Research Article

International Journal of Distributed

Sensor Networks
2017, Vol. 13(11)
A neighbor knowledge and velocity- Ó The Author(s) 2017
DOI: 10.1177/1550147717743699
based broadcast scheme for wireless journals.sagepub.com/home/ijdsn

ad hoc networks

Dingzhu Lu1,2 and Shoubin Dong2

Abstract
A neighbor knowledge-based broadcast scheme is proposed to reduce the latency for wireless ad hoc networks, yet
keeping the overhead at a reasonably low level and fulfilling the reliability. In the scheme, few Hello messages are inter-
changed to collect one-hop neighbor information. The collected information is used to calculate the neighbor density,
the ratio, and the number of one-hop uncovered neighbors, upon which the rebroadcast probability and delay are
adjusted adaptively. The way that the rebroadcast probability and delay are defined in neighbor knowledge-based broad-
cast scheme reduces the transmission overhead and restrains the traffic aggregation effectively. Next, a velocity-based
data distribution mechanism is proposed and extended to neighbor knowledge-based broadcast scheme to further
reduce the latency, forming neighbor knowledge and velocity-based broadcast scheme. It is stipulated that few higher-
velocity nodes are employed with bigger probability to rebroadcast the incoming message. The performance of the
schemes is evaluated by the simulation under diverse network configurations. The results show that they outperform
the existing broadcast schemes in overhead and especially in average end-to-end delay. Compared with flooding, they
reduce the overhead by 88.4% and the average end-to-end delay by 88.9% at most.

Keywords
Wireless ad hoc networks, rebroadcast probability, rebroadcast delay, normalized broadcast overhead, average end-to-
end delay, packet delivery ratio

Date received: 2 April 2017; accepted: 24 October 2017

Handling Editor: Daniel Gutierrez-Reina

Introduction decades, broadcasting has attracted much attention of

Broadcasting is a communication primitive in wireless
ad hoc networks. It is essential to resolve many issues
such as route discovery, route maintenance for unicast
or multicast and emergency or warning message disse-
mination in disaster, battlefield, and VANET scenario.
In some harsh conditions, an efficient broadcast
mechanism is vital and may be the only possible way to
disseminate crucial information.1 Because radio signals
are likely to overlap with each other in a same geogra-
phical area, a straightforward broadcasting by flooding
is usually very costly and will result in serious redun-
dancy, contention, and collisions, to which is referred
as the broadcast storm problem.2 Thus, during the past
the research groups. Many kinds of broadcast schemes
were proposed to minimize the number of transmis-
sions while attempting to ensure a high reliability and a
low latency.
1
School of Information & Engineering, The Open University of
Guangdong, Guangzhou, China
2
Computer Network and Communication Key Laboratory of
Guangdong, South China University of Technology, Guangzhou, China
Corresponding author:
Dingzhu Lu, Guangdong Key Laboratory of Computer Network, South
China University of Technology, Guangzhou 510006, China.
Email: dzlu@scut.edu.cn

