Self-Pruning based Probabilistic Approach to Minimize Redundancy Overhead for Performance Improvement in MANET

International Journal of Computer Networks & Communications (IJCNC) Vol.13, No.2, March 2021
DOI: 10.5121/ijcnc.2021.13202 15
SELF-PRUNING BASED PROBABILISTIC APPROACH
TO MINIMIZE REDUNDANCY OVERHEAD FOR
PERFORMANCE IMPROVEMENT IN MANET
Gyanendra Kumar Pallai1
, Meenakshi Sankaran2
and Amiya Kumar Rath3
1
Department of Computer Science & Engineering,
SOA Deemed-to-be University, Bhubaneswar, India,
2
VIT Business School, VIT Bhopal University, Bhopal, India
3
Department of Computer Science & Engineering, Veer Surendra Sai University of
Technology, Burla. Deputed as adviser to National Assessment & Accreditation
Council, Bengaluru, India
ABSTRACT
The Broadcast storm problem causes severe interference, intense collision and channel contention, which
greatly degrades the QoS performance metrics of the routing protocols. So, we suggest a neighbourhood
coverage knowledge probabilistic broadcasting model (NCKPB) integrating with AODV protocol with
knowledge on 2-hop neighbourhood coverage; a connectivity function to control a node’s forwarding
probability of retransmission to alleviate significant overhead redundancy. Our objective is to minimize the
broadcast RREQ overhead while ensuring fair retransmission bandwidth. We considered two more
important measures called Saved Rebroadcast and Reachability. The outcomes of NCKPB, Fixed
probability (FP) and Flooding (FL) routing schemes are examined under three major operating conditions,
such as node density, mobility and traffic load. The NS-2 results demonstrate the efficacy of the proposed
NCKPB model by illustrating its performance superiority over all key metrics such as redundancy overhead,
end to end latency, throughput, reachability, saved rebroadcast and collision contrast to FP and FL.
KEYWORDS
Broadcasting, Flooding, Self-pruning, Neighbour Coverage, Forwarding Probability, Connectivity
Metric.
1. INTRODUCTION
MANET is a dynamically self-organized network of freely moving nodes without any fixed
infrastructure [1] [2]. Arbitrary node mobility in ad hoc networks leads to regular errors in
synchronisation among source and destination nodes, thus growing the number of redundancy
overheads that degrade the efficiency of the routing protocol. Broadcasting [10] is a typical method
in MANET for exploring paths. In this method each node sends a RREQ message among its
neighbouring nodes within the transmitting range. Thus the RREQ message passes through
intermediary nodes across the network in multi-hop mode. A traditional flooding scheme [10] [11]
[12] enforces broadcasting, wherein each node retransmit a RREQ message upon receiving the
broadcast message for the first time. Flooding executes at every node to ensure the message must
reach to every node in the network. This results in severe interference, excessive collisions and
channel contention within mobile wireless networks, known as the "broadcast storm problem” [10]
[11] [12]. Channel contention [10] happens when a node X retransmits a message it receives from
node Y when another node Z tries to retransmit at the same time. Then node Z needs to wait for a
random amount of time, since it presumes that the channel is busy due to node Y. Node X will

16
compete substantially with node Y's common shared link, which increases the data transmission
latency time. Message collision [10] or overlapping node transmission is more likely to occur
because there is no detection process in the simple flooding scheme. Say node X and Y, for
example, attempt to retransmit the message simultaneously results in collisions for which the
message could not reach any other node say Z.
Basically broadcasting mechanism for route discovery process in MANET reactive protocols are
conducted via blind flooding. The core issue is how to reduce the amount of rebroadcasts thereby
preserving reasonable latency for retransmission and reachability of packets. We understand that a
huge volume of duplicate rebroadcasts imply a high degree of reachability, but at the same time
they degrade the efficiency of the protocol due to bandwidth wastage and collisions. Conversely,
limited amount of retransmission results in poor reachability since it causes the retransmission chain
to be split in such a way that certain hosts are unable to receive the packets sent. Hence, how to
lessen the amount of rebroadcasts whilst retaining fair bandwidth, overhead and packet reachability
is a significant issue.
This research paper discusses our proposed scheme integrating with AODV protocol using
neighbour knowledge information (NCKPB). To reduce unnecessary retransmissions, our method
blends the benefits of probabilistic and neighbourhood information with a connectivity metric. The
rest of this study is arranged as follows: We discuss the relevant studies on MANET broadcasting
on Section 2. We discuss the MANET routing protocols and route discovery process of AODV
protocol in Section 3. Section 4 reviews detail of our proposed broadcast scheme NCKPB on
AODV and its algorithm. In Section 5 we simulate our method using NS-2 [9] simulator. The
results of the simulation and contrast its output with FL and FP. In section 6, we summarize the
paper and identify approaches for future investigation.
2. RELATED WORK
Numerous broadcasting solutions have been introduced in recent times to optimize the MANETs'
retransmission techniques. The B.Williams and Camp [13] description of broadcasting systems is
based on probability, area and neighbourhood knowledge. Basic flooding [3] [10] [12] [13] is a
simplistic broadcasting solution which is quick to enforce with assured transmission of messages.
Every host rebroadcasts with a predetermined probability P (P=1) [19]. Simplistic FL offers
adequate transmission but often causes redundancy, contention, and collision. MANET has
suggested several probabilistic [14] [16] [17] [18] [19] [20] solutions as a way to reduce redundant
transmission and mitigate the adverse impact of the issue of broadcast storm. Fixed probabilistic
[10] [15] is the most basic probabilistic strategy, from the basis of which all subsequent dynamic
probabilistic schemes are defined. In FP, the probability P of each host is generally assumed to be
between 0.65 and 0.7. M.Bani-Yassein et.al. [17] performed a detailed ns-2 simulation to research
the efficiency of optimised FP and flooding strategy. A technique that combines the dynamic
probabilistic and fixed probabilistic approach to decrease the RREQ overhead for different node
speed has been implemented in the work [20]. The probability of forwarding is dynamically
regulated by each node, based on their neighbourhood density level in the network has been studied
in [18] [19]. A mixture of probabilistic and counter-based method was suggested by Qi Zhang
et.al.[16] contrasted with flooding and set probabilistic schemes for different network conditions
such as network size, speed and CBR connections. The suggested DP-AODV solution yields less
retransmissions, greater reachability, higher throughput and minimal latency by dynamically
regulating a node’s probability of forwarding utilising its neighbouring density. Jae-soo kim et.al.
[23] presents a broadcast technique that incorporates a probabilistic and area-based method. The
receiver node calculates the additional range by estimating the distance from the sender using signal
intensity and dynamically controls the forwarding probability. In flooding with self-pruning [21]
proposed by Lim and Kim, the broadcasting node appends its one-hop neighbours to the RREQ

