Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Peer-to-Peer Networking and Applications

https://doi.org/10.1007/s12083-019-00821-4

An improved congestion-aware routing mechanism in sensor


networks using fuzzy rule sets
G. Sangeetha 1 & M. Vijayalakshmi 1 & Sannasi Ganapathy 2 & A. Kannan 3

Received: 15 March 2019 / Accepted: 23 September 2019


# Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
Congestion-free routing of packets across a Wireless Sensor Network (WSN) is an important issue to be addressed for quick
response applications such as disaster management, healthcare systems, traffic control, etc. The existing routing algorithms have
been developed by mainly considering hop-count leading to a gap in addressing congestion problem in the routing process.
Congestion during routing leads to increase in packet drops, increased energy consumption and delay in delivery of packets at the
sink node. Therefore, it is necessary to propose an optimal congestion aware routing mechanism that considers network param-
eters such as congestion level and energy dissipation in addition to hop count factor. Hence in this paper, a traffic prudent
framework called Congestion Aware Routing using Fuzzy Rule sets (CARF) has been proposed for handling excess traffic
conditions by identifying the non-localized node paths, there after adding them to the existing localized node paths and selecting a
more reliable and a congestion alleviated path to sink node using fuzzy rule prediction. This results in reducing packet drops and
increasing energy utilization. The proposed framework CARF is organized into two operational segments namely 1) Multiple
Path Identification by positioning non-localized nodes and 2) Congestion Mitigated routing of data packets to sink node. First
operation employs the Positioning of a Non-Localized node algorithm that is used to compute the unknown coordinates of a
sensor node thereby utilizing them to form more packet transmission paths in addition to the existing paths. The Point In Which
Side (PIWS) hop count-based geometric method is utilized here for finding non-localized nodes in the sensor network. Second
operation uses an Enhanced Fuzzy-based Congestion Mitigation (ECFM) algorithm for estimation of congestion level in nodes
using fuzzy rule sets by considering incoming data packets per second, bandwidth size and path reliability. This CARF frame-
work has been simulated using NS-2 testbed. Depending on the two major network characteristics such as path reliability and
congestion level, a network path is chosen for routing data packets. From the experiments conducted in this work, it is proved that
the proposed CARF mainly alleviates congestion, besides reducing energy cost and interference such as shadowing and
attenuation.

Keywords CARF . Path reliability . Congestion level . ECFM . Energy cost . Interference

* G. Sangeetha
sangeetha.g@auist.net; sangcse10@gmail.com 1 Introduction

M. Vijayalakshmi Congestion prediction [3] in a Wireless Sensor Network


vijim@annauniv.edu (WSN) enables to efficiently route the packets to sink with
Sannasi Ganapathy minimal energy exhaustion. The amount of traffic is estimated
sganapathy@vit.ac.in prior in each node to follow a non-collision path to sink.
A. Kannan Excessive traffic occurs in a network due to sudden occur-
kannan.a@vit.ac.in rence of event triggers that may generate more packets to be
transferred to sink than usual. Event triggers are emergency
1
Department of Information Science and Technology, CEG, Anna situations that occur due to earthquakes, landslides and other
University, Chennai 600025, India
alert conditions. The main characteristics of an effective net-
2
School of Computing Science and Engineering, VIT University, work performing traffic estimation are to reduce packet drop
Chennai 600127, India
when routing packets to sink, apart from other factors includ-
3
School of Computer Science and Engineering, VIT University, ing negligible energy usage and delay [11].
Vellore 632 014, India
Peer-to-Peer Netw. Appl.

When congestion forecasting [12, 15] is followed in a net- There are three steps involved in CARF mechanism that in-
work, the employment of relay nodes can be minimized. The cludes the discovery of neighbouring nodes, foreseeing path
amount of energy utilized in relay nodes can be brought down reliability and congestion prediction. The Positioning of non-
to increase the lifetime of network. Apart from these nodes, localized nodes algorithm enables estimation of unknown co-
the energy span of individual nodes gradually increases con- ordinate nodes using number of hops. Path reliability is cal-
tributing to degradation in network lifetime. There is a pre- culated using fuzzy rules formed using neighbour distance
estimated route that can send data packets without loss to from sink and path energy as parameters. Reliability of a
reach sink. The existing congestion predicted routing [22] node’s neighbouring path decides on the selection of a greater
algorithms employ non-fuzzy estimations. Moreover, three durable path to destination without traffic intensity. This path
broad data flow actions namely congestion mitigation, con- reliability is used to estimate traffic level in nodes and minimal
gestion avoidance and reliable data transport are followed to traffic passable node is selected for routing. The proposed
eliminate traffic overloading across the network [21]. CARF framework saves considerable energy consumption,
Congestion mitigation [23, 24] comprises of congestion de- supports for reduction of hotspot problem and grows average
tection, notification and their corresponding remedial moves. lifetime of nodes. The major contributions of the proposed
Traffic avoidance action includes designing protocols to per- work include:
form traffic control, resource control, Medium Access Control
(MAC) layer improvement and building learning automatons & Positioning of non-localized nodes using a simple hop-
to pre-analyze the incoming traffic patterns. Reliable data de- based geometric algorithm and to create more paths in
livery [13] comprises of assured routing of data packets in addition to existing ones for routing packets to sink with-
hop-by-hop or end-to-end manner [28]. All the above out traffic overloading.
discussed mechanisms suggest an estimated value to take de- & Designing fuzzy inference rule sets for finding path reli-
cision in case of node congestion. Fuzzy based congestion ability to calculate path strength among neighbouring
prediction routing algorithms are proficient in comparison routes.
with traditional congestion predicted algorithms due to their & Retaining energy by pre-estimation of congestion level in
robust design of flexible rule sets. They provide a wide space nodes using fuzzy rule sets to achieve extended network
of selecting an optimal route decision is better and which is operation.
least good. & Enabling less activation of relay nodes and anchor nodes
In Fuzzy based congestion predicted routing [17, 20], the in the network to conserve energy.
reasoning process during data transport is simpler that con-
sumes less computational power in network. This increases The remaining part of the paper is structured as follows:
network lifespan as energy usage is conserved. Fuzzy based Section 2 presents the existing works in detail. In Section 3, an
adaptive congestion control [16] accounts for buffer occupan- outline of proposed framework is presented. Route identifica-
cy, traffic rate and participants as inputs for identifying and tion and congestion mitigated routing operations are discussed
alleviating congestion. There are no fuzzy rules regarding path in Section 4 and 5. Section 6 represents the performance study
reliability across routes to sink in the existing fuzzy algo- and proposed approach is concluded in Section 7.
rithms. In Priority based congestion control mechanism [19],
congestion score and buffer occupancy were the two inputs
used in decision making of data transmission rate. However, 2 Related works
the strength of a path is implicitly required to trace a mostly
accurate route from node to sink. If path sturdiness increases, a Previously many researchers have contributed many tech-
well-formed route can be used for routing packets to sink niques related to congestion control in WSNs. Gharajeh
without packet drops or delays thereby growing network per- et al. proposed a Dynamic Three Dimensional (3D) Fuzzy
formance. When congestion cleared routing takes place, there Routing Based on Traffic Probability (DFRTP) in WSNs [1].
is minimal traffic involved routing contributing to decreased In this technique, distance and number of neighbors were
hotspot problem. Many researchers have focussed on param- considered as the input parameters, yielding traffic probability
eter inputs such as traffic rate and buffer occupancy to decide as the output parameter. A 3D localization is done by
upon next hop for node, but have not taken into account the installing an external 3D accelerometer on well caliber nodes.
path strength as a parameter to improve the network durability The main idea is to route packets from source to sink nodes by
to transmit packets to sink. hop-by-hop delivery using fuzzy decision rules. DFRTP pro-
In order to solve the problems of energy draining and more vides greater network lifetime and packet delivery ration com-
packet drop when routing data to sink during congestion oc- pared to the other heuristic methods. However, determining
currence [25], we propose a new technique called Congestion- geographical coordinates using an external accelerometer is
Aware Routing using Fuzzy rule sets (CARF) in this paper. not much reliable.
Peer-to-Peer Netw. Appl.