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (http://www.uk.sagepub.com/aboutus/
openaccess.htm).
2 International Journal of Distributed Sensor Networks
Besides blind flooding,3,4 a basic classification of
broadcast schemes divides them into two main categories:
probabilistic broadcast scheme and deterministic broad-
cast scheme.1,5,6 The probabilistic broadcast scheme
includes probability-based schemes,2,7–11 counter-based
schemes,12–16 distance-based schemes,17–23 and location-
based schemes.24–27
Probabilistic schemes do not need any information
from the whole network and need not maintain a glo-
bal distributing topology, but, rather, start of building
a network with each broadcast domain. In probabilistic
schemes, all nodes are allowed to participate in the
broadcast process and forward the received broadcast
message with some probability. Therefore, the traffic
load would not aggregate on a part of nodes.
Furthermore, the interest in probabilistic broadcasting
schemes is due to their inherent low transmission over-
head, low processing complexity, low delay, and high
tolerance to frequent and rapid topological changes.
Balancing these benefits, though, is the disadvantage of
inability to guarantee full network coverage and, still,
the presence of some redundant transmissions.
Deterministic broadcast scheme includes tree-based
scheme,28–31 connected dominating set (CDS), and
neighbor elimination-based scheme.32–41 Tree-based
schemes construct a routing tree on which broadcast
messages are forwarded. In tree-based scheme broad-
cast redundancy minimization scheduling (BRMS),31 a
set of forwarding nodes are found first to minimize the
number of broadcast transmissions. Then, the scheme
constructs a forest of sub-trees based on the relation-
ship between each forwarding node and its correspond-
ing receivers. Finally, a broadcast tree is constructed by
connecting all sub-trees with a minimum number of
connectors. CDS and neighbor elimination-based
schemes maintain state on the neighborhood of each
node, via ‘‘Hello’’ packets. Upon the collected neighbor
information, each node decides its role (forward node
or non-forward node) in a specific broadcasting. For
example, in local broadcast algorithms (LBA)38 pro-
posed by Majid Khabbazian, every broadcasting node
selects at most one of its neighbors to rebroadcast the
received message. A node has to rebroadcast the
received message if it is selected to do. Other nodes that
are not selected have to decide whether to rebroadcast
upon a self-pruning condition called the coverage con-
dition. To evaluate the coverage condition, every node
u maintains a list Listucov (m) for every unique message
m. When node u receives a fresh message m for the first
time, Listucov (m) is initialized and filled with the IDs of
all neighbors of u. Listucov (m) is updated before the
rebroadcast delay ends, namely, the IDs of neighbors
covered by m are being removed from Listucov (m) during
the rebroadcast delay. Finally, if Listucov (m) is not
empty, u will rebroadcast m and select one node still in
the list to be the next forwarder. Deterministic schemes
use topological information to build a global forwarding
structure to reduce broadcast redundancy. They can be
proven to not degrade reachability compared with flood-
ing and probabilistic schemes, provided communications
are perfectly reliable. However, they do not tolerate fre-
quent topological changes well, because maintaining the
forwarding structure in dynamic environment brings
excessive control overhead. In high mobile environment,
deterministic schemes could not provide higher reachabil-
ity than probabilistic schemes do.
First, we proposed a neighbor knowledge-based
broadcast (NKB) scheme for wireless ad hoc networks.
In NKB scheme, few Hello messages are interchanged
between one-hop neighbors to collect the neighbor infor-
mation. The rebroadcast probability and the rebroadcast
delay on each node are adjusted adaptively according to
the obtained neighbor knowledge including the neighbor
density, the ratio, and the number of one-hop uncovered
neighbors. Second, a velocity-based data distribution
mechanism is proposed and extended to NKB scheme to
reduce the latency further, forming neighbor knowledge
and velocity-based broadcast (NKVB) scheme. In NKVB
scheme, few higher-velocity nodes are employed with big-
ger probability to forward the broadcast message, which
makes NKVB different from NKB scheme, and they do
accelerate the data transmission.
The rest of the article is organized as follows. Section
‘‘Related work’’ presents a brief review of broadcast
schemes. Section ‘‘NKB scheme’’ presents NKB scheme.
The velocity-based data distribution mechanism and
NKVB scheme are presented in section ‘‘NKVB scheme.’’
Performance evaluation of NKB and NKVB schemes by
simulations in the context of wireless ad hoc networks is
presented in section ‘‘Performance evaluation.’’ Finally,
section ‘‘Conclusion’’ concludes the article.
Related work
Probabilistic broadcast schemes only need to maintain
rough local topology knowledge, which makes it more
insensitive to node mobility than deterministic broad-
cast schemes. Probability-based schemes2,8 use some
basic understanding of the network topology to assign
a constant forwarding probability to each node to
rebroadcast. The fixed probability schemes do not need
to collect the beacon information, including density,
speed, and energy, and so on, upon which the optimal
rebroadcast probability is calculated in the adaptive
probability schemes.1 GOSSIP1( p)8 is a pure gossip
scheme with the fixed rebroadcast probability. When a
node first receives a route request, with probability p, it
broadcasts the request to its neighbors and with prob-
ability 1 – p, it discards the request. If the node receives
the same route request again, it is discarded. The fixed
probability schemes cost less control overhead. But the
Lu and Dong
constant rebroadcast probability makes the schemes
function not so well in dynamic and asymmetrical net-
work environment. In reliable gossip-based broadcast
protocol (RGB)7 and velocity and neighbor density-
based broadcast scheme (VDNB),11 the dynamic
rebroadcast probability is calculated upon the immedi-
ate number of neighbors of the node. Compared with
the fixed probability schemes, they are more suitable
for wireless ad hoc networks.
In dynamic counter-based broadcasting scheme
(DCB),15 the number of neighbors is used to determine
whether the node is located within dense, medium distri-
bution or sparse regions and assign Cmin, Cmid, and Cmax
thresholds, respectively. Each node needs to wait for a
random assessment delay. Within the delay, the number
of copies of the message received by a node is compared
with Cmin, Cmid, and Cmax, and then the corresponding
one of four different rebroadcast probabilities is assigned
to the node. In Humoud et al.,13 only two thresholds,
unlike in Yassein et al.,15 are used to evaluate the node
distribution and the rebroadcast probability is rougher.
In Tseng et al.,12 the decision to forward the broadcast
message is determined by the function C(n) where n is
the number of neighbors of the forwarding node.
In distance-based broadcast scheme,17 whether a
message is rebroadcasted or not is up to the relative dis-
tance between the mobile node and the previous sender.
When a node receives a broadcast message from a
neighbor node for the first time, its dmin is initialized
with the distance between them and the count is initia-
lized with 1. If dmin \ D (where D is the distance thresh-
old pre-defined), the node stops rebroadcasting. Else,
the node waits for a random number of slots, during the
waiting period, every time a copy of the broadcast mes-
sage is received, dmin and D are reassigned correspond-
ingly. If dmin \ D, the node stops rebroadcasting. Else,
if the waiting time is expired, the node rebroadcasts the
received message. In distance and cooperation based
broadcast (DCBB),23 four neighbor nodes at most are
determined to forward broadcast packets based on the
number of neighbors and the distance between neigh-
bors. The limited number of relay nodes is helpful to
reduce the redundancy of the scheme. DCBB is applica-
ble to densely distributed network environment.
In location-based broadcast schemes,26,27 the
rebroadcast probability of a node is in positive correla-
tion with its additional coverage ratio. In order to fig-
ure out an accurate additional coverage ratio, the
rebroadcast delay is set without exception on each node
to collect enough neighbor coverage information. In
scalable broadcast algorithm (SBA),26 three node-
stamping broadcast algorithms, named basic stamping,
advanced stamping, and hybrid stamping, are pro-
posed. Each node in advanced stamping algorithm col-
lects its one-hop neighbor information. IDs of each
forwarding node and its neighbors are appended to
3
stamp in advanced stamping scheme. When node r
receives a broadcast message m for the first time, it uses
the stamp of m to check whether all its neighbors N(r)
are already covered. If so, the rebroadcast is stopped;
otherwise, node r appends id(r) and id(N(r)) to the
stamp and schedules for rebroadcasting. Hybrid stamp-
ing utilizes two-hop neighbor information for pruning.
With hybrid stamping, each node further checks
whether some of its neighbors can also be reached by
other nodes in the stamp. When node r receives a new
message m, it checks whether its neighbors N(r) are all
contained in the stamp of m first. For each neighbor n
not contained in the stamp, node r then tries to check
whether a neighbor p in N(n) is in the stamp. If node p
has a higher priority than node r, node r can assume
node n will be covered by broadcasting from node p or
node with an even higher priority. If all uncovered
neighbors in N(r) will be covered by node with higher
priority, the rebroadcast is stopped by node r; other-
wise, node r appends id(r) and id(N(r)) to the stamp
and schedules for rebroadcasting.
In neighbor coverage-based probabilistic rebroad-
cast (NCPR),27 the proportion of the common neigh-
bors between node ni and its upstream node s uniquely
determines the rebroadcast delay Td(ni) of node ni, and
the rebroadcast delay of node ni is defined as follows
 
jN (s) \ N (ni )j
Td (ni ) = MaxDelay 3 1  ð1Þ
jN (s)j
where MaxDelay is a small constant delay; N(s) and
N(ni) are the neighbor sets of node s and ni, respectively.
|.| is the number of elements in a set. In NCPR, another
parameter, the connectivity factor, is defined to provide
the node density adaptation. The connectivity factor
Fc(ni) and the additional coverage ratio Ra(ni) of node
ni are defined as follows
NC
Fc (ni ) = ð2Þ
jN (ni )j
jU (ni )j
Ra (ni ) = ð3Þ
jN (ni )j
where Nc = 5.1774 log n, and n is the number of nodes
in the network. |U(ni)| is the number of uncovered
neighbors of node ni. By combining the additional cov-
erage ratio and the connectivity factor, the rebroadcast
probability of node ni is calculated as follows
Pre (ni ) = Fc (ni )  Ra (ni ) ð4Þ
The proposed rebroadcast delay is one of main con-
tributions of NCPR. It determines the rebroadcast
order, and then the more accurate additional coverage
ratio is obtained.27 We found that, in NCPR, more
neighbors a node has, more likely it is to rebroadcast
4 International Journal of Distributed Sensor Networks

the message in less time. Namely, the way that the

rebroadcast delay and the rebroadcast probability are
calculated is inclined to incur the traffic aggregation.
In NKB scheme, we focus on defining the rebroad-
cast delay and the rebroadcast probability more rea-
sonably to balance the traffic load, reduce the collisions
and lower the end-to-end delay, which improved the
reliability and the scalability of the scheme. For adapt-
ability purpose in mobile environment, for example in
disaster and battlefield scenario, few higher-velocity
nodes are employed with bigger probability to rebroad-
cast the message in enhanced NKVB scheme, which Figure 1. Neighbor set and uncovered neighbors set.
lowers the delay further.