17
packet. The receiver node matches the list of neighbours from the received RREQ message with its
own surrounded neighbour. W. Peng et.al [22] suggested a scalable broadcast algorithm (SBA)
focused on neighbour information within a distance of two hops. Xin Zhang Ming et al. [24]
developed a probabilistic rebroadcast protocol (NCPR) for neighbourhood coverage. The scheme
used a balanced rebroadcast delay to estimate neighbourhood information along with additional
neighbour coverage and a connectivity function. A neighbour based routing protocol DCFP [25]
was developed and tested against NCPR and AODV protocols to minimise the redundant RREQ
packets. DCFP dynamically investigates the neighbourhood knowledge using novel network
connectivity metric and preset parameters. An enhanced self-pruning broadcasting (ISB) algorithm
[26] is suggested, and its output is evaluated with dominant and other self-pruning schemes under
different transmission ranges utilising three-hop neighbour knowledge information. Huaqiang Xu
et.al [29] explored a probabilistic broadcast scheme TPB centred on nodes trust level. Dingzhu Lu
et.al.[27] has introduced a two neighbour knowledge based and velocity based schemes namely
NKB and NKVB. The efficiency of two systems is measured against other current broadcasting
structures for differing node size, maximum speed and CBR load. A density-aware probabilistic
strategy to alleviate broadcast storm effects for content-centric vehicular networks have been
suggested in the work [28] and its performance is compared with other related approaches.
3. ROUTING PROTOCOLS IN MANET
In order to overcome routing challenges in MANETs, many routing protocols [2] have recently
been studied. These protocols were categorised under three strategically defined routing
classifications [2] [3]: proactive, reactive, and hybrid. Table-driven routing protocols [5] [30] [31]
are considered proactive protocols, constantly keeping the updated information for whole network
in a routing table (RT). The greatest benefit of proactive protocols is that with minimal end-to - end
delay it transmits data immediately. However, Periodic update of the routing table (RT) raises the
overhead as compared to the reactive protocol. The Destination Sequence Distance Vector (DSDV)
and Optimized link state Routing (OLSR) is of proactive category. The reactive ones [3] [4] [7]
[30] [31] are an enhancement to table-driven routing that uses the method of route discovery only
when the source node desires it. Unlike proactive protocols, reactive protocols only hold knowledge
regarding active paths. Compared with the proactive category, this results in less redundancy
overhead. Ad hoc On-Demand Distance Vector Routing (AODV) and Dynamic Source Routing
(DSR) are two basic reactive MANET protocols. The Hybrid [6] incorporates the scalability benefit
of the reactive protocol plus proactive protocol performance. For instance, Zone Routing Protocol.
Hybrid protocols split their neighbouring nodes into two domains. One called the "intra-zone" and
uses reactive protocol, while the other designated "inter-zone" investigates by proactive protocol.
3.1. Route Discovery Mechanism in AODV
AODV has gained a considerable interest owing to its simplicity and low processing overhead. In
this scheme, a fresh path is established by a source node on demand. Unlike proactive, it does not
update its routing table (RT) periodically and thus incurs less control overhead. Any source node
has some packets to deliver to its destination , it first searches the path availability with its own
routing table (RT) [3] [7]. If the path is available in the RT, it literally follows the path to deliver the
message otherwise it invokes route discovery mechanism [3] [7]. First, the source broadcasts the
RREQ notification to all of its neighbourhood nodes. Any mobile host that receives the broadcast
message increases its hop count and builds a reversed route (for source and neighbour node) with an
update to RT. A unicast RREP message is sent to source either from destination or any node that
has fresh path information; otherwise the node propagates the RREP notification to all the
neighbours around it. The node receives an RREP message, it immediately generates the forward
route entry for the node (i,e. destination or intermediary node) RREP message was made. Using the

