You are on page 1of 17

79

CHAPTER 5: FUZZY LOGIC BASED CLUSTERING HIERARCHY FOR


NON-UNIFORM NODE DISTRIBUTIONS IN WSN

5.1 Fuzzy Logic based Energy efficient Clustering Hierarchy (FLECH) for WSN:

In Chapter 5 of this dissertation, Fuzzy Logic based Energy efficient Clustering

Hierarchy (FLECH), another fuzzy based clustering algorithm for non-uniform WSN is

proposed. In the non-uniform Wireless Sensor Network (WSN) [100], certain areas of

Region Of Interest (ROI) will be dense and others may be sparse. In this condition, mere

weight based or probabilistic cluster head (CH) election will affect the network lifetime

of WSN.

The probabilistic approach elects CHs without considering the various lifetime

parameters such as energy level, number of neighbor nodes, intra and inter cluster

distance, etc. It leads to election of non suitable nodes as CHs which affects the lifetime

of the WSN.

In the case of weight based approach, the CHs are elected based on the above said

parameters. When the nodes are not uniformly deployed over the ROI, the same nodes

are elected as CH frequently. It leads to premature death of that higher weight nodes and

affects the overall lifetime of the network. In overall literature survey, it is found that the

algorithms which blend both probabilistic approach and weight based approach gives

better results than following any one approach alone.

The proposed algorithm FLECH is designed in such a way that it uses both probability

and weight for CH election. In a non-uniform network, this blending of both the approach

gives a chance to all nodes in the network to get elected as CH.


80

5.2 Fuzzy System in FLECH:

FLECH used three input parameters for electing CHs and they are: (i) residual energy,

(ii) node centrality and (iii) distance to Base Station (BS). The output parameter is the

chance. Finally, CHs are elected based on this chance value.

 Residual energy: A CH node should have sufficient energy for its activities like

data reception, aggregation and transmission.

 Node centrality: Number of neighboring nodes within the communication radius

Rc of a node is termed as Node degree. Node centrality is a factor which

determines how a node is located in middle among its neighbors. The lower the

node centrality value represents the more chance of a node to act as CH.

∑ ⁄
= (5.1)
_

In Equation (5.1), ND is the node degree. Ntk_Dimension is the dimension of

ROI. For example, in a 100 m x100 m square area ROI, 100 is the

Ntk_Dimension.

 Distance to BS: The energy consumption for transmitting the data increases with

the distance between transmitter and receiver nodes. From energy conservation

perspective, the distance between CH and BS should be minimized.

FLECH has the output variable ‘Chance’.

 Chance: It represents the ability of a node to act as CH. The defuzzified value

will be a crisp output value through which the higher eligible nodes are finally

elected as CH.
81

1 Low
Medium
Degree of membership

0.8 High

0.6

0.4

0.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


Residual energy
Figure 5.1: Membership function for Residual Energy

1 Close
Reachable
Degree of membership

Distant
0.8

0.6

0.4

0.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


Node centrality
Figure 5.2: Membership function for Node Centrality
82

1
Nearby
D egree of m em bership

0.8 Average
Far
0.6

0.4

0.2

0 20 40 60 80 100 120 140 160


Distance to BS
Figure 5.3: Membership function for Distance to BS

The membership functions for the three input parameters are shown in Figure 5.1, Figure

5.2 and Figure 5.3. Low, Medium and High are the fuzzy linguistic variable for the

residual energy. The linguistic variables for node centrality are Close, Reachable and

Distant. Nearby, Average and Far are the linguistic variables for distance to BS. Low,

High, Close, Distant, Nearby and Far follows Trapezoidal membership function whereas

Medium, Reachable and Average follows triangular membership function.


83

1 very low
low
Degree of membership

0.8
rather low

0.6 low medium


medium
0.4 rather medium
rather high
0.2
high
0 very high
0 10 20 30 40 50 60 70 80 90 100
Chance
Figure 5.4: Membership Function for output variable ‘Chance’

The output variable chance has Very High (VH), High (H), Rather High (RH), High

Medium (HM), Medium (M), Low Medium (LM), Rather Low (RL), Low (L), Very Low

(VL) as its nine output linguistic variables. In this, Very High and Very Low have

trapezoidal membership function. All other output linguistic variables have triangular

membership function. Figure 5.4 depicts the output parameter chance’s membership

functions.