NKB scheme Uncovered neighbors set

In NKB scheme, the rebroadcast probability is adjusted
adaptively at each node according to its neighbor
knowledge. The neighbor knowledge includes the ratio
of uncovered neighbors, the number of uncovered neigh-
bor, and the neighbor density. The neighbor knowledge
is obtained at each node by exchanging Hello message.
In order to sufficiently pursue lower average end-to-
end delay, exploit more accurate uncovered neighbors
set, and avoid channel collisions, a proper rebroadcast
delay is set at each node.
For research purpose, the network can be repre-
sented by an undirected graph G = (V , E), where V is a
set of vertices (nodes) and E is a set of undirected edges.
An undirected edge (ni, nj) denotes the connection
between two neighboring nodes ni and ni. The neighbor
set N(ni) of node ni is defined as fni j(ni , nj ) 2 Eg. |N(ni)|
denotes the number of elements in set N(ni).
Collection of neighbor information
Hello message is used to collect a node’s one-hop
neighbor information in NKB scheme. To reduce the
overhead of Hello message, piggybacking is used.
Moreover, piggybacking of the Hello message by
broadcast packet could make the neighbor information
fresher. Only when the time elapsed from the last
broadcasting packet is greater than a Hello Interval,
the node needs to send a separate Hello message.
In NKB, Hello message of node ni includes the
following:
1. N(ni): a neighbor list of node ni;
2. M(ni): an ID list of broadcast messages received;
3. flag_leaf: if node ni is a leaf node, flag leaf is set
to 1. Otherwise, this field is excluded from Hello
message.
Definitions
leaf node: A leaf node has only one neighbor; it is a
special kind of node who locates at the boundary.
secondary leaf node: It is a kind of node who is the
neighbor of a leaf node.
When node ni receives a fresh broadcast message from
the upstream node ns, their common neighbors receive
it too because of the broadcast nature of radio com-
munication, as shown in Figure 1. Upon both node
ns’s neighbor list and its own neighbor list collected by
the Hello message, node ni can figure out who are
their common neighbors and deduce its uncovered
neighbors set. If node ni has more neighbors uncov-
ered by the broadcast message, it is reasonable for
node ni to rebroadcast the message with a bigger
probability. The uncovered neighbors set U(ni) of
node ni is defined as follows
U (ni ) = N (ni )  ½N (ni ) \ N (ns )  fns g ð5Þ
where N(ni) and N(ns) are the neighbor sets of node ni
and ns, respectively.
In NKB scheme, each node sets a rebroadcast delay
(defined in section ‘‘Rebroadcast delay’’) when it
receives a fresh broadcast message, to exploit its accu-
rate uncovered neighbors set. When node ni receives a
duplicate message during the rebroadcast delay, for
example from its neighbor node nj, node ni should
update its uncovered neighbors set U(ni) according to
formula (6). The uncovered neighbors set U(ni) is being
updated until the rebroadcast delay expires
U ðni Þ = U ðni Þ  ½U ðni Þ \ N (nj )  fnj g ð6Þ
Rebroadcast probability based on neighbor
knowledge
In proposed broadcast scheme, the neighbor knowl-
edge, including the ratio of uncovered neighbors, the
number of uncovered neighbors, and the neighbor den-
sity, is used to calculate the rebroadcast probability at
each node.
Ratio of uncovered neighbors. As soon as the rebroadcast
delay expires, node ni will calculate the ratio of uncov-
ered neighbors Ru(ni) as follows
Lu and Dong 5

jU (ni )j
Ru (ni ) = ð7Þ
jN (ni )j  1
where |.| is the number of elements in a set and
|N(ni)| – 1 is the number of node ni’s neighbors exclud-
ing the upstream node ns.
nj
If node ni has a big ratio of uncovered neighbors, it is
reasonable for ni to rebroadcast the received message ns ni U(n )
with a big probability to cover more fresh nodes. On i
the contrary, if node ni has a small ratio of uncovered
neighbors, the smallest possible rebroadcast probability
should be adopted by ni to reduce redundancy and col-
lisions without hurting the packet delivery ratio. Figure 2. Leaf node and secondary leaf node.

Density factor and absolute number of uncovered neighbor

where b and (1 – b) are the influence coefficients of
factor. We assume that node nh and ni are neighbors of
Ru(ni) and Fu(ni), respectively. If pk(ni) is greater than 1,
the upstream node ns and receive the same broadcast
we set pk(ni) to 1. The value of b is related to node den-
message from node ns. If they have the same ratio of
sity and node distribution. In even and dense network,
uncovered neighbors but the different numbers of uncov-
a big value of b assigned benefits reducing the colli-
ered neighbors, it is reasonable for the node with more
sions and the delay, and upgrading the reliability. A
uncovered neighbors to rebroadcast the message with a
small value of (1 – b) is helpful to improve the packet
bigger probability. Even when a node has a small ratio
delivery ratio in sparse area. A number of experiments,
of uncovered neighbors, but a big absolute number of
which are not emphases mentioned here, were done to
uncovered neighbors, it is worth rebroadcasting the
prove it. And as a result, the empirical value of b is
message with a big probability.
set to 0.8 in the following simulation in section
Besides the number of uncovered neighbors, the
‘‘Performance evaluation.’’
neighbor density also affects the rebroadcast probabil-
But, if node ni is an uncovered secondary leaf node,
ity. When node ni is in a denser area, a smaller rebroad-
whose leaf node is nj as shown in Figure 2, it should
cast probability adopted avails to reduce the redundancy
rebroadcast the message with probability 1 for node nj.
and the collisions. However, a bigger rebroadcast prob-
In addition, nj should give up the rebroadcast as a last
ability should be adopted to keep a higher packet deliv-
node on the forwarding path. Finally, we have the
ery ratio in a sparser area.
rebroadcast probability pk(ni) as follows
Xue and Kumar42 concluded that if each node con-
nects to more than 5.1774 logN of its neighbors, then pk (ni ) =
the probability of the network being connected is 8
>
> 0, U (ni ) = F
approaching 1 as N increases, where N is the number <
1, 9nj , nj 2 U (ni ) and N (nj ) = fni g
of nodes in the network. We use 5.1774 logN, denoted >  
>
: min 1, bRu (ni ) + (1b)Fu (ni )
by Nc, as a reference value to evaluate the absolute d(ni ) other
number of uncovered neighbors and the neighbor den-
sity. The density factor d(ni) and the absolute number of
ð11Þ
uncovered neighbor factor Fu(ni) are defined as follows

jN (ni )j
d(ni ) = ð8Þ Rebroadcast delay
NC
jU (ni )j The definition of the rebroadcast delay is critical for
Fu (ni ) = ð9Þ the scheme to achieve optimal performance. The com-
NC
mon neighbor ratio of node ns and node ni, denoted by
Rc(nsi), and the absolute number of uncovered neigh-
Rebroadcast probability based on neighbor knowledge.
bors influence the rebroadcast delay in a different way.
Combining the ratio of uncovered neighbors, the abso-
Rc(nsi) is defined as follows
lute number of uncovered neighbor factor, and the den-
sity factor, we have node ni’s rebroadcast probability jN (ns ) \ N (ni )j
pk(ni) based on its neighbor knowledge Rc (nsi ) = ð12Þ
jN (ns )j  1
b  Ru (ni ) + (1  b)  Fu (ni ) where |N(ns)| – 1 is the number of node ns’s neighbors
pk (ni ) = ð10Þ
d(ni ) excluding node ni.
6 International Journal of Distributed Sensor Networks