18
reverse path entry, the node passes the RREP message until it reached to source node. If the source
node obtains multiple RREPs along separate paths, the path with the highest destination sequence
number is selected.
4. PROPOSED SCHEME - NEIGHBOUR COVERAGE KNOWLEDGE
PROBABILISTIC BROADCASTING PROTOCOL (NCKPB)
The main purpose and motivation of our proposed scheme is to find a way of alleviating broadcast
storm problems that cause extreme collision and channel contention and optimise the performance
of routing protocols across all performance metrics. The suggested methodology is a reactive
approach to self-pruning broadcasting. A mobile host makes a decision on rebroadcast using 2-hop
neighbourhood knowledge with redundant information to minimize the redundancy overhead. For
explicit exploration of neighbourhood knowledge, every host shares "hello" packets periodically. If
the estimated time interval for rebroadcasting is short, it contributes to excessive collision or
contention. Moreover, if the neighbouring nodes of a node have been covered by previous
broadcasts, a node should not retransmit the packet.
The suggested NCKPB broadcast strategy is structured in four steps:
 Estimation of a reasonable rebroadcast delay for a node.
 Acquisition of neighbouring information to determine additional coverage area.
 A connectivity function to balance the amount of retransmissions.
 Regulating the probability of retransmissions based on the additional neighbour coverage
fraction with connectivity function.
4.1. Rebroadcast Delay
In order to optimise the neighbourhood coverage of the node effectively, it is important to
identify a reasonable delay in rebroadcasting. A node will be given a lower retransmission if there
were a large amount of similar neighbours instead we would set a higher limit.
The delay ratio Td-ratio of nr is calculated is as follows:
Td-ratio (nr) = (1)
Rebroadcast delay Trb-delay is estimated using max delay constant “ , we choosen 0.1 sec.
Trb-delay (nr) = △ x Td-ratio (nr) (2)
4.2. Additional Coverage Ratio
A receiver node nr accepts an RREQ from sender ns , the recipient nr compares the sender list to
its own neighbours. The receiving node nr would then compile an uncovered list of neighbours.
The prepared list contains all neighbour nodes of the receiver not covered by the sender node.
This uncovered neighbour list gives an approximation for estimating the extra coverage ratio of
the receiver. This means, if the receiving node nr rebroadcasts the packet, it will reach more
adjacent nodes.

19
The additional coverage ratio Acov-ratio of nr is as follows:
Acov-ratio (nr) = (3)
4.3. Connectivity Metric
We evaluate a connectivity metric Cmetric on basis of average amount of neighbours in the network
to control the quantity of RREQ retransmissions and computed as:
Navg-nb(nr) = (4)
N - Total number of network nodes, R - Transmission radius, A - Topology size, and
K - Connectivity constant (1.02).
Cmetric(nr) = (5)
4.4. Forwarding Probability
The forwarding probability Pfw is a product of connectivity metric Cmetric with additional coverage
ratio Acov-ratio and calulated as:
Pfw(nr) = Cmetric (nr) x Acov-ratio (nr) (6)
Measurement of the forwarding probability Pfw with a random value (0 to 1) provides knowledge
of the node's neighbour density (ie. dense or sparse). If the node’s forwarding probability is lower
than the random value, this signifies that node is found in dense region. To minimize number of
redundancy overhead, we must discard duplicate RREQ messages. However, if forwarding
probability is bigger than random value, then it is found in the sparse zone and the message needs
to be rebroadcasted.
Steps involved in our proposed scheme:
1. Each node sends its recent neighbouring list in the RREQ message. Upon
receiving themessage from sender ns, the receiver nr compares the sender list to
its own neighbours list.
If N (nr) ⊆ [N (ns) ⋃ { ns }] Cancel Rebroadcast (Duplicate msg)
2. Otherwise, RREQ msg is received for the first time, nr prepares a covered neighbour set
Bcov-set (nr, msg) = N (ns) ⋃ { ns }
3. Rebroadcast delay Trb_delay using equation (2) for nr is calculated.
4. Node nr updates the covered neighbour set for each msg and discards the duplicate
Bcov-set (nr, msg) = Bcov-set (nr, msg) ⋃ [N (ns) ⋃ { ns } ]
5. Rebroadcast decision is made after the delay expires; If the Bcov-set (nr, msg)
covers all theneighbours of nr, then the receiver nr not to rebroadcast
If N (nr) ⊆ Bcov-set (nr, msg) Cancel rebroadcast
6. Otherwise, additional coverage ratio Acov-ratio (nr) is computed using eq (3).
7. Neighbour density Navg-nb and connectivity metric Cmetric estimated using eq (4) and
(5).

20
8. Forwarding probability Pfw of nr is evaluated using eq (6).
9. Finally, the rebroadcast probability Pfw is compared against random value
ranging between0 and 1 for determining RREQ retransmissions.
4.5. Algorithm - Neighbour Knowledge Probabilistic Broadcasting (NCKPB)
Definitions: msg - RREQ message, N(nr) - Neighbour set of nr, N(ns) - Neighbour
set of ns,Bcov-set(nr, msg) - Broadcast cover set of node nr for msg.
1. When a node nr receives a RREQ msg from ns:
2. If N(nr) ⊆ [ N(ns) ⋃ { ns } ] Cancel Rebroadcast
3. If (RREQ msg received for the first time)
4. Bcov-set(nr,msg) = N(ns) ⋃ { ns }
5. Calculate Rebroadcast delay:
6. Td-ratio(nr) =
7. Trb-delay(nr) = △ x Td-ratio(nr)
8. Schedule the rebroadcast timer for Trb-delay(nr)
9. End if
10. While (nr receives duplicate msg during the timer period)
11. Update the broadcast cover set:
12. Bcov-set(nr,msg) = Bcov-set(nr,msg) ⋃ [ N(ns) ⋃ { ns } ]
13. Discard msg
14. End while
15. If (Timer for rebroadcasting node nr expires)
16. If N(nr) ⊆ Bcov-set(nr,msg) Cancel rebroadcast
17. Else
18. Calculate the coverage ratio:
19. Acov-ratio(nr) =
20. Calculate average network neighbour density:
21. Navg-nb(nr) =
22. Calculate connectivity factor:
23. Cmetric(nr) =
24. Calculate forwarding probability:
25. Pfw(nr) = Cmetric(nr) x Acov-ratio(nr)
26. If random (0,1) ≤ Pfw(nr)
27. Rebroadcast msg
28. Else
29. Discard msg
30. End if