Congestion elimination in WSNs can be done in two ways: logic algorithm to compute traffic congestion in network and
First solution is to detect congestion, send implicit or explicit provides a considerable Quality of Service (QoS) to manage
notifications and perform the removal process by incorporat- traffic rate in WSN. This method prolongs network lifespan
ing traffic control algorithms [2, 3]. The second solution is to and decreases packet loss across the network. A more real-
carry out congestion avoidance technique that aims at time implementation of the methodology is required to study
predetermining a traffic block-free path to route the packets the traffic distribution. Ghaffari et al. [10] outlined an energy
from source to sink nodes [2]. The methods to avoid conges- effective routing mechanism using A-star algorithm. The
tion include usage of exponential back-off protocols, priority above stated EERP algorithm assures minimal hop count
node-level assignments and resource controlled allocation in and higher link quality. A considerable delay is involved dur-
the sensor network. Jiang et al. devised a localization algo- ing A* search that increases complexity even for lighter pack-
rithm called Based on the Hops for LArge-Scale WSNs et transfer.
(BHLA), where the unknown coordinate nodes can commu- Munir et al. [14] devised a congestion evaluation method
nicate with known anchor nodes under unknown communica- using QoS management and control block that is implemented
tion range [4]. The distance measure of an anchor node to an at node and sink levels using fuzzy method. This technique
unknown node is changed into hops via flooding. The unique supports network variations, sorts traffic and minimizes pack-
coordinate information contained in an area was found using et drop for priority based event-driven data traffic. The effect
the Point In Which Side (PIWS) method and intersection be- of increasing congestion related parameters and its effect on
tween communication radius of both anchor and unknown wide QoS (Quality of Service) architecture need to be
nodes. The goal of this algorithm is to find out the geograph- analysed. Saurabh et al. [16] proposed a Fuzzy Based
ical point locations of unknown nodes in unrecognized posi- Adaptive Congestion Control (FBACC) that provides a fuzzy
tions. BHLA algorithm is range-free, crops down node energy congestion estimation considering factors such as buffer oc-
cost, reduces time delay, thereby removing interference. The cupancy, participants and traffic rate. This method conserves
correctness of positioning points can be verified by enhancing retransmission energy due to minimal packet loss using traffic
PIWS phase. adaption. The learning activity about incoming data rate and
Chen et al. [5] presented a Range-free Localization based participants in the network consumes more time delay and
on Anisotropy of Nodes (RLAN) to trace node positions un- causes traffic within nodes. Jilani et al. [18] framed a
der varying magnitude across various directions. RLAN in- Hierarchical Tree Based Congestion Control using Fuzzy
creases localization accuracy compared to other range-free Logic (HTCCFL) for heterogeneous traffic flow in WSN
localization methods. Moreover, other effective distance esti- using fuzzy technique using packet service ratio, number of
mation methods other than Euclidean distance can be tested to competitors and buffer occupancy. Energy consumption is
find out the average hop distance. Annie et al. [6] presented a reduced by dynamic rate adaptation or source selects an alter-
predictive rate reservation schemes such as rate reduction nate path from hierarchical tree to control congestion. The
(RR), rate regulation (RG) and split protocol (SP). Time delay exact selection on the usage of rate adjustment and alternate
is a major factor to be considered for the reservation schemes path choosing from tree formation is ambiguous. Majid et al.
and more dynamic methodology for choosing more neighbor [19] designed a priority-based congestion control mechanism
nodes. As a result of reviewing the above techniques, a new using fuzzy rules in WSNs. Congestion score and buffer oc-
routing scheme to route the Clear-to-Send (CTS) and Request cupation are given as inputs to fuzzy rate controller to regulate
to Send (RTS) data packets in data-link layer by data transmission to sink. There is significant minimization of
predetermining congestion at each node using hop-by-hop packet loss rate and delay. Less number of traffic parameters is
movement. The RTS/CTS communication methods are used involved in decision making of traffic free route to sink node.
in IEEE 802.11 protocol to lower window frame interference In the past, many researchers worked in areas of energy
and collision due to unseen nodes. aware routing [29–34], fuzzy based route selection methods
A Fuzzy Temporal Congestion Prediction (FTCP) model for effective routing and congestion aware routing techniques
was designed to route the data packets along a least congested [35–38] for effective routing in WSN. Damaso et al. [39]
path [7]. The congestion-aware framework includes two proposed a reliability model considering battery power as
phases: Congestion Prediction and Localization phases. The the key factor. Blocks and models were used to express the
temporal fuzzy rules are mapped across these two phases. sensor reliability elements using energy factor. Deif et al. [40]
Hatamian et al. [8] proposed a fuzzy rate controller with a modeled the sensor nodes based on three modes namely on,
congestion aware routing method that assigns a priority for relay and off states. An energy-based search algorithm to dis-
the data packets. Integration of fuzzy logic and greedy ap- cover sensor node combinations based on modes was pro-
proach decreases packet drop and average energy usage in posed to yield a complete path to sink. Mostafaei et al. [41]
nodes. More upstream traffic has to be tested for delay in a devised a Distributed Learning Automaton (DLA) to select a
distributed environment. Sara et al. [9] listed a type-2 fuzzy set of smallest number of nodes to achieve end-to-end
Peer-to-Peer Netw. Appl.