The rebroadcast delay ratio Rd(ni) and the rebroad-

cast delay Td(ni) are defined as follows

1  Fu (ni )  Rc (nsi ) N (ns ) \ N (ni ) 6¼ F
Rd (ni ) =
0 N (ns ) \ N (ni ) = F or 9nj , nj 2 U (ni ) and N (nj ) = fni g ð13Þ

D  Rd (ni ) N (ns ) \ N (ni ) 6¼ F
Td (ni ) =
0 N (ns ) \ N (ni ) = F or 9nj , nj 2 U (ni ) and N (nj ) = fni g ð14Þ

where D is a small constant delay. When there is no The first two points are obviously reasonable. We
common neighbor between node ni and ns or ni is a sec- present the proof for the rationality of the third point
ondary leaf node, Rd(ni) and Td(ni) are set to 0. If Rd(ni) in the following.
is smaller than 0, we set Rd(ni) to 0.
Next, the details that the rebroadcast delay is defined Proof 
are explained. Suppose node ni receives a fresh broad- Ru (ni ) = x
cast message from its upstream node ns, their common Ru (nj ) = y
neighbors receive the message too. 8
>
> jU (ni )j
>
< jN (n )j  1 = x
i
1. If node ni has more uncovered neighbors, )
namely a bigger Fu(ni), it should rebroadcast the >
> jU (nj )j
>
: =y ð15Þ
received message with a smaller rebroadcast jN (nj )j  1
8
delay to make the message cover more fresh > jU (ni )j
>
< jN (ni )j = 1 +
nodes as soon as possible. In this condition, the x
smaller delay is helpful to accelerate the mes- )
>
> jU (n j )j
sage propagation and reduce the average end- : jN (nj )j = 1 +
y
to-end delay.
2. If N(ns) \ N(ni) = F, namely there is no com- Note that
mon neighbor between node ns and node ni, 
jN (ni )j  jU (ni )j = jN (ns ) \ N (ni )j
node ni rebroadcasts the message without delay ð16Þ
jN (nj )j  jU (nj )j = jN (ns ) \ N (nj )j
(Td(ni) = 0) will not incur the collision. If
9 nj , N (nj ) = fni g, namely node ni is a secondary If Rc(nsi) . Rc(nsj), then
leaf node, node ni has to rebroadcast the mes- jN (ns ) \ N (ni )j jN (ns ) \ N (nj )j
sage for the boundary node nj, and that node ni .
jN (ns )j  1 jN (ns )j  1 ð17Þ
rebroadcasts the message without delay
(Td(ni) = 0) is helpful to accelerate the message ) jN (ns ) \ N (ni )j . jN (ns ) \ N (nj )j
propagation. Formula (18) can be derived from formulas (16)
3. If bigger the common neighbor ratio of node ns and (17).
and ni is, shorter the rebroadcast delay of node
ni should be. Suppose node ni and node nj are N (ni )j  jU (ni )j . jN (nj )j  jU (nj )j ð18Þ
neighbors of the upstream node ns. If the com- Formula (19) can be derived from formulas (15)
mon neighbor ratio of node ns and node ni is and (18).
bigger than the common neighbor ratio of node
ns and node nj (Rc (nsi ).Rc (nsj )), node ni should jU (ni )j jU (nj )j
1+  jU (ni )j . 1 +  jU (nj )j
rebroadcast the message earlier than node nj. x y
    ð19Þ
Then, more nodes, namely these common neigh- 1 1
)  1 jU (ni )j .  1 jU (nj )j
bors between ns and ni, can update their own x y
uncovered neighbors set upon the received 8
neighbor list from node ni. Next, their rebroad- < Ru (ni )  Ru (nj )
>
cast probability will be lowered to reduce redun- Ru (ni ) = x
>
:
dant messages. If Rc (nsi ).Rc (nsj ) and Ru (nj ) = y
Ru (ni )  Ru (nj ), the rebroadcast by node ni 
x\1, y\1 ð20Þ
within smaller delay will cover more additional )
fresh nodes in less time than by node nj. It is xy
   
helpful to improve the channel utilization and 1 1
) 0\  1  1
reduce the average end-to-end delay. x y
Lu and Dong 7

Combining formulas (19) and (20), we draw the con- big challenge for most of existing broadcast schemes to
clusion as follows minimize the harm of node mobility. This section pro-
poses a velocity-based data distribution mechanism. It
jU (ni )j . jU (nj ) ð21Þ stipulates that few higher-velocity nodes should be
selected to rebroadcast the message with a big probabil-
ity without delay to accelerate the data propagation.
NKB algorithm description The mechanism is extended to NKB scheme to reduce
In NKB algorithm, the rebroadcast probability of node average end-to-end delay further, forming NKVB
ni is determined by its neighbor knowledge. The formal scheme.
description of NKB scheme is shown in the following.
The algorithm consists of three phases. In the first Velocity-based data distribution mechanism
phase, U(ni) is initialized. If U(ni) = F, node ni refuses
to rebroadcast the received message, else, Td(ni) is cal- Suppose node ni has ki neighbors, denoted by
culated. Next, if U(ni) 6¼ F, U(ni) is being updated fnj j(ni , nj ) 2 E, j = 1, 2, . . . , ki g, whose velocities are
before Td(ni) expires. Finally, when Td(ni) expires and denoted by vj ( j = 1, 2,., ki). The minimum and max-
U(ni) 6¼ F, pk(ni) is calculated and the message is imum velocity of them are denoted by vmin and vmax,
rebroadcasted with probability pk(ni). respectively