21
5. PERFORMANCE ANALYSIS
5.1. Simulation Scenarios and Parameters
We have made an effort to examine and assess the effectiveness of our proposed NCKPB model for
an approach to FL and FP. All routing schemes are investigated under three main operating
scenarios of the MANET network, such as speed, density and load. Mobile nodes are uniformly
dispersed across 1000 x 1000 metres of geographical region with random way point mobility model
(RWP) [8]. For each network node the probability of retransmission varies from 0 to 1. The
simulation period for all three scenarios is set at 900 seconds. First, we performed detailed
simulation to study the density impacts, where the amount of nodes ranging from 75 to 200 nodes.
The second scenario analysis explores the effect of high speed on three separate routing protocols.
150 numbers of mobile hosts are distributed uniformly to study the influence on the mobility. The
mobility of the node scales from 5 to 30 m/sec with zero pauses. The third aspect of network
analysis discusses the outcomes of the traffic loads provided. 125 node numbers are distributed
uniformly to evaluate the effect of load differences from 5 to 25 CBR connections. The node speed
chosen is 10 m/sec with zero pause time for both traffic load and node density scenarios. Maximum
of 10 CBR links are arbitrarily chosen for traffic load.
Table 1. Simulation parameters
Simulation Parameter Value
Simulator NS2
MAC protocol 802.11
Topology size 1000 m x 1000 m
Radio propagation model Two-ray ground
Node’s transmission radius 250 m
Bandwidth 2 mbps
Interface queue length 50 packets
Mobility model Random way point
Traffic type Constant Bit Rate (CBR)
Packet length 512 bytes
Packet transmission rate 4 packets/second
Number of Nodes 75, 100, 125, 150, 175 , 200
Maximum node’s speed 5, 10, 15, 20, 25, 30 m/s
Number of traffic flows 5, 10, 15, 20, 25
Pause time 0 second (Continuous mobility)
Simulation time 900 seconds
5.2. Performance Metrics
We have assessed three different broadcasting schemes using following performance metrics:

22
5.2.1. Throughput
Throughput measures the amount of successful data packets that are transmitted in a given amount
of time over a communication channel.
5.2.2. End-to-End delay (ENE)
Average delay is the amount of time the data packet spends to reach at the destination.
5.2.3. RREQ overhead
It defines the amount of route request (RREQ) packets propagated over the simulation period.
5.2.4. MAC collision
This metric determines the number of data packets (CBR) which are dropped in collisions per
second over the MAC layer.
5.2.5. Reachability (RE)
It is the fraction of the nodes received the broadcasted RREQ packets over the total amount of
nodes in the network.
5.2.6. Saved Rebroadcast (SRB)
This metric calculates the ratio of RREQ packets received minus transmitted over the total amount
of RREQ packets received in the network.

23
5.3. Performance Results
5.3.1. Effects of Network Density
5.3.1.1. Routing Overhead
The effect of node densities on RREQ packet retransmission is plotted in Figure 1 for three
protocols. During the simulation study, the redundancy overhead in the network reflects the amount
of RREQ messages produced and propagated. Performance outcome of overhead routing shows
that NCKPB surpasses the other two schemes, while FP outperforms FL. Figure 1 shows that as the
network density grows from 75 to 200 numbers of nodes, all the protocols severely suffers with
increased number of RREQ overheads. Overall analysis clearly indicates that the volume of RREQ
packets transmitted across the network for NCKPB is lowered by 35% and 56% contrasts to FP and
FL.
Figure 1. Routing Overhead vs. Number of Nodes
5.3.1.2. Collision
The effect of the MAC collision on all three routing schemes is seen in Figure 2. With the increase
of node density from 75 to 150 numbers of nodes, the number of collisions increases significantly.
It is obvious because the average percentage of RREQ overhead retransmissions substantially
grows higher with each of the routing protocol. Ultimately, the individual node's bandwidth gets
impaired and ultimately raises the risk of packet loss and drop. The NCKPB protocol demonstrates
its superiority over the other two methods of routing. The simulation findings specifically indicate
that the overall collision rate for NCKPB is reduced by 33% and 53% compared to FP and FL.

24
Figure 2. MAC Collision vs. Number of Nodes
5.3.1.3. End to End Delay
Figure 3 reflects the delay performance of three distinct routing schemes with increased network
density. It is observed that average delay gets increased with the increased network density. The
reason for the increased delay is attributed primarily to excessive collisions, channel contention and
the amount of retransmissions in the MAC layer due to unwanted drop. This raises the delay in
reaching the destination of the CBR packets. Enhanced routing efficiency with lower transmission
of RREQ packets increases delay performance. The simulation results indicate that compared to the
other two variants, the volume of broadcast retransmissions is considerably smaller with the
proposed NCKPB model and has proved to be a better routing mechanism. This effectively ensures
the efficiency of the NCKPB algorithm by decreasing the delay to 25% and 35% contrast to FP and
FL.
Figure 3. End-to-End Delay vs. Number of Nodes
5.3.1.4. Throughput
The findings in Figure 4 depict the average throughput performance among all the routing protocols
with node density. Results clearly reveal that the average throughput of the network decreases with
increased network density. It is obvious, because higher number of redundancy control overheads
contributes to excessive network congestion. Probability of packets collisions and contention of the