The fuzzy if-then rules of FLECH are given in Table 5.1. These rules are developed

based on Mamdani [66] method, which was simpler and yields better results. In total, 27

rules are there based on the combination of different linguistic variables.


84

Table 5.1: FLECH fuzzy if-then rules

S.No Residual Energy Node Centrality Distance to BS Chance


1 Low Distant Nearby Very low
2 Low Reachable Nearby Low
3 Low Close Nearby Rather low
4 Low Distant Average Very low
5 Low Reachable Average Low
6 Low Close Average Rather low
7 Low Distant Far Very low
8 Low Reachable Far Low
9 Low Close Far Rather low
10 Medium Distant Nearby Low medium
11 Medium Reachable Nearby Medium
12 Medium Close Nearby Rather medium
13 Medium Distant Average Low medium
14 Medium Reachable Average Medium
15 Medium Close Average Rather medium
16 Medium Distant Far Low medium
17 Medium Reachable Far Medium
18 Medium Close Far Rather medium
19 High Distant Near Rather high
20 High Reachable Near High
21 High Close Near Very high
22 High Distant Average Rather high
23 High Reachable Average High
24 High Close Average Very high
25 High Distant Far Rather high
26 High Reachable Far High
27 High Close Far Very high
85

The output from Fuzzy Inference System (FIS) is again a fuzzy linguistic variable. Center

of Area method is used to defuzzify the output to a crisp value.

5.3 Phases in FLECH:

FLECH has two phases: (i) Cluster formation phase and (ii) Data processing phase.

5.3.1 Cluster Formation phase:

At the start of this phase, a hello message will be broadcast by all the nodes in its

communication radius Rc. The hello message is used to calculate the node degree of a

node. All the nodes in the network have to select a value between 0 and 1 in random. The

nodes whose values are less than the predetermined threshold value Pthres will be elected

as the probationary_CH nodes. The probationary_CH will execute its FIS to compute

their chance values. Then, the probationary_CH nodes will broadcast Probationary_CH

message to all the nodes within Rc. The Probationary_CH message will contain the

details like node id and chance value. If a probationary_CH node receives

Probationary__CH message with higher chance than its own, it withdraws from CH

competition by broadcasting Quit_CH message otherwise it broadcast an Elected_CH

message to all nodes within radius RC. The non-CH joins the nearest CH to it by sending

Join_Req message containing the node id.

5.3.2 Data Collection phase:

The elected CHs nodes generate the TDMA schedule for its cluster members and

broadcast it to them. Member nodes send the data to their respective CH during its

allotted time slots. In other time slots they will be in sleep state in order to conserve the
86

energy. The CH nodes after collecting the data from all the CMs aggregate them into a

single data message. The aggregated data will be finally reported to BS from CHs. All the

CHs should be in Wake up state all through the time unlike member nodes. All these

activities of CH lead to increased energy consumption. In order to achieve balance energy

consumption, the CH role is rotated among all the nodes in every rounds. The total CM to

CH communication energy cost is given in Equation (5.2)

_ _ =∑ ∑ (, ) (5.2)

where k is the number of clusters in the network, m is the number of CMs within a

cluster, and ETx(j,CHi) represents the transmission energy between member node j to its

CH in the ith cluster of the network.

The total aggregation cost in the network is depicted in Equation (5.3), where mi is the

number of CMs in the ith cluster, l is the number of bits in the data message and EperDataBit

is the aggregation cost for a single bit.

=∑ . . (5.3)

The communication cost of aggregated data from CHs to the BS is depicted in Equation

(5.4)

=∑ ( , ) (5.4)

5.3.3 Time Complexity of FLECH:

In FLECH, all the nodes generates a random value between 0 and 1 to decide whether it

is probationary CH or not for the current round and it takes n computation. Then the
87

elected probationary nodes compute its fuzzy cost and it takes p computations. Each

probationary node decided itself to be CH for current round or not by comparing with

other nearby probationary CH and it takes px computations. Finally k numbers of CH are

elected. Then member node will be sending Join_Req for the nearest CH node and it

takes mk computations.

( )= + + + (5.5)

( )= + (1 + ) + = ( ) (5.6)

In Equation (5.5) and (5.6), n = number of nodes in the network, p = number of

probationary CHs, x=number of neighbor probationary CH, k = number of cluster heads,

m = number of members

FLECH algorithm is shown in Figure 5.5.