reliability. From the research works done so far, QoS parameter area and this is the reason for detecting the non-localized nodes.
Path energy on a wider basis was used to calculate reliable paths to The above framework involves two primary algorithms namely
sink. Addition of neighbour distance parameter measures the Positioning of Non-Localized nodes and Enhanced Fuzzy-based
shorter distant node that can be traversed without energy loss. Congestion Mitigation (ECFM). The first algorithm constructs
Thus, the neighbour distance and path energy parameters are more multiple paths including the paths discovered with unknown co-
sufficient to estimate path reliability across network paths. Ullah ordinate nodes thereby preventing sensor battery dying by even
et al. [29] describes an efficient routing metric for smart meter distribution of data traffic. This algorithm adopts the functioning of
networks to be deployed in smart cities. This proposed metric traditional DV-Hop (Distance-Vector) and PIWS range-free local-
considers the residual energy and queue utilization of neighboring ization methods for positioning unknown coordinate nodes in sen-
nodes. This metric enables congestion-aware routing for smart sor environment. The latter algorithm applies fuzzy rules for esti-
meter networks or Advanced metering infrastructure (AMI). mating path reliability across network edges and determining con-
Song et al. [30] represented a unified modeling scheme called gestion level in all neighbouring node paths to suggest an ideal
generalized logical sphere (GLS) that can be mainly applied for path to be followed to reach sink.
space-terrestrial networks. The features including load balancing, The output obtained is a crisp number indicating conges-
smart routing, cognitive healing and hierarchical shielding of GLS tion level within a node along the routing path to sink.
were discussed. Therefore, a less traffic jammed path is chosen to achieve
However, the existing approaches discussed above consider good packet delivery and limit energy consumption.
only the energy at nodes. Congestion aware routing cannot be
performed effectively by utilizing only the node residual energy 3.2 Design of Fuzzy Inference Rule sets for estimation
parameter. Hence, it is necessary to propose a new routing mech- Path Reliability and Congestion Level.
anism that considers hop-count, congestion level, energy of nodes
and path energy in order to improve the overall network Fuzzy rule sets are framed for computing both path reliability and
performance. congestion level values in the proposed framework. The proposed
CARF design adopts Mamdani fuzzy inference process to provide
optimal decisions. The technique consists of four steps:
3 The proposed CARF framework
a) Fuzzifying input variables.
When the number of incoming packets gradually increases than b) Valuation of IF-THEN rules.
the restricted buffer size of a node while routing data to reach sink, c) Dissemination of rule outputs.
congestion occurs. The amount of data traffic shoots up due to d) Defuzzification output as a crisp number.
occurrence of sudden event triggers resulting in data jamming [26,
27]. This causes a greater packet drop that needs to be checked and Determination of path reliability and congestion level pre-
repaired using a congestion prediction algorithm. Apart from se- diction is performed using a fusion of both triangular and
vere packet loss, traffic collision can also result in energy drain and trapezoidal membership functions. Triangular membership
end-to-end delays that affect the entire network. There is a consid- function is calculated using Eq. 1.
erable degradation in network throughput that prevents sensitive
data from not reaching sink. Our goal is to construct a fuzzy set 8
>
> 0; x≤ a
model for predicting data traffic in prior for the sensor nodes to >
> x−a
frame a reliable route path for transferring packets to sink node. < ; a ≤ x≤ b
TriangularMFðx; a; b; cÞ b−a
¼ c–x ð1Þ
>
> ; b ≤ x≤ c
>
>
3.1 Key Methodology of CARF : c−b
0; c ≤x
The presented paper proposes a congestion-free routing frame- Trapezoidal membership function is computed by Eq. 2.
work for data transmission from source to sink nodes in WSNs.
The devised Congestion-Aware Routing Mechanism using Fuzzy
sets (CARF) consists of operations such as situating non-localized TrapezoidalMFðx; a1; b1; c1Þ
nodes, by which more paths for even traffic distribution could be 8
>
>
0; x≤ a1
constructed and there after predicting an optimal congestion elim- >
> x–a1
>
inated path to sink by pre-estimation of incoming traffic congestion < b1–a1 ; a1≤ x ≤ b1
>
level in network nodes. Here, the non- localized or unknown nodes ¼ 1; b1≤ x ≤ c1 ð2Þ
>
> d1−x
refer to the unrevealed node vertices hidden in sensor network. >
> ; c1≤ x ≤ d1
>
>
The point where the event trigger has occurred is important to be : d 1–c1
0; d 1≤ x
known as there are a large number of sensor nodes in deployed
Peer-to-Peer Netw. Appl.

The Centre of mass is used here as the defuzzification tech- ii) Form coordinate sets, determine area covered by PIWS.
nique. The defuzzified single valued output is given in Eq. 3. Form overlapping area between coordinates.

p
Step 3: Overlap the detected area by finding cartesian prod-
∑ xi ρ ð xi Þ uct of coordinate sets N(n-1)xN(n).
i¼1
Z¼ p ð3Þ Step 4: Calculate Centre of Gravity (COG). Position of a
∑ ρ ð xi Þ non-localized node = COG(Overlap area). If overlapping not
i¼1 satisfied, replace basis coordinate with another anchor co-
ordinate within same hop range. Repeat step 3.
where xi denotes each candidate in the output universe of
Step 5: For hop value = n, calculate location point LP = (x-
discourse, ρ(xi) is the defines degree of membership of xi
x1)(y2-y1) - (y-y1)(x2-x1) using PIWS to find overlap area.
and p is the count of candidates. These valuations complete
Traverse steps 1 to 4.
finding resultant fuzzy values to decide the selection of route
The unknown coordinate positions are located to find out
path to sink node.
all possible efficient paths to sink using Algorithm 1. For
example in the dividing area, if the sensor network has only
2 hops, then 1 hop area has 2 anchor nodes and 2 hop area has
4 Multiple PATH identification operation 3 anchor nodes. This implies 2 coordinate sets: {(ax11,by21)},
{((ax11,by11), (ax22,by22)), ((ax11,by11), (ax23,by23))} and
Finding multiple paths in CARF framework is composed of {(ax12,by21)}, {((ax12,by21), (ax22,by22)), ((ax12,by12), (ax23,-
finding non-localized nodes and constructing paths after lo- by23))}. The centre of gravity of overlapped area by nodes
cating the unknown nodes in sensor deployed area as shown constitutes the location of an unknown node. The above ap-
in Fig. 1. New network paths along localized nodes are in- proach allows to determine the nearer, middle and farther
cluded to the existing paths to provide additional routing nodes.
packets to destination. A simpler manual localization method Consider ‘n’ number of hops with ‘an’ anchor nodes having
is carried on based on PIWS [4] calculation. Consider the node coordinate points to be x and y. The unknown coordinate
anchor and non-localized nodes for distance estimation phase node can be found in sensor region using the following
as shown in Fig. 2 and Fig. 3. If there are three a-level coor- moves. Divide the sensor area to find the coordinate sets and
dinates, then N1 is given as N1 = {(ax11,ay11), (ax12, ay12), collect their information. For example, assume 4- hop value,
(ax13,ay13). For node b-level coordinate set with four points the coordination information set, N 4 = {(x4(1) ,y4(1)), (x-
(b1,b2,b3,b4), N2 = {(bx21,by21), (bx22,by22), (bx23,by23), 4(2),y4(2)),….,(x4a(n), y4a(n))}, where a(n) denotes n number
(bx24,by24)}. For c-level coordinate set with more than four of anchor nodes in 4- hop area. Hence, for n-hop value, un-
points N3 = {(cx31, cy31),…,(cx36,cy36)}. In this way, area of known coordinate information is collected utilizing a general-
node coordinates is divided and superpositioned. ized computation set Nn = {(xh1(a1),yh1(a1)),(xh2(a + 1),yh2(a +
1)),….,(xh(n)a(n),yh(n)a(n))}, where h(n) represents n-hop area.
Algorithm 1: Working steps in Positioning of Non-Localized Calculate overlapped area covered by anchor nodes and basis
nodes. node with respect to non-localized node using PIWS geomet-
Step 1: Segregating information: ric method. Select a closer basis anchor (another localized
anchor) node in addition to two anchor nodes for finding un-
Assume a0 collects minimum number of hops from an- known node. Three bisector lines are used to evenly separate
chor nodes a1, a2 and a3 using flooding. This procedure the connected path of anchor nodes that are closer to unknown
records Location information and NodeID. node. To find the point where the unknown node is located, a
Location = {1,1,1} (number of hops). basic geometrical determination is utilized. Assume two
NodeID = {a1,a2,a3}. points of anchor nodes to be A1(x1,y1) and A2(x2,y2), the
location on which side the point (unknown node) is lying
Step 2: Partitioning area: can be found using the calculation, LP = (x-x1)(y2-y1) - (y-
y1)(x2-x1), where LP denotes point location. If LP < 0, point
Involves coordinate information. If hop value = 1 then, lies on left side, LP > 0 point lies on right and LP = 0 marks
Coordinate information, N1 = ((x11,y11), (x12,y12), ….., point exactly lying on line. This way overlapping area can
(x1a,y1a)). surround the unknown node more accurately.
If hop value>3, calculate area using PIWS method. The overlapped region has to be computed to yield the
unknown node. The intersection between communication
i) Select an anchor node in N1 in random and set as a base ranges of anchor nodes placed within range of non-
point (x11,y11). positioned node implies required node region that is to be
Peer-to-Peer Netw. Appl.

