Engineering from Hunan University, China,
in 1999. From 1999 to 2003, he was a Post-
doctoral Fellow with Central South University,
China. Currently, he is a professor in the Depart-
ment of Automotive and Mechanical Engineer-
ing, Changsha University of Science and Tech-
nology. He is the author or coauthor of over 180 technical papers and six
academic books. His research has been funded by the National Natural
Science Foundation of China, the National High Technology Research and
Development Program of China, National Science and Technology Sup-
port Program of China, etc. His current research interests focus on Vehicle
Safety Theory.
CAO et al.: EPBP FOR EMERGENCY MESSAGE DISSEMINATION IN VEHICULAR NETWORKS
9
Baofeng Ji received the Ph.D. degree from
Southeast University, Nanjing, China in 2013
with Information and Communication Engineer-
ing. He was a post-doctoral fellow from 2015 in
the school of information science and engineer-
ing, Southeast University, Nanjing, China. He
was an assistant professor in Henan University
of Science and Technology from 2013. His cur-
rent research interests include signal processing
for wireless communications, MIMO communi-
cations, cooperative relaying systems, and sta-
tistical signal processing. He is the author or coauthor of more than 40
papers and holds several patents.
Chunguo Li received his bachelor from
Shandong University in 2005, Ph.D. from the
Southeast University in 2010, all in wireless
communications. From June 2010, he joined in
the faculty of Southeast University in Nanjing,
where he becomes the Associate Professor since
April 2012. From July 2012 to June 2013, he did
the postdoctoral research in Concordia Univer-
sity, Montreal, Canada. From July 2013 to Au-
gust 2014, he joined the DSL laboratory super-
vised by Prof. John M. Cioffi. Dr. Li’s research
interests are in the MIMO relay communications, green communications,
next generation of WiFi. He is currently an Associate Editor for Circuits,
Systems and Signal Processing, the editor for KSII Transactions on Internet
and Information Systems and has served for many IEEE conferences such
as ICC, ICASSP as the TPC member.