25
channel severely affects throughput performance of all the protocols. The average range of
throughput performance lies between 17 kbps to 16.78 kbps for NCKPB, 16.95 kbps to 17.75 kbps
for FP and 16.91 kbps to 16.7 kbps for FL with the differences in network density. The proposed
model NCKPB shows its superiority over throughput performance due its combined advantage of
neighbour knowledge and controlled transmission of broadcasting. Outcome results specifically
reveal that the total average throughput performance achieved by NCKPB, FP and FL is 17.03
kbps, 16.83 kbps and 16.78 kbps respectively.
Figure 4. Throughput vs. Number of Nodes
5.3.1.5. Reachability (RE)
Figure 5 reveals the influence of various node densities on the simulation's reachability. From the
results of the simulation it is found that comparison to FP and FL, NCKPB performs well without
compromising reachability. Simulation results explicitly indicate that the ratio of nodes obtaining
the RREQ message across the network is growing steadily with node density. NCKPB, FP and FL
have an average reachability performance of 97.6 %, 95.4 %, and 99.6 % respectively. It is obvious
for the proposed NCKPB that the retransmission probabilities are dynamically regulated by each
network node resulting in improved results compared to FP.
Figure 5. Packet Reachability vs. Number of Nodes

26
5.3.1.6. Saved Rebroadcast (SRB)
Figure 6 demonstrates the FP and FL saving rebroadcast performance over proposed NCKPB
routing algorithm. This saved rebroadcast metric measures the proportion of duplicate messages
that can be saved from the total network RREQ generated. The higher the amount of rebroadcast
saved, the better the protocol performance. However, sending very few rebroadcast packets induces
poor reachability, which eventually contributes to a lack of protocol functionality. The number of
broadcast transmissions substantially grows with increased density. NCKPB's saved re-broadcast
performance improves linearly whereas no improvement is found either in blind flooding or fixed
probabilistic scheme. Because of NCKPB’s dynamic neighbouring information probability
broadcast scheme, it tends to save redundancy broadcasting at low density by around 40 % and
gradually saves up to 60 % in higher node density networks. The total percentage of NCKPB's
saved rebroadcast packets is approximately more than 52% of the total RREQ generated and
transmitted over the network.
Figure 6. Saved Rebroadcast vs. Number of Nodes
5.3.2. Effects of Node Mobility
Figure 7 exhibits the performance measure of RREQ overheads on three separate routing variants
against node mobility. Overhead routing transmissions gradually increases with every increase in
node speed (5 to 20 m/s). In contrast, an increase in the mobility beyond 20 m/s (i.e. 25 and 30 m/s)
is reported to have a positive effect on all the routing schemes. Because higher node speed
significantly reduces the path length between the nodes. The connectivity becomes stronger and the
probability of link failures and its repair gradually decreases. It is observed that our proposed model
NCKPB works better in high speed network. The experimental findings specifically demonstrate
that the average routing overhead of NCKPB is greatly decreased by 30% and 42% relative to FP
and FL schemes.

27
Figure 7. Routing overhead vs. Maximum Speed for 150 nodes
5.3.2.2. Collision
The simulation result in Figure 8 demonstrates the node mobility impact of routing protocols on
packet collisions. Findings clearly represents that the rate of collisions gradually rises due to rapid
increase in overhead transmissions. With low to moderate node mobility (5 to 20 m/s), the path
length gets increased between source to destination pairs. And eventually, the network struggles
from factors such as severe interference and strong congestion. However, higher node mobility
causes a favourable impact on protocol routing. The redundancy broadcast retransmissions are
greatly reduced due to the dynamic control mechanism of the NCKPB, thereby increasing the
performance of the protocols for all the metrics considered. Our proposed NCKPB eliminates
packet collisions by more than 29% and 40% relative to FP and FL.
Figure 8. MAC Collision vs. Maximum Speed for 150 nodes
The simulation findings in Figure 9 display the delay performance of NCKPB against other two
variants with respect to speed. It is clearly evident that the average delay parameter rises with rise in
node mobility. The delay increased by 22% for NCKPB, 35% for FP and 45% for FL with every
increase of speed (5 - 20 m/s). The reason of such delays is attributed to a variety of factors, such as

28
delays during path compilation, packet queuing times, and propagation and transmission delays to
the MAC layer. Higher collision and packet loss, all these factors imply larger delays. With higher
node speed (i.e. 25 and 30 m/s), the node moves faster towards the outer node, thereby reducing the
length of the route between the intermediate nodes. A secure network is less susceptible to
connectivity loss and thus creates less control overheads. As a result, network queue data packets
do not have a long wait period to reach the target. The proposed NCKPB model reveals its
superiority in reducing delay performance by approximately more than 30% and 42%, respectively
compared to FP and FL.
Figure 9. End-to-End Delay vs. Maximum Speed for 150 nodes
5.3.2.4. Throughput
The simulation results illustrated in Figure 10 indicate throughput over mobility for three routing
protocols. Outcome indicates that the performance of the throughput over all protocols is slightly
decreased from 5 to 20 m/s for each increase in speed. Huge amount of overhead controls to create
a precise routing table (RT) that essentially affects the bandwidth and delay of the individual nodes.
At higher node mobility (25 m/s), all protocols function well, enabling greater network
connectivity. NCKPB, FP and FL schemes have an optimum throughput performance of 13.05
kbps, 12.96 kbps and 12.91 kbps, respectively, at a speed of 25 m/s. The efficiency of the
throughput really depends on the delay and channel bandwidth. The findings obtained explicitly
indicate that NCKPB is dominant in terms of the throughput rate compared to others. The average
throughput rate over simulation for NCKPB is 12.93 kbps, for FP 12.9 kbps and for FL 12.85 kbps.