Fig. 1 Proposed CARF ECFM


framework
Fuzzy Fuzzy Inference
Inference Rule set for
Rule set for estimating Fuzzification
Path Congestion
Reliability Level
Defuzzification

Crisp-valued
Positioning of Congestion Level based
Non-Localized hop to next node
nodes

Optimal
congestion-free
Sensor path to sink node

Area

Unknown nodes Anchor Nodes

localized. A square geometric shape with respect to hop re- unknown, discard and fix a new basis anchor node and follow
gion is more preferable for intersection than a circle as it can the above listed procedure. This node position estimation is
possess equal sharp dimensions to separate specific area applied for n-hop regions in the deployed sensor area.
nodes. Depending on the hop area and number of anchor
nodes, the different combinations of known coordinate sets
taken for area computation. When the network has n hops,
then n coordinate sets are constructed with anchor nodes 5 Congestion mitigated routing
a(n) within hop area h(n). For instance, a network consists of
2-hop valued region, with 1-hop region having one anchor Congestion Mitigated packet routing in CARF requires the
node N1 = (x11,y11) and 2-hop region consisting of three an- estimation of both path reliability and congestion level among
chor nodes N2 = ((x21,y22),(x22,y22),(x23,y23)). Here traversing sensor nodes. The design of fuzzy rules is divided into two
hops from one region to another uses cartesian product be- sets. First fuzzy set is framed for path reliability and a second
tween N1 and N2 sets. This equals one coordinate set of fuzzy set for prediction of excess traffic in the network.
N (1)(2) = ((x 11 ,y 11 )(x 21 ,y 22 ), (x 11 ,y 11 )(x 22 ,y 22 ), (x 11 ,y 11 )
(x23,y23)). The unknown coordinate node position is known
by finding the COG pointed towards axle location. If position

2R
c1
a1
a0
b1 a a3 a2
a0 a1 a
a C
a 2
C

a
a
C
Unknown Nodes
3
Anchor Nodes Non-LocalizedNodes Anchor Nodes
Fig. 2 Sensor node region Fig. 3 Superposition of sector areas
Peer-to-Peer Netw. Appl.

5.1 A. Fuzzy mapping for path reliability Rule 2: If Neighbor_Distance is fair and Path_Energy is
higher then Path_Reliability is medium_high.
Reliability is a network measure that is used to identify the Rule 3: If Neighbor_Distance is far and Path_Energy is
most optimal connected pathway for nodes to transmit packets medium then Path_Reliability is low.
to destination node. We utilize a mesh sensor network that can Rule 4: If Neighbor_Distance is far and Path_Energy is
be visualized as an undirected graph ‘G’. Let N represent the higher then Path_Reliability is certainly_low.
set of nodes performing a group reliable data transmission to
sink and ‘V’ is the set of all vertices in G. Then, the network G In next phase, we evaluate congestion level using fuzzy
is N-group connected, if all of the vertices of N is contained in rules to enable nodes to select favorable route to sink without
V belong to a single component of G. The main parameters any traffic collision.
that are considered to output path reliability are neighbor node
distances and total energy across paths. The neighbor distance
is calculated using Euclidean measure considering all possible
5.2 B. Fuzzy mapping for congestion level estimation
proximate nodes from current node. The fuzzy relations for
stating path strength are listed below in Table 1. Congestion Level is predicted in this part to decide on the
amount of traffic present across different routing paths to sink.
The two remarkable IF-THEN rules to select path fitness
are given below and remaining rules follow these base The denotations used for congestion_level prediction are
implications. greater_high, high, certainly_high, medium_high, medium,
medium_low, certainly_low, low and greater_low.
Case 1: If Neighbour_Distance is Far and Path_Energy is Fuzzification is executed on the membership variables
Low then Path Reliability is Greater_low. data_rate, bandwidth_size and reliability_status. Data_rate
has linguistic terms namely higher, medium, lower flows
Case 2: If Neighbour_Distance is Near and Path_Energy
is High then Path Reliability is Greater_high. and bandwidth_size marking highly, medium and less occu-
pied threshold limits. Path hop status can be derived from the
Neighbor proximity and path residual energy are the major previously determined path reliability that evaluates a well-
parameters required to estimate the path strength to route defined path to sink within N-node group. Initially, reliability
packets to sink without packet drop. A triangular membership from a route to sink node is found using path reliability.
function is used for evaluation of path reliability. Table 3 lists the congestion level decision that is used by
Neighbor_Distance marks the average distance of neighbors individual nodes in a path to opt for a traffic-free route to sink.
from sink and Path_Energy is the amount of residual energy This reduces burden of relay nodes by prior decision making
along neighbor edges. For example, consider values for of choosing appropriate routes for transmitting data to sink.
Neighbor_distance = 45, then occurrence position lies be- There is an equal distribution of energy among all nodes in the
tween Fair and Far distances. If the Path_Energy = 0.48, po- network instead of nodes dying near the base station. By con-
sition lies in between medium and high ranges. The fuzzy gestion level estimation, the energy spent on excess traffic
rules for above relation can be defined from Table 1 as given data exchanges can be prevented.
below. The range limits to trace congestion in individual nodes is
carried on using the following assumptions for fuzzy variable
Rule 1: If Neighbor_Distance is fair and Path_Energy is extensions in Table 3. The output congestion state consists of
nine fuzzy values stating congestion level to be greater high,
medium then Path_Reliability is medium.
high, less high, medium high, medium, medium low, less low,
Table 1 Fuzzy rule set to compute path reliability low and more low. Taking into fuzzy combinations of data
rate, channel capacity and hop status the concentration of con-
Neighbour_Distance Path_Energy Path_Reliability gestion level is given from the above listed fuzzy rules. When
there is more congestion level in network, then the congested
Near High Greater_high
Near Medium High Table 2 Fuzzy variable coordinates for Path Reliability selection
Near Low Certainly_high
Fair High Medium_high Neighbour_Distance Path_Energy
Fair Medium Medium
Fuzzy indices Input limits Fuzzy indices Input limits
Fair Low Medium_low
Far High Certainly_low Near 0–35 Low 0.0–0.35
Far Medium Low Fair 30–60 Medium 0.25–0.65
Far Low Greater_low Far 50–140 High 0.45–1
Peer-to-Peer Netw. Appl.