88

FLECH Algorithm:

1 N = Total number of nodes


2 i = sensor node
3 for i=1:N
4 x=rand(0,1)
5 if (x<Pthres)
6 state (i)=probationary CH
7 RE(k) =residual energy of the node.
8 NC(k)=Centrality of the node within its Network Dimension
9 dist_to_BS(k)=distance to BS
10 chance = FuzzyChance(RE(k), NC(k), dist_to_BS(k))
11 broadcasts probationary_CH
12 else
13 state (i)=CM
14 end
15 for m=1:N
16 x (m)=list of all neighbor probationary_CH
17 this=current sensor node’ m’
18 if (state (m)==probationary CH)
19 status=1
20 for t=1:|x(m)|
21 if (this.chance>chance(t))
22 continue
23 else
24 status=0
25 break
26 end
27 end
28 if (status==1)
29 state (m)=Final CH
30 advertise Elected_CH
31 else
32 state (m)=CM
33 advertise Quit_CH
34 end
35 end
36 for w=1:N
37 if (state (w)==CM)
38 send Join_Req to closest CH
39 end
40 exit

Figure 5.5: FLECH algorithm


89

5.4 Simulation Setup:

The proposed algorithm FLECH is compared with Low Energy Adaptive Clustering

Hierarchy (LEACH), Cluster Head Election mechanism using Fuzzy logic (CHEF),

Energy aware distributed Clustering Protocol using Fuzzy approach (ECPF) and Energy

Aware Distributed Clustering using Fuzzy Logic approach (EADC-FL). LEACH is a well

known standard clustering protocol to be compared, whereas ECPF and CHEF uses fuzzy

logic based clustering. EADC-FL is taken for comparison since it is designed specifically

for non-uniform network like FLECH. In order to verify the efficiency of FLECH over

other clustering algorithms, the three network scenarios as discussed in Chapter 3 are

used.

The control message in the algorithms has been given the size of 25 bytes. Data messages

take the size of 500 bytes which is much larger than control messages. In FLECH, the

control message transmissions occur during the cluster formation phase and data message

transmissions during data collection phase. In the simulations, these control message

communication costs are also included.

The desired percentage of CH in LEACH is set as 0.1 in all the scenarios. The Popt and

Pthres value is set as 0.3 for CHEF and FLECH respectively. α value in ECPF is set as 1

which results in dynamic clustering for every round. The communication radius Rc of the

nodes is set as 40 m for all the three scenario mentioned in Chapter 3. The network size

used for the simulation is 200 m x 200 m square area. As mentioned in Chapter 4, BS is

located at (100, 100), (200, 200) and (100, 250) in scenario 1, 2 and 3 respectively. All

the results values are based on the average of 30 simulations. For each scenario, 10

simulation runs are conducted by varying network deployment. So in total 30 simulations


90

are conducted for all the three scenarios. MATLAB environment is used for our

simulations since it supports the easy way of building FIS engines for CHEF, ECPF,

EADC-FL and FLECH through its toolbox environment.

5.5 Results and Analysis:

5.5.1 Average Energy consumption:

The average energy consumption of data collection is an important metric for analyzing

the performance of the algorithms. Figure 5.6 depicts the average energy consumption in

the data collection phase in the network.

0.35
E n e rg y c o n s u m p tio n in J o u le s

Scenario 1
0.3
0.2618 Scenario 2
0.2451 Scenario 3
0.25
0.2214 0.2204 0.2091
0.2105 0.2004
0.2
0.1610 0.1558 0.1498 0.1494
0.1545 0.1500 0.1442 0.1420
0.15

0.1

0.05

0
LEACH CHEF ECPF EADC-FL FLECH
Clustering algorithms
Figure 5.6: Average energy consumption

In all the scenario, LEACH lags behind other algorithms due to its probabilistic way of

electing the CH. CHEF somewhat reduces the energy by electing the CH using residual

energy and local distance. But in certain cases, the local distance makes the nodes having

higher number of neighbor nodes to get less chance of getting elected as CH. Though
91

ECPF, EADC-FL shows improvement than LEACH and CHEF, the non-inclusion of

residual energy in computing the fuzzy cost made it to lag behind FLECH. FLECH

founds to show improvement because of its blending of probabilistic and fuzzy based

weight CH election.

5.5.2 Network Lifetime:

Like in Chapter 4, here also First Node Die (FND), Half Node Die (HND) and Most

Node Die (MND) are used for the performance evaluation. Figure 5.7, 5.8 and 5.9 depicts

the network lifetime up to last node die condition for all the algorithms. The number of

rounds completed for every 10 node death is shown in the Figures 5.7, 5.8 and 5.9

respectively for scenario 1, 2 and 3.

