elsevier 2015 modified hs wsn

8/19/2019 Elsevier 2015 Modified HS WSN

1/35

Accepted Manuscript

Title: An Improved Harmony Search Based Energy-EfficientRouting Algorithm for Wireless Sensor Networks

Author: Bing Zeng Yan Dong

PII: S1568-4946(15)00807-8

DOI: http://dx.doi.org/doi:10.1016/j.asoc.2015.12.028

Reference: ASOC 3385

To appear in: Applied Soft Computing

Received date: 9-5-2015Revised date: 25-11-2015

Accepted date: 16-12-2015

Please cite this article as: Bing Zeng, Yan Dong, An Improved Harmony

Search Based Energy-Efficient Routing Algorithm for Wireless Sensor

Networks, (2015),

http://dx.doi.org/10.1016/j.asoc.2015.12.028

This is a PDF file of an unedited manuscript that has been accepted for publication.

As a service to our customers we are providing this early version of the manuscript.

The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that

apply to the journal pertain.

http://dx.doi.org/doi:10.1016/j.asoc.2015.12.028http://dx.doi.org/10.1016/j.asoc.2015.12.028http://dx.doi.org/10.1016/j.asoc.2015.12.028http://dx.doi.org/doi:10.1016/j.asoc.2015.12.028


2/35


3/35

Page 2 of 34

A c c e p t e d

M a n u s

c r i p t

A new encoding of harmony memory for

routing in WSNs is proposed

A new generation method of a newharmony for routing in WSNs is proposed

The dynamic adaptation is introduced forthe parameter HMCR

An effective local search strategy is proposed

An energy efficient objective functionmodel is proposed

Improved Harmony Search Based EnergyEfficient Routing Algorithm (IHSBEER)

The experimental results clearly show thatIHSBEER outperforms other algorithms

raphical abstract (for review)


4/35

Page 3 of 34

A c c e p t e d

M a n u s

c r i p t

An Improved Harmony Search Based Energy-Efficient

Routing Algorithm for Wireless Sensor Networks

Bing Zenga, Yan Dongb,∗

aState Key Lab of Digital Manufacturing Equipment & Technology, School of Mechanical

Science and Engineering, Huazhong University of Science and Technology, 1037 Luoyu Road,

Wuhan, Chinab Department of Electronics and Information Engineering, Huazhong University of Science and

Technology, 1037 Luoyu Road, Wuhan, China

Abstract

Wireless sensor networks (WSNs) is one of the most important technologies in

this century. As sensor nodes have limited energy resources, designing energy-

efficient routing algorithms for WSNs has become the research focus. And be-

cause WSNs routing for maximizing the network lifetime is a NP-hard problem,

many researchers try to optimize it with meta-heuristics. However, due to the un-

certain variable number and strong constraints of WSNs routing problem, most

meta-heuristics are inappropriate in designing routing algorithms for WSNs. This

paper proposes an Improved Harmony Search Based Energy Efficient RoutingAlgorithm (IHSBEER) for WSNs, which is based on harmony search (HS) algo-

rithm (a meta-heuristic). To address the WSNs routing problem with HS algo-

rithm, several key improvements have been put forward: First of all, the encoding

of harmony memory has been improved based on the characteristics of routing in

WSNs. Secondly, the improvisation of a new harmony has also been improved.

We have introduced dynamic adaptation for the parameter HMCR to avoid the

prematurity in early generations and strengthen its local search ability in late gen-

erations. Meanwhile, the adjustment process of HS algorithm has been discarded

to make the proposed routing algorithm containing less parameters. Thirdly, an

effective local search strategy is proposed to enhance the local search ability, so as

to improve the convergence speed and the accuracy of routing algorithm. In addi-

tion, an objective function model that considers both the energy consumption and

the length of path is developed. The detailed descriptions and performance test

results of the proposed approach are included. The experimental results clearly

∗Corresponding author

Email addresses: [email protected] (Bing Zeng), [email protected]

(Yan Dong)

Preprint submitted to Applied Soft Computing November 25, 2015

anuscript


5/35

Page 4 of 34

A c c e p t e d

M a n u s

c r i p t

show the advantages of the proposed routing algorithm for WSNs.