Table 3 Fuzzy rules to analyze congestion density in nodes

Data_Rate Bandwidth_Size PathHop_Status Congestion_Level

Higher_Flow Highly_occupied Weakly_connected Greater_high


Higher_Flow Medium_ocupied Passably_connected High
Higher_Flow Less_occupied Strongly_connected Certainly_high
Higher_Flow Highly_occupied Weakly_connected Medium_high
Higher_Flow Medium_occupied Passably_connected Medium
Higher_Flow Less_occupied Strongly_connected Medium_Low
Higher_Flow Highly_occupied Not_connected Certainly_Low
Higher_Flow Medium_occupied Weakly_connected Low
Higher_flow Less_occupied Strongly_connected Greater_Low
Medium_flow Highly_occupied Not_connected Greater_high
Medium_flow Medium_ocupied Weakly_connected High
Medium_flow Less_occupied Strongly_connected Certainly_high
Medium_flow Highly_occupied Not_connected Medium_high
Medium_flow Medium_ocupied Weakly_connected Medium
Medium_flow Less_occupied Strongly_connected Medium_Low
Medium_flow Highly_occupied Not_connected Certainly_Low
Medium_flow Medium_occupied Weakly_connected Low
Medium_flow Less_occupied Strongly_connected Greater_Low
Lower_flow Highly_occupied Not_connected Greater_high
Lower_flow Medium_occupied Weakly_connected High
Lower_flow Less_occupied Strongly_Connected Certainly_high
Lower_flow Highly_occupied Not_Connected Medium_high
Lower_flow Medium_occupied Weakly_connected Medium
Lower_flow Less_occupied Strongly_connected Medium_Low
Lower_flow Highly_occupied Not_connected Certainly_Low
Lower_flow Medium_occupied Weakly_connected Low
Lower_flow Less_occupied Strongly_connected Greater_Low

node and path will be avoided for future data transmission is carried on by transporting packets to sink. If congestion
until reaching normal traffic condition. Moreover, the conges- level is greater high, high or certainly high efficiency
tion information is broadcasted to all the nodes of the network percentage is calculated per node by subtracting the num-
in order to make them to select routes without congestion. ber of retransmitted packets from packets transferred in
An Enhanced Fuzzy-based Congestion Mitigation current node divided by the total packets transferred.
(EFCM) algorithm is used for predicting the optimal route The total packet loss percentage is found from efficiency
to be taken from source to sink node during data trans- at congested node. If packet loss is greater than 1/2th of
port. If the congestion level involves low, certainly low bandwidth size, then find nearest node using Euclidean
and greater low or medium high, medium and medium length to construct more number of node paths to sink
low states, a normal operation of congestion-free routing node.

Table 4 Fuzzy variable limits for Congestion Level decision

Data_Rate Bandwidth_Size PathHop_Status

Fuzzy Input Limits Fuzzy Input Limits Fuzzy Input Limits


indices Indices Indices

Higher_flow 145–200 Highly_Occupied 0.55–1 Weakly_connected 0.0–0.38


Medium_flow 44–160 Medium_Occupied 0.22–0.65 Passably_connected 0.30–0.75
Lower_flow 0–60 Low_occupied 0–0.35 Strongly_connected 0.5–1
Peer-to-Peer Netw. Appl.

1
Table 5 Values for fuzzy candidate set
Low Medium High
Fuzzy variables Minimum crisp value Maximum crisp value
0.5
Neighbor_Distance 0 130
Path_Energy 0.0 1
0
Data_Rate 0 200
Bandwidth_Size 0 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

PathHop_Status 0 1 Fig. 5 Fuzzy membership functions for path energy

Step 9: For N c = U n , discover node positions using


Algorithm 2: Enhanced Fuzzy-based Congestion Mitigation Algorithm 1. Perform working steps 1 to 8.
(ECFM) algorithm. The better reliable path for packet transfer is chosen by
Nc: Current hop node in network. following steps indexed in Algorithm 2. Table 4 gives the
Nn, Dn: List of ‘n’ neighbors for a node, distance between Nc input ranges for the fuzzy points used to predict congestion
and neighbors. intensity in nodes. The initial and end crisp values for all used
NLn, An: Number of non-localized nodes and anchor fuzzified variables is tabulated in Table 5.
nodes in a network.
DR, BS, PHS, CL: Incoming Data Rate, Bandwidth Size,
Path Hop Status and Congestion level marking crisp fuzzy 5.3 C. A Test Case Evaluation for estimating
values in nodes. Congestion Level.
P L , P T, P RT: Total packet loss, transmitted and
retransmitted packets in nodes. Consider a case where the value of Neighbor_distance is 45
ni, mi: Network nodes in the Euclidean space. and Path_Energy is 0.38. Using IF-THEN rules and variable
Step 1: Determine min(Dn) from a set of n neighboring ranges given in Tables 1 and 2, output values are computed
nodes when Nc = An. and defuzzification is done to obtain crisp number values.
When Nc = NLn, find NLn by following steps in Algorithm 1. Figures 4, 5 and 6 project the membership functions for input
Step 2: Compute path reliability and congestion level for variables.
min(Dn) neighbor node using the above listed fuzzy rules in Table 2. A triangular membership function is used when neighbor
Step 3: Follow fuzzy rule sets and consider corresponding distance is fair and trapezoidal function to find the neighbor
CL in Table 3. distance when far as follows. The same procedure is followed
Step 4: Select Nn with maximum path reliability and minimum for path energy.
congestion level CL for next hop from Nc to sink using fuzzy sets. The value A = 45 represents the fair neighbor distance from
Step 5: Repeat steps 2 and 3 to find optimal node Nn until sink then,
sink is reached. f(A; x,y,z) = max(min((A-x)/(y-x), (z-A)/(z-y)),0).
Step 6: If CL = GL||CL||L||ML||M||MH then continue with f(A; x,y,z) = max(min((45–30)/(50–45), (60–45)/(60–
normal operation using steps 1 to 5. 50)),0) = 1.
Step 7: For CL = CH||H||GH then compute packet loss in If A = 65 is the far neighbor distance from sink then,
current node during last transmission by determining the PL = f(A; w,x,y,z) = max(min((A-w)/(x-w),1,(z–A)/(z-y)),0).
100- ((PT-PRT)/PT). f(A; w,x,y,z) = max(min((65–50)/(85–65), 1,(141–65)/
Step 8: If PL > 1/2(BS), node is under greater congestion. (141–140)),0) = 0.75.
Find next nearest node to sink node using Euclidean length Similarly, to determine path energy a combination of both
p
using d(n,m) = √ ∑ ðni −mi Þ2 in p-dimensional space. If PL < triangular and trapezoidal membership functions are utilized.
i¼1 When A = 0.48 represents medium path energy along the
1/2BS, perform steps 1 to 5. edge node,
f(A;x,y,z) = f(0.48,0.25,0.45,0.65).
1 = max (min((0.48–0.25)/(0.45–0.25),(0.65–0.48)/(0.65–
0.45)),0) = 0.85.
Near 1 Fair Far
When A = 0.52 marks high path energy along edge net-
0.5 work node,
f(A;w,x,y,z) = f(0.52,0.45,0.65,1,1.1).
0
= max(min((0.52–0.45)/(0.65–0.45),1, (1.1–0.52)/(1.1–
0 20 40 60 80 100 120 140 1)),0) = 0.35.
Fig. 4 Fuzzy membership functions for neighbor distance Applying the above values to rules 1 to 4 mentioned earlier.
Peer-to-Peer Netw. Appl.