29
Figure 10. Throughput vs. Maximum Speed for 150 nodes
Figure 11 displays the findings with respect to reachability for all routing protocols against node
mobility. Specifically, the results indicate that performance in terms of reachability is marginally
compromised, irrespective of routing schemes. The flood strategy ensures all mobile hosts receive
packets sent at the risk of enhanced traffic generated by redundant RREQs. Conversely, redundant
messages even result in higher collisions, which ultimately cause immense message drops, thereby
directly impacting reachability. At a lower speed network, each node has a higher level of
neighbouring nodes. The network is more influenced by overhead control, collisions and additional
delays. The outcome over the entire simulation shows that the NCKPB reachability is 96 %, FP is
97.8% and FL is 99% respectively.
Figure 11. Packet Reachability vs. Maximum Speed for 150 nodes

30
Figure 12 indicates the average rebroadcasting performance of NCKPB compared to FP and FL
with varying speeds. The lower speed network is more liable to stagnate and the amount of adjacent
nodes per node is greater. The generation and propagation of overhead routing is higher to ensure
the efficient data communications to the destination. The NCKPB model reached a ratio of more
than approximately 50% SRB from 5 to 20 m/s of node mobility. In contrast, faster speeds (i.e. 25 -
30 m/s) have a predictably favourable impact on the efficiency of the proposed model. Greater
speed ensures substantially lower overhead routing when the length of the route from source to
destination is shortened. Higher speed network experiences less overhearing between participating
nodes that minimises congestion and improves efficiency. The overall saving of rebroadcast for
NCKPB is more than 52%, whereas no improvement is found either in FL or FP scheme.
Figure 12. Saved Rebroadcast vs. Maximum Speed for 150 nodes
5.3.3. Effects of Traffic Load
Findings represented in Figure 13 presents the RREQ overhead incurred by NCKPB, FP and FL for
5 to 25 flows of CBR connections. Experimental outcomes on traffic load reveal that NCKPB has
distinct performance gain compared to other two variants. More fresh paths are required when the
number of source-destination pairs have been increased. It triggers a proportionate number of new
path discoveries to carry the CBR packets to their destination. As a response, enormous amount of
RREQ packets are produced that have a negative impact on all performance metrics. In overall, the
total volume of redundant RREQ retransmissions of NCKPB decreased by around 25% and 50%
relative to FP and FL for variable traffic load.

31
Figure 13. Routing Overhead vs. Traffic Load for 125 nodes
5.3.3.2. Collision
In Figure 14, the effect of traffic load on packet collisions is demonstrated. Subsequent results of
simulations revealed that for all different loads from 5 to 25 (Max connections), NCKPB followed
by FP performs better than FL. Compared to other two variants, NCKPB demonstrates superiority
over collision rate among other two protocols. This is attributed to the successful framework of the
proposed model with a fair retransmission delay that dynamically governs the retransmission
probability of nodes. The simulation result clearly illustrates that with all connection pairs, NCKPB
beats collision reduction by 26% and 33% compared to FP and FL.
Figure 14. MAC Collision vs. Traffic Load for 125 nodes
NCKPB’s average latency performance, contrasting with FP and FL with varying traffic loads has
been illustrated in Figure 15. With the increased number of maximum connections, the delay
measures have been significantly impacted. It is clear, as the overhead redundancy is significantly
enhanced with traffic load variations. These redundant RREQ packets overload the network queue,

32
collisions and channel contentions. The network still struggles from longer delays from source to
destination. Such overhead redundancy involves the transfer of packets suffering from elevated
delay for extra time in the network queue. The NCKPB incurred the lowest overall end to end delay
of 30% and 44% respectively relative to FP and FL.
Figure 15. End-to-End Delay vs. Traffic Load for 125 nodes
5.3.3.4. Throughput
The findings in Figure 16 indicate that throughput performance is fairly comparable for all the
routing protocols. The increase in average throughput of all the distinct routing variants increases
with increase in connection pairs (5 to 25 max connections). The throughput performance of
NCKPB dominates over other two variants. NCKPB reaches optimum throughput of 29.5 kbps,
28.27 kbps for FP and 28.72 kbps for FL, respectively at the traffic load of 25 CBR connections.
The average network throughput rating of NCKPB is 18.5 kbps, 18.35 kbps for FP and 18.21 kbps
for FL at low and high traffic loads (i.e. 5-25 source-destination pairs).
Figure 16. Throughput vs. Traffic Load for 125 nodes

33
Figure 17 illustrates the effect on the reachability of the simulation of different traffic loads.
Flooding system ensures that any node in the network receives packets at the cost of redundant
overheads to reach optimum connectivity. However, the finding indicates that the reachability
performance of simple flooding is also compromised by distinct traffic flows. In contrast, our latest
NCKPB protocol, on the other side, is a good way to retain broad reach. The NCKPB's reachability
improves with more nodes. The average percentage of reachability attained by NCKPB is 94.5%,
FP is 96% and FL is 97%.
Figure 17. Packet Reachability vs. Traffic Load for 125 nodes
Figure 18 demonstrates the effect of traffic load variations on saved rebroadcasts for all variants.
Traffic loads vary from 5 to 25 CBR source-destination pairs. With increased source-destination
pairs, each coverage region has a higher number of retransmission candidates. Proportionate
volumes of RREQ packet transfers are required for new route discoveries. Simulation findings
clearly indicate that the performance of NCKPB with respect to SRB metrics continues to improve
with each increase in traffic load, with the exception of a marginal loss of performance at 20 CBR
connections. However, better performance improvements are found when the traffic flow is heavier
(25 CBR connections). Contrast to other two variants, our proposed scheme NCKPB has its
dominance and the average ratio of saved rebroadcasts is more than 45% over the generated and
retransmitted overhead RREQ in the network.