vmax = max(v1 , v2 , . . . , vj )
j = 1, 2, . . . , ki ð22Þ
NKVB scheme vmin = min(v1 , v2 , . . . , vj )
Several studies43 draw the conclusion that the mobility The arithmetic mean of these neighbors’ velocities v
increases the capacity of networks. However, it still is a and the standard deviation of their velocities s are cal-
culated respectively as follows
Algorithm 1. Neighbor knowledge-based broadcast algorithm. !
1 X ki
v = vj + vi ð23Þ
Definitions: ki + 1 j = 1
m: the received broadcast message.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
Timer(ni, m, Td(ni)): timer of node ni for m message, initialized X ki qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
with the value of Td(ni). s= (vj  v)2 + (vi  v)2 ð24Þ
random(0, 1): a random function whose range is the interval ki + 1 j = 1
from 0 to 1.
1: While ni receives a fresh message m from ns do The velocity-based data distribution mechanism sti-
2: U(ni ) = N(ni )  ½N(ni ) \ N(ns )  fns g
pulates strictly that node ni is allowed to rebroadcast
3: if U(ni ) = F then
4: pk(ni) = 0 and return the received message with the probability pv(ni) only if
5: endif vi is bigger than (v + s). The rebroadcast probability
6: if N(ns ) \ N(ni ) = F or 9nj , N(nj ) = fni g then pv(ni) is defined by formula (25) as follows
7: Td(ni) = 0
8: else vi  (v + s)
9: Td (ni ) = D3Rd (ni ) pv ð ni Þ = ð25Þ
10: Set a Timer(ni, m, Td(ni)) vmax  (v + s)
11: endif
12: endwhile
The overhead incurred by these employed high-
13: while ni receives a duplicate m from nj before velocity nodes is very limited. Assume that the node
Timer(ni,m,Td(ni)) expires do velocity obeys uniform distribution between vmin and
14: Uðni Þ = Uðni Þ  ½Uðni Þ \ N(nj )  fnj g vmax, the percentage PU of the nodes who take part in
15: if U(ni ) = F then computing the rebroadcast probability is 0.21 calcu-
16: pk(ni) = 0 and return
17: endif
lated by formula (27)
18: endwhile 8
19: While Timer(ni, m, Td(ni)) expires do > vmax + vmin
< vU =
20: if 9nj , N(nj ) = fni g then 2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
21: pk(ni) = 1 2 ð26Þ
>
: s = (vmax  vmin )
22: else U
23: pk (ni ) = min(1, (b:Ru (ni ) + (1  b):Fu (ni ))=d(ni ))
12
24: endif PU = 1  FU (vU + sU )
25: if random(0, 1)  pk(ni) then  
26: broadcast(m) and return vU + sU  vmin
=1  ð27Þ
27: endif vmax  vmin
28: endwhile
= 0:21
8 International Journal of Distributed Sensor Networks

Similarly, suppose the node velocity obeys normal Table 1. Simulation parameters.
distribution N(a, s2), the percentage PN of the nodes
who take part in computing the rebroadcast probability Simulation parameters Value
is 0.1587 calculated by formula (29) Mobility model RWM
 Topology size 1000 3 1000 m2
vN = a Numbers of nodes 50, 100, 150, 200, 250, 300
ð28Þ Transmission range 250 m
sN = s
Bandwidth 2 Mbps
PN = 1  FN (vN + sN ) Traffic type CBR
CBR connection 10, 20, 30, 40, 50, 60
= 1  Fa, s (vN + sN ) Packet size 512 bytes
= 1  Fa, s (a + s) Packet rate 4 packets/s
  Pause time 0s
(a + s)  a ð29Þ Max mobility velocity (m/s) 0, 5, 10, 15, 20, 25
=1  F
s MAC protocol 802.11 DCF CSMA/CA
Hello interval 0.1 s
= 1  F(1) Constant delay (D) 0.01 s
= 0:1587 Fixed probability(Gossip1(p)) 0.65

Note that ultimately, only few of nodes, who take
part in computing the rebroadcast probability, are
really employed to rebroadcast the message, so the pro-
posed velocity-based data distribution mechanism
incurs little overhead. It is also reconfirmed by the
simulation result in section ‘‘Performance evaluation.’’
NKVB algorithm description
In order to collect the neighbor nodes’ velocity, a new
field vi, the velocity of node ni, needs to be added to
Hello message. In NKVB algorithm, the rebroadcast
probability of node ni is determined by both its velocity
and its neighbor knowledge. The formal description of
the NKVB algorithm is shown in the following.
The algorithm consists of three phases. In the first
phase, U(ni) is initialized. If U(ni) 6¼ F and vi .(v + s),
node ni broadcasts the message with probability pv(ni)
without delay. If the message is not broadcasted since
the random number does not fall in the rebroadcast
probability interval from 0 to pv(ni), Td(ni) is initialized.
Next, if U(ni) 6¼ F, U(ni) is being updated before Td(ni)
expires. Finally, when Td(ni) expires and U(ni) 6¼ F,
pk(ni) is calculated and the message is rebroadcasted
with probability pk(ni).
Performance evaluation
In this section, NKB and NKVB schemes are com-
pared with flooding,4 Gossip1(p),8 DCB,15 NCPR,27
and LBA,38 by simulation way using the NS-2(v2.35)44
simulator. The IEEE 802.11 MAC with distributed
coordination function (DCF) is used as the MAC pro-
tocol. The mobility model is based on the random
waypoint model (RWM) in a square of size
1000 3 1000 m2. In the simulation of our schemes,
the empirical value of b is set to 0.8. The rebroadcast
probability p of Gossip1(p) is set to 0.65. The detailed
simulation parameters are listed in Table 1.
DCF: distributed coordination function.
We have used the following performance metrics for
comparison purposes:
 Normalized broadcast overhead: normalized
broadcast overhead indicates the normalized
transmissions of the broadcast operation per
node. It defines the ratio of the total transmis-
sions, including the broadcast packets and extra
control packets to the broadcast packets per
node. It is measured by bytes per broadcast byte
per node.
 Average end-to-end delay: the average delay of
successfully delivered broadcast packets from
source to destinations.
 Packet delivery ratio: the percentage of nodes
received the broadcast message.
The simulation is divided to three parts, and in each
part, we evaluate the impact of one of the following fac-
tors on the performance of broadcast schemes. In the
simulation results, each data point represents the aver-
age of 40 trials of experiments. The confidence level is
95%, and the confidence interval is shown as a vertical
bar in the figures:
 Mobility velocity. The max mobility velocity of
nodes is modulated in range of 0–25 m/s to eval-
uate how the node mobility affects the perfor-
mance of the schemes. In this part, the number
of nodes and the number of CBR connections
are set fixedly to 200 and 50, respectively.
 Number of nodes. The number of nodes is modu-
lated in range of 50–300 in a fixed square to
evaluate how the node density affects the perfor-
mance of the schemes. In this part, the max
mobility velocity of nodes and the number of
Lu and Dong 9

Algorithm 2. Neighbor knowledge and velocity-based broadcast

algorithm.