Fig. 6 Fuzzy membership 1


GL
functions for output path
reliability L CL ML M MH H CH GH
1
0.5

0.5
0

0 10 20 30 40 50 60 70 80 90 100

Rule1:Minimum value(1,0.85) = 0.85. number of nodes in an area of 200 m × 200 m with an initial
Rule2:Minimum value(1,0.35) = 0.35. energy of 0.5 J. The number of nodes employed is varied
Rule3:Minimum value(0.75,0.35) = 0.35. between 50 to 100 during start and succeeding rounds. The
Rule4:Minimum value(0.75,0.35) = 0.35. experiment is carried out on 100 nodes for more than 500
Choosing maximum of values from the above listed rules rounds. Initially a performance comparison of Packet
gives 0.85. The crisp value ranges from 50 to 140. During the Delivery Ratio (PDR), Network Lifetime (NL) and Time
defuzzification process, this value is substituted for calculat- Delay (TD) between three algorithms namely proposed
ing X that outputs Path_Reliability as 62.9. Fuzzy rules are CARF, DFRTP and EERP is performed. Later, variations for
framed on individual nodes during packet transmission and energy utilization and packet drop ratio are shown.
the greater path reliability score is selected to ensure better Table comparisons for average packet reachability and left
way to sink. The same illustrated process is applied to find over energy between various techniques are illustrated. The
congestion level along neighboring nodes. Fuzzification is packet delivery ratios are distinguished between fuzzy and
performed based on the rules g iven in Table 3. non-fuzzy methods. Fuzzy algorithms included are proposed
Defuzzification is done to yield a crisp number range for con- CARF, Priority based Congestion Control, HTCCFL and
gestion level. The node having least congestion level value is FBACC. Non-fuzzy algorithms included for correlation are
chosen among the neighbors. This procedure is repeated till CARF, EERP, TADR, FACC and STCP (Table 6).
sink node is reached to route data. Neighbor distance is mea- CARF technique is compared with the previous heuristic
sured in meters and data flow rate is given in packets per routing methods such as EERP and DFRTP. Experimental
second. A trapezoidal membership function is used for predic- results are obtained comparing Packet Delivery Ratio
tion of traffic across the transmission routes. (PDR), Network Lifetime (NL) and Time Delay (TD) under
the exchange of data packets. The results obtained are tabu-
lated as shown in Fig. 7.
Figure 5 shows the overall energy utilized for 500 rounds
6 Simulation analysis of packet transfer. Tables 1 and 2 describe the number of
successfully received packets in sink and retained nodes in
The proposed CARF algorithm has been evaluated using NS- network for different techniques in contrast with the devised
2 testbed. Simulation is performed by deploying a variable method. The mean residual energy in sensor nodes is shown in
Fig. 8. The proposed mechanism has an average increase of 15
Table 6 Simulation nodes in comparison with DFRTP and 28 nodes in relation
parameters Parameters Values

Simulation area 200 × 200 m2


No. of nodes 50–100 Performance Analysis
Transmission range 60 m 100
90
Bandwidth 5 Mbps 80
Percentile Ratio

Data packet size 512 bytes 70


60
Start energy 0.5 J 50 Proposed CARF
40
Maximum buffer size 10 packets/s DFRTP
30
Initial battery power 200 J 20 EERP
10
Number of sinks 1 0
Distribution scenario Uniform random PDR NL TD
Performance Metrics
Simulation time 200 s–1000s
Fig. 7 Comparison of performance level between various techniques
Peer-to-Peer Netw. Appl.

Energy Consumption Table 8 Residual energy comparison between CARF and prevailing
80 methods for 500 s time span
70
Number of alive nodes

60 Initial Average remaining energy (J)


50
Energy (J)
Proposed CARF CARF Mohammed Ghaffari et al. Hatamian et al.
40
Mohammed et al. [1] et al. [1] [8] [10]
30
Ghaffari et al. [8]
20
Hatamian et al. [10] 1 0.75 0.40 0.10 0.8
10
0 2 1.20 0.85 0.19 0.11
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
3 1.82 1.39 0.42 0.25
Initial energy (J)
4 2.61 1.70 0.80 0.54
Fig. 8 Energy utilization among various data traffic control techniques
5 3.20 2.57 1.58 1.12

with EERP. There is a visible growth in the increased number


of nodes in CARF algorithm. Tables 7 and 8 illustrate the 70% packet delivery ratio exhibited by HTCCFL among
mean number of packets reaching sink and remaining energy fuzzy methods. On the other end, STCP projects 73% delivery
in nodes in CARF and other algorithms. Initially, for 10 live ratio between non-fuzzy methods.
nodes data reachability in sink for CARF is 10,940 packets Table 9 indicates the maximum and minimum time delay
compared to other algorithms [1, 8, 10] transmitting 10,058, range values between CARF, fuzzy and non-fuzzy based al-
8770 and 7910 packets respectively. There is a scale increase gorithms. TADR has a lower delay period up to 1 s, succeeded
in our devised algorithm having 5338 packets more sink by CARF with 1.2 s. FACC has a greater delay of more than
reachability compared to DFRTP, 10638 packets in EERP 3 s in relation with other congestion prediction mechanisms.
and 12,340 packets in priority based congestion control Figure 11 shows the network span for a sequence of 1000
methods. Thus CARF projects a steady increase in transfer- rounds. Initially at 200 rounds, proposed algorithm has a net-
ring packets with respect to node increase. In Table 8, there is work lifetime of 98% compared to other fuzzy and non-fuzzy
an extension of 2.08 J in CARF in relation with other algo- techniques. In non-fuzzy methods, CARF has an increase of
rithms for 5 J energy run on simulation of 500 s time period. 4%, 9%, 13% and 18% in contrast with the existing EERP,
At the end of 500 s time sequence, CARF tops the existing TADR, FACC and STCP algorithms. In fuzzy mechanisms,
algorithms with residual energy of 3.02 J. CARF has a considerable shoot up of 7%, 16% and 19% in
Figure 9 shows the packet drop ratio showing the entire relation with Priority based congestion control, HTCCFL and
loss distribution along network during variable transmission FBACC algorithms.
rounds. There is a minimal loss ratio percentage of 25% in When proposed CARF mechanism is compared with
CARF compared with other techniques [1, 8, 10] marking ECRM [29] scheme for smarter grid based networks, the av-
40%, 58% and 70% loss ratios. A comparison of packet de- erage power consumption required for data transmission to
livery ratios between fuzzy and non-fuzzy methods is de- receiver node as percentage ratio can be compared. The packet
scribed in Fig. 10. The highly packet delivering algorithm is receiving nodes are evaluated with receiver level percentage
shown as CARF with 92% successful transmission in both ratios as seen in Fig. 12 indicating level of data loss in net-
fuzzy and non-fuzzy methods. At the lowest level, there is work. When ECRM has 30 active nodes power consumed
gradually decreases to 1.01 mW when RX ratio is 100%
which is slightly greater than CARF operation resulting in
Table 7 Average reachable packets in sink for CARF and existing 1 mW. For 60 working nodes, ECRM serves with slightly
mechanisms

No. of live Number of routed packets to sink


Packet Loss Ratio
80
nodes
70
Packet drop rate/s

CARF Mohammed Ghaffari Hatamian 60


et al. [1] et al. [8] et al. [10] 50
Proposed CARF
40
30
50 54,000 50,030 45,000 39,200 20
Mohammed et al. [1]

100 43,050 41,500 26,170 23,100 10 Ghaffari et al. [8]


0
150 39,085 36,000 18,200 14,600 0 15 30 45 60 75 90 105 120 135 150 Hatamian et al. [10]
200 27,040 21,540 13,800 9000 Time (s)

250 15,038 9700 4400 2702 Fig. 9 Overall packet fall distribution in congestion control
methodologies
Peer-to-Peer Netw. Appl.