34
Figure 18. Saved rebroadcast vs. Traffic Load for 125 nodes
6. CONCLUSIONS
Extensive simulations using NS-2 are performed to analyze the three routing schemes NCKPB, FP
and FL in various network scenarios such as network densities, traffic loads and mobility. Our main
goal of this research analysis is to develop a model that blends the advantages of knowledge of
neighbourhood coverage with the probabilistic method. The proposed scheme bypasses the
conventional way of route discovery process usually found in AODV. Excessive redundant RREQ
retransmissions and collisions are significantly reduced in broadcasting using the following
procedures: estimating a reasonable rebroadcast delay for a node, acquisition of neighbouring
information to determine additional coverage area, a connectivity function to regulate the
probability of retransmissions. The experimental findings show that in all network situations,
NCKPB proves its dominance over two other variants to minimize the diffusion storm problems.
The simulation outcome demonstrates that our well-designed model NCKPB greatly reduce the
RREQ overhead with tremendous enhancement in terms of QoS performance metrics such as MAC
collision, throughput, delay, reachability and saved rebroadcast. In the further scope of our research
study, we intend to implement a more robust adaptive probabilistic scheme, taking into account an
optimal path selection using cost function of various routing metrics that guarantees a committed
level of QoS.
CONFLICTS OF INTEREST
The authors declare no conflict of interest.
REFERENCES
[1] Murthy, C. S. R., & Manoj, B. S. (2004). Ad hoc wireless networks: Architectures and protocols, portable
documents. Pearson education.
[2] Toh, C. K. (2001). Ad hoc mobile wireless networks: protocols and systems. Pearson Education.
[3] Perkins, C., Belding-Royer, E., & Das, S. (2003). RFC3561: Ad hoc on-demand distance vector (AODV)
routing..
[4] Johnson, D., Hu, Y. C., & Maltz, D. (2007). The dynamic source routing protocol (DSR) for mobile ad hoc
networks for IPv4 (Vol. 260). RFC 4728.
[5] Perkins, C. E., & Bhagwat, P. (1994). Highly dynamic destination-sequenced distance-vector routing
(DSDV) for mobile computers. ACM SIGCOMM computer communication review, 24(4), 234-244.
[6] Haas, Z. J., Pearlman, M. R., & Samar, P. (1999). Zone routing protocol for ad-hoc networks. Internet
Draft, draft-ietf-manet-zrp-02. txt, work in progress.