Definitions:
m: the received broadcast message.
Timer(ni, m, Td(ni)): timer of node ni for m message, initialized
with the value of Td(ni).
random(0, 1): a random function whose range is the interval
from 0 to 1.
1: While ni receives a fresh message m from ns do
2: U(ni ) = N(ni )  ½N(ni ) \ N(ns )  fns g
3: if U(ni ) = F then
4: p(ni) = 0 and return
5: endif
6: if vi . ðv + sÞ then
7: pv ðni Þ = (vi  (v + s))=(vmax  (v + s))
8: if random(0, 1)  pv(ni) then
9: broadcast(m) and return
10: endif
11: endif
12: if N(ns ) \ N(ni ) = F or 9nj , N(nj ) = fni g then Figure 3. Mobility velocity versus normalized broadcast
13: Td(ni) = 0 overhead.
14: else
15: Td (ni ) = D3Rd (ni )
16: Set a Timer(ni, m, Td(ni)) overhead curves are almost flat due to the fixed
17: endif rebroadcast probabilities which are not affected by the
18: endwhile node mobility. The normalized broadcast overhead of
19: while ni receives a duplicate m from nj before
Timer(ni,m,Td(ni)) expires do the other schemes increases proportional to the node
20: Uðni Þ = Uðni Þ  ½Uðni Þ \ N(nj )  fnj g mobility, namely, the node mobility has a negative
21: if U(ni ) = F then impact on the reduction of overhead. It is because the
22: p(ni) = 0 and return node mobility directly brings about more control pack-
23: endif
ets to repair the hurt communication link and to main-
24: endwhile
25: While Timer(ni, m, Td(ni)) expires do tain the changed network topology. Furthermore, the
26: if 9nj , N(nj ) = fni g then topological variation may incur some collisions fol-
27: pk(ni) = 1 lowed by the retransmissions. In some schemes, the
28: else node mobility as well gives rise to the imprecise neigh-
29: pk (ni ) = min(1, (b:Ru (ni ) + (1  b):Fu (ni ))=d(ni ))
30: endif
bor information collected and the unnecessary rebroad-
31: if random(0, 1)  pk(ni) then casts. As shown in Figure 3, when the mobility velocity
32: broadcast(m) and return is bigger than 15 m/s, there is a significant increase in
33: endif the normalized broadcast overhead, especially for LBA
34: endwhile and NCPR schemes.
In LBA scheme, a node refuses to rebroadcast the
received message only if it learned that all of its neigh-
CBR connections are set fixedly to 15 m/s and bors have been covered. In fact, it is almost impossible
20, respectively. for each node to acquire the intact neighbor coverage
 Number of CBR connections. The number of ran- knowledge in mobile environment. Then, many super-
domly chosen CBR connections is modulated in fluous rebroadcasts are executed, leading to some inci-
range of 10–60 with a fixed packet rate to evalu- dental collisions and retransmissions. Therefore, in
ate how the traffic load affects the performance LBA scheme, the redundant broadcast is the main
of the schemes. In this part, the max mobility composition of the normalized broadcast overhead. In
velocity of nodes and the number of nodes are mobile environment, DCB scheme tends to define a
set fixedly to 15 m/s and 200, respectively. rough threshold and count a wrong number of redupli-
cative packets received, which leads inevitably to the
redundant rebroadcasts and collisions. Compared with
Effects of node mobility LBA and DCB schemes, the normalized broadcast
Figure 3 shows that flooding and Gossip1(0.65) are overhead of NKB, NKVB, and NCPR schemes is
deserved a high normalized broadcast overhead. The affected more mildly by the node mobility since their
result is driven comprehensively by the high node den- rebroadcast probabilities do not rely on the precise
sity and their big rebroadcast probabilities. But their neighbor knowledge. Compared with Gossip1(0.65)
10
Figure 4. Mobility velocity versus average end-to-end delay.
and flooding, NKVB scheme saves 57.5%–84.9% over-
head and 65.4%–88.4% overhead, respectively, as
shown in Figure 3. Compared with NCPR scheme,
NKVB scheme saves 16.2%–26.9% overhead.
Figure 4 shows how the node mobility affects the
average end-to-end delay. Both Gossip1(0.65) and
flooding keep a stable and higher average end-to-end
delay. For the other schemes, as elaborated in Figure 3,
the topological change incurs more collisions and
retransmissions. Definitely, it leads to higher latency.
The incomplete topology knowledge exceedingly ampli-
fies the rebroadcast probability in LBA scheme, and
then many redundant messages are rebroadcasted,
which leads to a mass of channel competitions and col-
lisions. Definitely, LBA deserves the worst average
end-to-end delay among all schemes, especially when
the nodes are moving at high velocity.
In DCB scheme, the average end-to-end delay
increases with the node mobility due to the incremental
redundant rebroadcasts and collisions incurred by the
inaccurate threshold and count of reduplicative packets.
The rest of three schemes only need to maintain rela-
tively rough topology knowledge, so, the node mobility
has less influence on their average end-to-end delay
than that of LBA and DCB. Both NKB and NKVB
schemes outperform NCPR scheme far in average end-
to-end delay, as shown in Figure 4, thanks to their rea-
sonable rebroadcast delay and probability. Compared
with Gossip1(0.65) and flooding, NKVB scheme
reduces 53.4%–58.3% delay and 60.3%–63.6% delay,
respectively. Compared with NCPR scheme, NKVB
scheme reduces 25.2%–42.3% delay. And, there is an
apparent difference that the first half of the curve of
NKVB scheme shows a downward trend, which thanks
to a part of higher-velocity nodes hired to rebroadcast
the messages. These nodes do accelerate the propaga-
tion of messages.
International Journal of Distributed Sensor Networks
Figure 5. Mobility velocity versus packet delivery ratio.
As shown in Figure 5, the node mobility almost has
no impact on the packet delivery ratio of flooding.
Flooding always keeps its reliability at a high level,
even when nodes are at high velocity. But the packet
delivery ratio of the other schemes is affected in a nega-
tive way. The inherent deterministic broadcast mechan-
ism makes LBA scheme obtain the best packet delivery
ratio when the node mobility velocity keeps lower than
5 m/s and also makes it failed in high-velocity scenario.
For DCB scheme, the changing of the topology reduces
the accuracy of the threshold significantly, leading to
the serious declining in the packet delivery ratio. The
packet delivery ratio of NCPR, NKB, and NKVB
schemes is influenced gently by the node mobility,
because their rebroadcast does not rely on the accurate
neighbor knowledge. Compared with the other three
schemes, except for in low velocity scenario, NKB and
NKVB scheme obtain a better packet delivery ratio in
most scenarios. The simulation result above confirms
that probabilistic broadcast schemes are more insensi-
tive to node mobility than deterministic broadcast
schemes as mentioned before, since they only need to
maintain rough local topology knowledge.
Effects of node density
How the node density affects the running of the broad-
cast schemes is illustrated in this section. As shown in
Figure 6, at the very start, the normalized broadcast
overhead of all schemes decreases with the increasing
number of nodes. It is because the network connectivity
is being enhanced by the joined nodes, and then every
rebroadcast becomes more efficient to cover more fresh
nodes. However, the continued increasing nodes inevi-
tably bring about the collisions followed by retransmis-
sions, which leads to the increasing normalized
Lu and Dong 11

Figure 7. Number of nodes versus average end-to-end delay.

Figure 6. Number of nodes versus normalized broadcast
overhead.