120
location of sensor nodes also affects sensor communication.
Packet Delivery Percentile 100
If there is long-haul distance data transportation, there may be
80
certain packet drops if the sensor deployment area grows larg-
CARF
60
Priority based CC
er with a transmission range TX < 50 m and bandwidth size
40 HTCCFL along with path hop status has to be known accurately to
20
FBACC determine congestion level. However, CARF has to incorpo-
rate methods to check energy balancing and packet drops dur-
0
25 50 75 100 125 150 ing greater distant transmissions. CARF uses connected mesh
Number of rounds
or random topology to handle traffic conditions, yet can still
(a) utilize the efficiency of grid topology when there is increase
number of sensor nodes.
120
The total cost involved in computation of congestion alle-
Packet Delivery Percentile

100
viated routing is as follows. Assume w1, w2 and w3 are the
80
CARF weighted factors in network whose values are positive.
60 EERP Moreover, the routing cost ‘R’ is computed in this work using
40
TADR the formula.
FACC
20 STCP R ¼ ðw1*No:of Hops þ w2*Energy Consumed by routed Packets in nodes
0
þw3*Energy required for Communicating among nodesÞ=ðw1 þ w2 þ w3Þ
25 50 75 100 125 150
Number of rounds Based on the experiments conducted in this work, the
(b) values of w1, w2 and w3 have been fixed as 2, 1 and 1. The
Fig. 10 Comparison of packet delivery ratios in CARF and other a Fuzzy weights are chosen depending on specific congestion con-
and b Non-Fuzzy Techniques trolled routing application in proposed scenario. The cost is
optimized by choosing the optimal path in terms of distance
and the congestion level at nodes to get a minimum value of R.
lesser power consumption with 1.02 mW than CARF with
1.035 mW. Thus, both CARF and ECRM are subjected to
least variations in power utilization across the network. 7 Conclusion and future scope
The above listed resultant graphs and table readings evi-
dently shows that CARF performs better in terms of energy In this paper, a framework on Congestion Aware
utilization, packet delivery, data reachability and network life- Routing using Fuzzy rule set has been proposed for
time. When compared to the existing mechanisms, there is an optimal packet routing in WSNs. Pre-estimating traffic
acceptable delay variation when the number of transmission congestion level at each node and routing data packets
rounds increase. There are several parametric values in which to sink in hop-by-hop manner are the main advantages
CARF works efficiently and under certain conditions where of the proposed model. CARF reduces the maximum
the framework can still be improvised. The geographical work performed by relay nodes in case of overloaded
nodes and thereby increasing energy distribution in the
network. The proposed CARF design decreases packet
loss and time delay, thereby increasing network lifetime
Table 9 Variable delay limits for CARF and various algorithms for a compared to other fuzzy and heuristic methods used for
span of 800 rounds routing in WSNs. The proposed algorithm retains more
than 75% residual energy in the network at a minimum
Congestion prediction and alleviation algorithms Varying delay ranges
initial energy level of 1 J to enable routing of data to
(sec)
sink. There is a better packet delivery ratio of 92%
CARF 0–1.2 when CARF is compared with other fuzzy and non-
Priority based CC 0–1.5 fuzzy congestion prediction methods. The above find-
FBACC 0–1.4 ings deduce that CARF is a well-sustained algorithm
HTCCFL 0–4.9 in terms of good packet delivery and retains a maxi-
EERP Greater than 2.0 mum network lifetime when compared to the existing
TADR 0.1–1 methodologies. Future work includes the utilization of
FACC Greater than 3.0 other membership inference system with their types to
STCP 0–2.0, greater than 2.0 sketch the fuzzy rules and corresponding subsequent
variations can be observed. A more accurate localization
Peer-to-Peer Netw. Appl.

Fig. 11 Network Lifetime 100


between CARF and a Non-fuzzy
methods b Fuzzy methods
80

Percentile value
CARF
60
EERP

40 TADR
FACC
20 STCP

0
200 400 600 800 1000
Number of rounds

(a)

100
90
80
70
Percentile value

60 CARF
50 Priority based CC
40
HTCCFL
30
FBACC
20
10
0
200 400 600 800 1000
Number of rounds

(b)

analytical informatics system can be framed to identify References


the correct nodes that form an optimal path to reroute
packets to sink by balancing excess traffic and reducing 1. Gharajeh MS, Khanmohammadi S (2016) DFRTP: dynamic 3D
fuzzy routing based on traffic probability in wireless sensor net-
routing delay. The functionalities of spherical distance-
works. IET Wireless Sens Syst 6(6):211–219
based load balancing and routing derived from extra 2. Kafi MA, Djenouri D, Ben-Othman J, Badache N (2014)
space and terrestrial GLS model [30] can be incorporat- Congestion control protocols in wireless sensor networks: a survey.
ed with the proposed work. IEEE Commun Surv Tutorials 16(3):1369–1390
3. Ghaffari A (2015) Congestion control mechanisms in wireless sen-
sor networks: a survey. Netw Comput Appl Elsevier 52:101–115
4. W. Jiang, P. Wan, Y. Wang, W. Su, D. Liang, “A localization algo-
1.4 rithm based on the hops for large-scale wireless sensor networks”,
1.35
Average Power Utilization (mW)

IEEE International Conference on Wireless Communication and


1.3 Sensor Network, pp. 217–221, 2014
ECRM N=30 5. Chen L, Pang L, Zhou B, Zhang J, Liu Z, Luo Q, Sun L (2015)
1.25
RLAN: range-free localisation based on anisotropy of nodes for
1.2 ECRM N=60 WSNs. Electron Lett 51(24):2066–2068
1.15 CARF N=30 6. Rajan AU, Raja SVK, Jeyasekar A, Lattanze AJ (2015) Energy-
1.1 CARF N=60 efficient predictive congestion control for wireless sensor networks.
IET Wireless Sensor Syst 5(3):115–123
1.05
7. Mukherjee S, Dasgupta P (2013) A fuzzy real-time temporal logic.
1 Int J Approx Reason Elsevier 54:1452–1470
20% 40% 60% 80% 100%
8. Hatamian M, Bardmily MA, Asadboland M, Hatamian M, Barati H
Receiver (RX) %
(2016) Congestion-aware routing and fuzzy-based rate controller
Fig. 12 Mean power consumption percentage with RX value using node for wireless sensor networks. Radio Eng 25(1):114–123
populations of 30 and 60
Peer-to-Peer Netw. Appl.