35
[7] Das, S. R., Perkins, C. E., & Royer, E. M. (2000, March). Performance comparison of two on- demand
routing protocols for ad hoc networks. In Proceedings IEEE INFOCOM 2000. Conference on Computer
Communications. Nineteenth Annual Joint Conference of the IEEE Computer and Communications
Societies (Cat. No. 00CH37064) (Vol. 1, pp. 3-12). IEEE.
[8] Camp, T., Boleng, J., & Davies, V. (2002). A survey of mobility models for ad hoc network research.
Wireless communications and mobile computing, 2(5), 483-502.
[9] The NS Manual (Formerly NS Notes and Documentation) (2011) The VINT Project: A Collaboration
between Researchers at UC Berkeley, LBL, USC/ISI, Xerox PARC. 5 November.
[10] Tseng, Y. C., Ni, S. Y., Chen, Y. S., & Sheu, J. P. (2002). The broadcast storm problem in a mobile ad hoc
network. Wireless networks, 8(2-3), 153-167.
[11] Tseng, Y. C., Ni, S. Y., & Shih, E. Y. (2003). Adaptive approaches to relieving broadcast storms in a
wireless multihop mobile ad hoc network. IEEE transactions on computers, 52(5), 545-557.
[12] Lim, H., & Kim, C. (2001). Flooding in wireless ad hoc networks. Computer Communications, 24(3-4),
353-363.
[13] Williams, B., & Camp, T. (2002, June). Comparison of broadcasting techniques for mobile ad hoc
networks. In Proceedings of the 3rd ACM international symposium on Mobile ad hoc networking &
computing (pp. 194-205).
[14] Sasson, Y., Cavin, D., & Schiper, A. (2003, March). Probabilistic broadcast for flooding in wireless mobile
ad hoc networks. In 2003 IEEE Wireless Communications and Networking, 2003. WCNC 2003. (Vol. 2,
pp. 1124-1130). IEEE.
[15] Reina, D. G., Toral, S. L., Johnson, P., & Barrero, F. (2015). A survey on probabilistic broadcast schemes
for wireless ad hoc networks. Ad Hoc Networks, 25, 263-292.
[16] Zhang, Q., & Agrawal, D. P. (2005). Dynamic probabilistic broadcasting in MANETs. Journal of parallel
and Distributed Computing, 65(2), 220-233.
[17] Bani-Yassein, M., Ould-Khaoua, M., Mackenzie, L. M., & Papanastasiou, S. (2006). Performance analysis
of adjusted probabilistic broadcasting in mobile ad hoc networks. International journal of wireless
information networks, 13(2), 127-140.
[18] Hanashi, A. M., Siddique, A., Awan, I., & Woodward, M. (2009). Performance evaluation of dynamic
probabilistic broadcasting for flooding in mobile ad hoc networks. Simulation Modelling Practice and
Theory, 17(2), 364-375.
[19] Abdulai, J. D., Ould-Khaoua, M., Mackenzie, L. M., & Mohammed, A. (2008, June). Neighbour coverage:
A dynamic probabilistic route discovery for mobile ad hoc networks. In 2008 International Symposium on
Performance Evaluation of Computer and Telecommunication Systems (pp. 165-172). IEEE.
[20] Yassein, M. B., Khaoua, M. O., Mackenzie, L. M., Papanastasiou, S., & Jamal, A. (2006, October).
Improving route discovery in on-demand routing protocols using local topology information in MANETs.
In Proceedings of the ACM international workshop on Performance monitoring, measurement, and
evaluation of heterogeneous wireless and wired networks (pp. 95- 99).
[21] Lim, H., & Kim, C. (2000, August). Multicast tree construction and flooding in wireless ad hoc networks.
In Proceedings of the 3rd ACM international workshop on Modeling, analysis and simulation of wireless
and mobile systems (pp. 61-68).
[22] Peng, W., & Lu, X. C. (2000, August). On the reduction of broadcast redundancy in mobile ad hoc
networks. In 2000 First Annual Workshop on Mobile and Ad Hoc Networking and Computing. MobiHOC
(Cat. No. 00EX444) (pp. 129-130). IEEE.
[23] Kim, J. S., Zhang, Q., & Agrawal, D. P. (2004, November). Probabilistic broadcasting based on coverage
area and neighbor confirmation in mobile ad hoc networks. In IEEE Global Telecommunications
Conference Workshops, 2004. GlobeCom Workshops 2004. (pp. 96-101). IEEE.
[24] Zhang, X. M., Wang, E. B., Xia, J. J., & Sung, D. K. (2012). A neighbor coverage-based probabilistic
rebroadcast for reducing routing overhead in mobile ad hoc networks. IEEE transactions on mobile
computing, 12(3), 424-433.
[25] Ejmaa, A. M. E., Subramaniam, S., Zukarnain, Z. A., & Hanapi, Z. M. (2016). Neighbor-based dynamic
connectivity factor routing protocol for mobile ad hoc network. IEEE Access, 4, 8053- 8064.
[26] Rab, R., Sagar, S. A. D., Sakib, N., Haque, A., Islam, M., & Rahman, A. (2017). Improved self- pruning
for broadcasting in ad hoc wireless networks. Wireless Sensor Network, 9(2), 73-86.
[27] Lu, D., & Dong, S. (2017). A neighbor knowledge and velocity-based broadcast scheme for wireless ad
hoc networks. International Journal of Distributed Sensor Networks, 13(11), 1550147717743699.
[28] Tizvar, R., & Abbaspour, M. (2020). A density-aware probabilistic interest forwarding method for content-
centric vehicular networks. Vehicular Communications, 23, 100216.
[29] Xu, H., Si, H., Zhang, H., Zhang, L., Leng, Y., Wang, J., & Li, D. (2020). Trust-Based Probabilistic
Broadcast Scheme for Mobile Ad Hoc Networks. IEEE Access, 8, 21380-21392.

36
[30] Pallai, G. K., Meenakshi, S., Rath, A. K., & Majhi, B. (2015). Comparative Study of On- Demand and
Table-Driven Routing Protocols in MANET. In Computational Intelligence in Data Mining-Volume 2 (pp.
369-379). Springer, New Delhi.
[31] Pallai, G. K., Meenakshi, S., Rath, A. K., & Majhi, B. (2014). A Quantitative Analysis and Behavioral
Study of Routing Protocols in MANET.
AUTHOR
Prof. Gyanendra Kumar Pallai is an alumni of Utkal University Bhubaneswar. He is
pursuing his research in Computer Science & Engineering at SOA Deemed-to-be
university, Bhubaneswar and had teaching experience of more than two decades. His area
of research interests is Mobile ad-hoc networks, Data communication and networking,
Computer architecture and Operating systems.
Dr. Meenakshi Sankaran is an alumni of National Institute of Technology Rourkela. She is working as
Senior Assistant Professor and also serving as NAAC Program Coordinator of VITBS,
VIT Bhopal University. She had 15 years of teaching experience and possesses Research
& Administrative experience of more than a decade. Her research interest mainly focused
on Competency Mapping & Skill development for Generation Z, Educational Psychology,
Management Information systems, Information & Communications Technology,
Emotional Intelligence, Mobile Adhoc Networks and Database Management systems.
Dr. Amiya Kumar Rath is a Professor in Computer Science & Engineering, Burla. At
present he is an Senior adviser, ICT for National Assessment & Accreditation Council
(NAAC), India. He contributed over 150 research papers to many national, international
journals & conferences, and author of eight books published by reputed publishers. Earned
four awards for contribution to academics and research.

Self-Pruning based Probabilistic Approach to Minimize Redundancy Overhead for Performance Improvement in MANET

More Related Content

What's hot (19)

Similar to Self-Pruning based Probabilistic Approach to Minimize Redundancy Overhead for Performance Improvement in MANET (20)

More from IJCNCJournal (20)

Recently uploaded (20)

Self-Pruning based Probabilistic Approach to Minimize Redundancy Overhead for Performance Improvement in MANET