900

800
Number of rounds

700

600
LEACH
CHEF
500 ECPF
EADC-FL
FLECH
400
0 10 20 30 40 50 60 70 80 90 100
Number of dead sensors
Figure 5.7: Network lifetime in scenario 1
92

600

550
Number of rounds

500

450

400

350 LEACH
CHEF
300 ECPF
EADC-FL
250
FLECH
200
0 10 20 30 40 50 60 70 80 90 100
Number of dead nodes
Figure 5.8: Network lifetime in scenario 2

550
LEACH
500 CHEF
Number of rounds

ECPF
450 EADC-FL
FLECH
400

350

300

250

200
0 10 20 30 40 50 60 70 80 90 100
Number of dead nodes
Figure 5.9: Network lifetime in scenario 3

Figure 5.10 depicts the FND of the simulated algorithms. It is observed from Figure 5.10

that FLECH increases the lifetime of WSN in scenario 1 than LEACH by 63.18%, CHEF

by 28.12%, ECPF by 10.80% and EADC-FL by 3.30%. In case of scenario 2, FLECH

increases lifetime than LEACH by 35.10%, CHEF by 16.54%, ECPF by 13.35%, EADC-

FL by 3.76%. On considering scenario 3, the lifetime of the network is increased by


93

FLECH than LEACH by 33.94%, CHEF by 30.35%, ECPF by 10.60%, EADC-FL by

4.28%.

700
656
635 LEACH
592
600 CHEF
Number of rounds

512 ECPF
500 EADC-FL
FLECH
402
400
319 331
292 280 292
300 284 264
245
218 224
200

100

0
Scenario 1 Scenario 2 Scenario 3
FND in various scenarios
Figure 5.10: First Node Die (FND) in various scenarios

HND of the network is depicted in Figure 5.11. It is observed from Figure 5.11 that

FLECH gives better increased lifetime in scenario 1 than LEACH by 26.44%, CHEF by

20.28%, ECPF by 13.33% and almost equal to EADC-FL. In scenario 2, FLECH

increases the lifetime than LEACH by 13.09%, CHEF by 20.55%, ECPF by 9.95%,

EADC-FL by 3.71%. On considering scenario 3, FLECH shows improvement than

LEACH by 45.39%, CHEF by 40.31%, ECPF by 4.98% and EADC-FL by 4.73%.


94

800 763 765


675
LEACH
700 CHEF
636
Number of rounds
605 ECPF
600
EADC-FL
500 475 FLECH
458 442
420 432 421 422
394
400
304 315
300

200

100

0
Scenario 1 Scenario 2 Scenario 3
HND in various scenarios
Figure 5.11: Half Node Die (HND) in various scenarios

1000
LEACH
800 801 CHEF
800
Number of rounds

ECPF
692 713
EADC-FL
638 FLECH
600
534 512
489 512 483 485
473
450
359
400
331

200

0
Scenario 1 Scenario 2 Scenario 3
MND in various scenarios
Figure 5.12: Most Node Die (MND) in various scenarios

Fig 5.12 depicts the MND conditions of the clustering algorithms. EADC-FL and FLECH

are giving equal performance in scenario 1. But FLECH shows improvement in scenario

2 and 3. Thus when the BS is moved out of ROI, FLECH performance is better than

others.
95

FLECH shows better improvement than others due to the proper election of CH using

appropriate parameters. Also it ensures the proper rotation of CH responsibility among all

the nodes through the initial probabilistic selection of probationary CH. ECPF and

EADC-FL lags behind FLECH because of pure weight based election of CH. In the case

of both, the highly eligible nodes are frequently elected as CH which leads to premature

death of certain nodes. As discussed in Chapter 4, CHEF lags behind ECPF, CHEF,

EADC-FL and FLECH because of its inappropriate input variable local distance. And

LEACH shows poor performance than all others due to its inappropriate CH election

mechanism.

5.6 Conclusion:

In the case of non-uniform distribution of nodes in the networks, mere metric based CH

election will deplete energy faster for certain nodes. In this chapter, a new fuzzy based

clustering algorithm FLECH is proposed for WSN where non-uniform node distribution

is followed. FLECH combines both probabilistic and metric based approaches for

electing CH. FLECH shows better increased lifetime than other fuzzy based clusters in

non-uniform node deployment.

You might also like