9. S. Ghanavati, J. Abawajy, D. Izadi, "A fuzzy technique to control 26. Huang X, Feng S, Zhuang H (2011) Jointly optimal congestion
congestion in WSN", Proceedings of International Joint Conference control, channel allocation and power control in multi-channel
on Neural Networks, Dallas, Texas, USA, August 4–9, pp. 1806– wireless multihop networks. J Comput Commun Elsevier 34:
1810, 2013 1848–1857
10. Ghaffari A (2014) An energy efficient routing protocol for wireless 27. Godoy PD, Cayssials RL, Garino CGG (2018) Communication
sensor networks using A-star algorithm. J Appl Res Technol 12(4): channel occupation and congestion in wireless sensor networks. J
815–822 Comput Electr Eng Elsevier:1–13
11. Ren F, He T, Das SK, Lin C (2011) Traffic-aware dynamic routing 28. Sharma VK, Kumar M (2017) Adaptive congestion control scheme
to alleviate congestion in wireless sensor networks. IEEE Trans in mobile ad-hoc networks. Peer-to-Peer Networking and
Parallel Distrib Syst 22(9):1585–1599 Applications, Springer Science 10:633–657
12. Antoniou P, Pitsillides A, Blackwell T, Engelbrecht A, Michael L 29. Ullah R, Faheem Y, Kim B (2017) Energy and congestion-aware
(2013) Congestion control in wireless sensor networks based on routing metric for smart grid AMI networks in smart city. IEEE
bird flocking behavior. J Comput Netw Elsevier 57:1167–1191 Access 5:13799–13810
13. Y. G. Iyer, S. Gandham, S. Venkatesan, “STCP: a generic transport 30. Song F, Zhou Y, Chang L, Zhang H (2019) Modeling space-
layer protocol for wireless sensor networks”, IEEE, 14th terrestrial integrated networks with smart collaborative theory.
International Conference on Computer Communications and IEEE Netw 33(1):51–57
Networks, ICCCN Proceedings, pp. 449–454, 2005
31. Kumar SM, Ganapathy S, Vijayalakshmi M, Kannan A (2017) An
14. S. A. Munir, Y. W. Bin, R. Biao, M. Jian, “Fuzzy logic based
intelligent secured and energy efficient routing algorithm for
congestion estimation for QoS in wireless sensor network”, IEEE
MANETs. Wirel Pers Commun 96(2):1753–1769
Wireless Communications and Networking Conference (WCNC)
32. Logambigai R, Ganapathy S, Kannan A (2018) Energy–efficient
proceedings, pp. 4339–4344, 2007
grid–based routing algorithm using intelligent fuzzy rules for wire-
15. Yin X, Zhou X, Huang R, Fang Y, Li S (2009) A fairness-aware
less sensor networks. Comput Electr Eng Elsevier 68:62–75
congestion control scheme in wireless sensor networks. IEEE Trans
Veh Technol 58(9):5225–5234 33. Wang Q, Lin D, Yang P, Zhang Z (2019) An energy-efficient com-
16. S. Jaiswal, A. Yadav, "Fuzzy based adaptive congestion control in pressive sensing-based clustering routing protocol for WSNs. IEEE
wireless sensor networks", Sixth International Conference on Sens J 19(10):3950–3960
Contemporary Computing (IC3), pp. 433–438, 2013 34. Dhand G, Tyagi SS (2019) SMEER: secure multi-tier energy effi-
17. J. Wei, B. Fan, Y. Sun, "A Congestion Control Scheme Based on cient routing protocol for hierarchical wireless sensor networks.
Fuzzy Logic for Wireless Sensor Networks", 9th International Wirel Pers Commun 105(1):17–35
Conference on Fuzzy Systems and Knowledge Discovery, pp. 35. Sangeetha G, Vijayalakshmi M, Ganapathy S, Kannan A (2018) A
501–504, 2012 heuristic path search for congestion control in WSN. Lect Notes
18. Sayyad J, Choudhari NK (2014) Hierarchical tree based congestion Netw Syst Springer 11:485–495
control using fuzzy logic for heterogeneous traffic in WSN. Int J 36. Robinson YH, Julie EG, Kumar R, Son LH (2019) Probability-
Curr Eng Technol 4(6):4136–4143 based cluster head selection and fuzzy multipath routing for
19. M. Hatamian, H. Barati, "Priority-based Congestion Control prolonging lifetime of wireless sensor networks. Peer Peer Netw
Mechanism for Wireless Sensor Networks using Fuzzy Logic", Appl:1–15
6th International Conference on Computing, Communication and 37. Preeth SKSL, Dhanalakshmi R, Kumar R, Mohamed Shakeel P
Networking Technologies (ICCCNT), pp. , 2015 (2018) An adaptive fuzzy rule based energy efficient clustering
20. Rezaee AA, Pasandideh F (2018) A fuzzy congestion control pro- and immune-inspired routing protocol for WSN-assisted IoT sys-
tocol based on active queue management in wireless sensor net- tem. J Ambient Intell Humaniz Comput:1–13
works with medical applications. Wirel Pers Commun 98:815–842 38. Nisha UN, Basha AM (2018) Triangular fuzzy-based spectral clus-
21. Sergiou C, Antoniou P, Vassiliou V A comprehensive survey of tering for energy-efficient routing in wireless sensor network. J
congestion control protocols in wireless sensor networks. IEEE Supercomput:1–26
Commun Surv Tutorials 16, 2014(4, Fourth Quarter):1839–1859 39. Damaso A, Rosa N, Maciel P (2014) Reliability of wireless sensor
22. Chen S, Yang N (2006) Congestion avoidance based on lightweight networks. Sensors MDPI 14:15760–15785
buffer management in sensor networks. IEEE Trans Parallel 40. Deif D, Gadallah Y (2017) A comprehensive wireless sensor net-
Distrib Syst 17(9):934–946 work reliability metric for critical Internet of Things applications.
23. Fang W, Chen J, Shu L, Chu T, Qian D (2010) Congestion avoid- EURASIP J Wirel Commun Netw 145:1–18
ance, detection and alleviation in wireless sensor networks. J
41. Mostafaei H (2019) Energy-efficient algorithm for reliable routing
Zhejiang Univ Sci C (Comput Electron) 11(1):63–73
of wireless sensor networks. IEEE Trans Ind Electron 66(7):5567–
24. Karenos K, Kalogeraki V (2010) Traffic management in sensor 5575
networks with a mobile sink. IEEE Trans Parallel Distrib Syst
21(10):1515–1530
25. Wei K, Guo S, Li X, Zeng D, Xu K (2016) Congestion control in Publisher’s note Springer Nature remains neutral with regard to jurisdic-
social-based sensor networks: a social network perspective. Peer tional claims in published maps and institutional affiliations.
Peer Netw Appl Springer Science 9:681–691
Peer-to-Peer Netw. Appl.

Sangeetha Ganesan Ganapathy Sannasi has com-


has completed Masters from the pleted his M.E. and PhD from
Department of Computer Science Anna University, Chennai.
and Engineering. She is currently Currently he is working as an
a research scholar in the Assistant Professor (Senior) in
Department of Information VIT University-Chennai
Science and Technology at Anna Campus. His areas of interest are
University, Chennai. Her research Artificial Intelligence, Wireless
interests include wireless adhoc Sensor Network, Data Mining,
and sensor networks, congestion and Recommendation systems.
control and traffic analysis.

Vijayalakshmi Muthuswamy is Kannan Arputharaj is currently


currently working as an working as a Professor in the
Associate Professor in the School of Computer Science and
Department of Information Engineering, VIT, Vellore, India.
Science and Technology in He has published more than 350
College of Engineering, Guindy, Papers in journals and Conference
Anna University, Chennai. She Proceedings. His areas of Interest
has completed PhD under the fac- are DBMS, AI and Networks.
ulty of Information and
Communication Engineering in
Anna University, Chennai. Her
research interests include Mobile
Databases, Mobile Cloud
Computing and Machine
Learning.

You might also like