broadcast overhead afterward. When the number of

nodes exceeds 200, the overhead curves of flooding,
LBA, and NCPR start a noticeable uptrend. The over-
head curves of NKB and NKVB schemes are relatively
flat for their self-adjusted rebroadcast probability upon
the network density. Compared with Gossip1(0.65) and
flooding, NKB and NKVB schemes save 49.2%–
67.5% overhead and 59.8%–75.8% overhead, respec-
tively, as shown in Figure 6. Compared with NCPR
scheme, NKB and NKVB schemes save 6.6%–41.4%
overhead.
Different from the other schemes, the normalized
broadcast overhead of DCB scheme almost decreases
with the increasing number of nodes all the time.
Because the value of threshold Cmax taken by DCB is
not big enough in dense scenario, so the messages are Figure 8. Number of nodes versus packet delivery ratio.
rebroadcasted with smaller probability, reducing the
collisions and the retransmissions. Accordingly, its
average end-to-end delay also decreases with the delay of flooding, compared with the other schemes, as
increasing number of nodes as shown in Figure 7. shown in Figure 7. But flooding still keeps a stable and
Unfortunately, the inaccurate threshold Cmax also leads high packet delivery ratio. In NCPR, NKB, and
directly to the lower packet delivery ratio, as shown in NKVB schemes, the node density is taken into consid-
Figure 8. eration when defining the rebroadcast probability and
Figures 7 and 8 show how the node density acts on the rebroadcast delay, so the changing node density has
the average end-to-end delay and the packet delivery less negative impact on their delay and delivery ratio.
ratio. As mentioned before, the overhead decreases In NKB and NKVB schemes, some special nodes such
with the increasing nodes in sparse network environ- as leaf node deserve a special treat in the process of
ment. The increasing nodes make the rebroadcast more broadcasting. Their rebroadcast parameters are defined
efficient, namely, lower delay and higher delivery ratio. more thoughtfully to deal with the node density than
In dense network environment, on the contrary, the the other schemes do, which makes both schemes more
increasing nodes incur more collisions and more reliable and efficient in complicated network environ-
retransmissions, increasing the end-to-end delay and ment. As shown in Figures 6–8, NKB and NKVB
deteriorating the delivery ratio. The increasing node schemes outperform the other schemes especially in
density has a more serious and negative impact on the average end-to-end delay by 88.9% at most.
12 International Journal of Distributed Sensor Networks

Figure 10. Number of CBR connections versus average end-

Figure 9. Number of CBR connections versus normalized
to-end delay.
broadcast overhead.

Effects of traffic load

How the traffic load affects the running of the broad-
cast schemes is illustrated in this section. As shown in
Figure 9, without exception, the normalized broadcast
overhead of all schemes increases with the growing traf-
fic load, since more traffic load leads to more conten-
tions and collisions. The high rebroadcast probability
adopted by flooding and Gossip1(0.65) brings about a
serious congestion, and the heavy traffic load aggra-
vates the congestion further. LBA scheme rebroadcasts
a great many redundant messages, so its overhead
experiences a noticeable increase as shown in Figure 9.
For DCB scheme, the main of the overhead originates
from the retransmission.
In NCPR scheme, the rebroadcast delay of a node is
inversely proportional to the common neighbors
between the upstream node and itself. Generally, it Figure 11. Number of CBR connections versus packet
means that the nodes with more neighbors will rebroad- delivery ratio.
cast the message more possibly after a smaller rebroad-
cast delay. Then, these nodes will forward more packets
in less time than the other nodes around. Namely, the rebroadcast delay and the rebroadcast probability in
way that the rebroadcast delay and the rebroadcast NKB and NKVB scheme. It makes the traffic load dis-
probability are defined is inclined to incur the traffic tribution relatively uniform and more reasonable. The
aggregation. The situation becomes worse as the traffic contentions and collisions are restrained effectively.
load increases. As shown in Figure 9, there is a remark- Therefore, the traffic load has less effect on NKB and
able overhead fluctuation arising when CBR connec- NKVB scheme than that on the other schemes.
tions of NCPR scheme rise from 50 to 60. At the same Compared with flooding, the propose schemes save
time, there is a remarkable increasing of the average 79.2% overhead and reduce 79.1% delay at most, as
end-to-end delay and an obvious decreasing of the shown in Figures 9 and 10.
packet delivery ratio as shown in Figures 10 and 11,
respectively.
Besides the ratio of uncovered neighbors and the Conclusion
neighbor density, the absolute number of uncovered We proposed NKB and NKVB schemes for wireless ad
neighbors is taken in consideration to calculate the hoc networks. In both schemes, the neighbor density
Lu and Dong
and coverage information are collected by Hello mes-
sage to calculate the rebroadcast parameters in near
real-time at each node. To reduce the overhead of
Hello message and upgrade the validity of neighbor
knowledge, piggybacking is used. One-hop neighbor
information collected keeps the control overhead at a
low level. The flexible calculation of the rebroadcast
probability and delay guarantees a low average end-to-
end delay as well as a high packet delivery ratio in com-
plicated mobile environment. Moreover, the way we
defined the rebroadcast delay and the rebroadcast
probability is beneficial to balance the traffic load and
improve the scalability of the schemes. In order to
reduce the latency further, few higher-velocity nodes
are employed to rebroadcast the received message in
NKVB scheme. It makes NKVB scheme more adapta-
ble to mobile environment. The simulation results show
that the proposed schemes outperform the existing
broadcast schemes in both reliability and overhead,
especially in latency.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: This work is supported by National Natural Science
Foundation (NNSF) of China (61300198 and 61070092), and
by Guangdong Natural Science Foundation (2015A030
308017) and Guangdong Science and Technology Key Project
(2015B010131009).
References
1. Reina DG, Toral SL and Johnson P. A survey on prob-
abilistic broadcast schemes for wireless ad hoc networks.
Ad Hoc Netw 2015; 25: 263–292.
2. Tseng YC, Ni SY, Chen YS, et al. The broadcast storm
problem in a mobile ad hoc network. Wirel Netw 2002;
8(2): 153–167.
3. Obraczka K, Viswanath K, Tsudik G, et al. Flooding for
reliable multicast in multi-hop ad hoc networks. Wirel
Netw 2001; 7(6): 64–71.
4. Jetcheva J, Hu Y, Maltz D, et al. A simple protocol for
multicast and broadcast in mobile ad hoc networks
(internet-draft), IETF MANET working group, 2001,
https://tools.ietf.org/html/draft-ietf-manet-simple-mbcast-01
5. Williams B and Camp T. Comparison of broadcasting
techniques for mobile ad hoc networks. In: Proceedings
of the 3rd ACM international symposium on mobile ad hoc
networking and computing, Lausanne, 9–11 June 2002,
pp.194–205. New York: ACM.
6. Ruiz P and Bouvry P. Survey on broadcast algorithms
for mobile ad hoc networks. ACM Comput Surv 2015;
48(1): 1–35.
13
7. Wang G, Lu D, Jia W, et al. Reliable gossip-based
broadcast protocol in mobile ad hoc networks. In: Pro-
ceedings of the first international conference on mobile ad-
hoc and sensor networks, Wuhan, China, 13–15 December
2005, pp.207–218. Berlin: Springer.
8. Haas ZJ, Halpern JY, Li L, et al. Gossip-based ad hoc
routing. IEEE ACM T Netw 2006; 14(3): 479–491.
9. Slavik M and Mahgoub I. Stochastic broadcast for
VANET. In: 7th IEEE consumer communications and net-
working conference, Las Vegas, NV, 9–12 January 2010,
pp.205–209. New York: IEEE.
10. Zhou L, Cui G, Liu H, et al. NPPB: a broadcast scheme
in dense VANETs. Inf Technol J 2010; 9(2): 247–256.
11. Lu D, Dong S, Tao M, et al. A velocity and neighbor
density-based broadcast scheme in mobile ad hoc net-
works. J High Speed Netw 2015; 21(3): 221–235.
12. Tseng YC, Ni SY and Shih EY. Adaptive approaches to
relieving broadcast storms in a wireless multihop mobile
ad hoc network. IEEE T Comput 2003; 52(5): 545–573.
13. Humoud S, Mackenzie L and Abdulai J. Neighbour-
hood-aware counter-based broadcast scheme for wireless
ad hoc networks. In: IEEE GLOBECOM workshops,
New Orleans, LA, 30 November–4 December 2008, pp.1–
6. New York: IEEE.
14. Mohammed A, Ould-Khaoua M, Mackenzie LM, et al.
Probabilistic counter-based route discovery for mobile ad
hoc networks. In: Proceedings of international conference
on wireless communications and mobile computing, Leip-
zig, 21–24 June 2009, pp.1335–1339. New York: ACM.
15. Yassein MB, Nimer SF and Al-Dubai AY. A new
dynamic counter-based broadcasting scheme for mobile
ad hoc networks. Simul Model Pract Th 2011; 19(1):
553–563.
16. Liarokapis D and Shahrabi A. Fuzzy-based probabilistic
broadcasting in mobile ad hoc networks. In: IFIP wireless
days, Niagara Falls, ON, Canada, 10–12 October 2011,
pp.1–6. New York: IEEE.
17. Liarokapis D, Shahrabi A and Komninos A. DibA: an
adaptive broadcasting scheme in mobile ad hoc networks.
In: Seventh annual communication networks and services
research conference, Moncton, NB, Canada, 11–13 May
2009, pp.224–231. New York: IEEE.
18. Gang W, Shigang W, Cai L, et al. Research and realiza-
tion on improved MANET distance broadcast algorithm
based on percolation theory. In: International conference
on industrial control and electronics engineering, Xi’an,
China, 23–25 August 2012, pp.96–99. New York: IEEE.
19. Kokuti A and Simon V. Adaptive multihop broadcast
protocols for ad hoc networks. In: 8th international sym-
posium on communication systems, networks & digital sig-
nal processing, Poznan, 18–20 July 2012, pp.1–6. New
York: IEEE.
20. Banerjee A, Foh CH, Yeo CK, et al. Performance
improvements for network-wide broadcast with instanta-
neous network information. J Netw Comput Appl 2012;
35(3): 1162–1174.
21. Slavik M and Mahgoub I. Spatial distribution and chan-
nel quality adaptive protocol for multihop wireless broad-
cast routing in VANET. IEEE T Mobile Comput 2013;
12(4): 722–734.
14
22. Mostafa A, Vegni AM and Agrawal DP. A probabilistic
routing by using multi-hop retransmission forecast with
packet collision-aware constraints in vehicular networks.
Ad Hoc Netw 2014; 14(3): 118–129.
23. Liu X, Yu Y, Zheng Y, et al. Distance and cooperation
based broadcast in wireless ad hoc networks. L N Inst
Comp Sci So 2017; 177: 168–178.
24. Bako B, Kargl F, Schoch E, et al. Advanced adaptive
gossiping using 2-hop neighborhood information. In:
Global telecommunications conference, New Orleans, LO,
30 November–4 December 2008, pp.1–6. New York:
IEEE.
25. Wang Q, Shi H and Qi Q. A dynamic probabilistic broad-
casting scheme based on cross-layer design for manets.
Int J Mod Educ Comput Sci 2010; 2(1): 40–47.
26. Wu C and Li C. Intelligent broadcast in wireless ad hoc
networks using live packet information. In: 9th interna-
tional conference on information, communications and sig-
nal processing, Tainan, China, 10–13 December 2013,
pp.1–5. New York: IEEE.
27. Zhang X, Wang E, Xia J, et al. A neighbor coverage-
based probabilistic rebroadcast for reducing routing over-
head in mobile ad hoc networks. IEEE T Mobile Comput
2013; 12(3): 424–433.
28. Banerjee S, Misra A, Yeo J, et al. Energy-efficient broad-
cast and multicast trees for reliable wireless communica-
tion. In: IEEE wireless communications and networking
conference, New Orleans, LA, 16–20 March 2003,
pp.660–667. New York: IEEE.
29. Ruiz P, Dorronsoro B, Bouvry P, et al. Information dis-
semination in VANETs based upon a tree topology. Ad
Hoc Netw 2012; 10(1): 111–127.
30. Chen F and Kao J. Game-based broadcast over reliable
and unreliable wireless links in wireless multihop net-
works. IEEE T Mobile Comput 2013; 12(8): 1613–1624.
31. Duc TL, Le DT, Zalyubovskiy VV, et al. Towards broad-
cast redundancy minimization in duty cycled wireless sen-
sor networks. Int J Commun Syst 2017; 30(6): e3108.
32. Liu H, Wan P, Jia X, et al. Efficient flooding scheme
based on 1-hop information in mobile ad hoc networks.
In: 25th IEEE international conference on computer com-
munications, Barcelona, 23–29 April 2006, pp.1–12. New
York: IEEE.
International Journal of Distributed Sensor Networks
33. Woon W and Yeung KL. Variable power broadcasting
based on local information for source-dependent broad-
casting protocols. In: IEEE wireless communications and
networking conference, Cancun, Mexico, 28–31 March
2011, pp.796–801. New York: IEEE.
34. Stojmenovic I, Khan A and Zaguia N. Broadcasting with
seamless transition from static to highly mobile wireless
ad hoc, sensor and vehicular networks. Int J Parall Emer
Distrib Syst 2012; 27(3): 225–234.
35. Schleich J, Thi H and Bouvry P. Solving the minimum
M-dominating set problem by a continuous optimization
approach based on DC programming and DCA. J Comb
Optim 2012; 24(4): 397–412.
36. Li D, Kim D, Zhu Q, et al. Minimum total communica-
tion power connected dominating set in wireless net-
works. In: 7th international conference on wireless
algorithms, systems, and applications, Yellow Mountains,
China, 8–10 August 2012, pp.132–141. Berlin: Springer.
37. Ros FJ, Ruiz PM and Stojmenovi I. Acknowledgment-
based broadcast protocol for reliable and efficient data
dissemination in vehicular ad hoc networks. IEEE T
Mobile Comput 2012; 11(1): 33–46.
38. Khabbazian M, Blake I and Bhargava V. Local broad-
cast algorithms in wireless ad hoc networks: reducing the
number of transmissions. IEEE T Mobile Comput 2012;
11(3): 402–413.
39. Haas ZJ and Nikolov MV. Towards optimal broadcast
in wireless networks. IEEE T Mobile Comput 2015; 14(7):
213–222.
40. Kok GX, Chow CO and Ishii H. Reducing broadcast
redundancy in wireless ad-hoc networks with implicit
coordination among forwarding nodes. Wireless Pers
Commun 2015; 81(1): 253–278.
41. Nowak S, Nowak M and Grochla K. Global queue prun-
ing method for efficient broadcast in multihop wireless
networks. Commun Comput Inf Sci 2016; 659: 214–224.
42. Xue F and Kumar PR. The number of neighbors needed
for connectivity of wireless networks Wirel Netw 2004;
10(2): 169–181.
43. Grossglauser M and Tse D. Mobility increases the capac-
ity of ad hoc wireless networks. IEEE ACM T Netw 2002;
10(4): 477–486.
44. NS2, http://www.isi.edu/nsnam/ns