Keywords: Wireless Sensor Networks, routing algorithms, energy-efficient,

meta-heuristic, harmony search algorithm

1. Introduction

In recent years, due to the rapid progress of micro-electro-mechanical-system

(MEMS) and wireless communication technology, Wireless Sensor Networks have

obtained dramatic development. As one of the most important technologies in the

21st century and the core technology of the Internet of things (IoT) [1], WSNs5

are profoundly changing the human life, and its application scopes become widerand wider, especially, the application of Small-scale Wireless Sensor Networks

(SWSNs) which contains dozens to several hundreds of sensor nodes. Large-

scale Wireless Sensor Networks can be hierarchical sensor networks comprised

of some SWSNs [2, 3]. The wide applications are involved with both civilian and10

military scenarios, including environmental monitoring, surveillance for safety

and security, automated health care, intelligent building control, traffic control,

object tracking, smart homes and smart grid, etc. [4, 5, 6, 7, 8, 9, 10, 11, 12].

A WSN comprises of numerous small devices, i.e., sensor nodes, which con-

tain sensing (measuring), computing, and communication component that ensure15

an administrator to observe and react to events and phenomena in a specific area

called a sensor field [1]. These sensor nodes scattered in the sensor field can col-

lect local environmental information, process them into useful data packets, and

send the packets to the sink node by multi-hop. The sink node transmits the pack-

ets to administrator via internet or GPRS. Most of these sensor nodes suffer from20

the same constraint: limited battery life, limited transmission power, low memory

and limited processing capabilities [1]. With the dramatic development of hard-

ware technology, the CPU and flash memory are becoming smaller, more powerful

and cheaper. As a result, the memory and processing capabilities of sensor nodes

will not be the most important obstacle for the application of WSNs. However,25

the battery technology has failed to obtain a breakthrough yet. Obviously, the

energy capacity of sensor nodes will be the key bottlenecks for the developmentof WSNs in a very long time. So the research on improving energy efficiency of

WSNs, so as to prolong the network lifetime, is still the focus at present and in the

future. The technical ways and methods of improving energy efficiency of WSNs30

mainly involve the energy efficient routing algorithms and clustering algorithms.

This paper is devoting to develop an energy efficient routing algorithm for WSNs.

For the wide applicability range of WSNs, WSNs routing protocols must be

application-based [13], which means that designing a WSNs routing algorithm

that meets the requirements of all application is impossible. Instead it is of impor-35

2


6/35

Page 5 of 34

A c c e p t e d

M a n u s

c r i p t

tance that designing general routing algorithms which somehow can be applied to

some application and meanwhile balance the energy consumption to increase thenetwork lifetime as far as possible, which is an important and challenging issue,

as well as the focus of developing the routing protocols for WSNs. As routing

in WSNs for maximum lifetime has proven to be a NP-hard problem [14], more40

and more researchers try to develop heuristic and meta-heuristic based routing al-

gorithms to address it. However, there is enormous scope for improvements in

energy efficiency for some WSNs applications. And on account of the uncertain

variable number and strong constraints of the routing problem in WSNs, most

meta-heuristics are inappropriate to design routing algorithms for WSNs. This45

paper tries to develop an energy efficient routing algorithm for WSNs based on an

improved Harmony Search (HS) algorithm, a meta-heuristic.The energy-saving mechanisms based on the management of the node status

[15], allowing turning nodes from sleep mode to transmitting/receiving mode, has

not been taken into account in this paper. Because these mechanisms are normally50

implemented in Physical and MAC layer.

The rest of this paper is organized as follows. Section 2 describes the state-

of-the-art of routing protocols for WSNs, mainly discusses the routing algorithms

based on heuristic and meta-heuristic approaches. In section 3, the IHSBEER

algorithm is described in detail, in conjunction with another approach, which is55

the core of this paper. Section 4 presents the experimental results performed to

evaluate the proposed routing algorithm. Related discussion about the proposedrouting algorithm is presented in section 5. The last section is the conclusions and

topics for further work.

2. Related work60

WSNs and mobile ad hoc networks (MANETs) have a lot in common, such as

they both involve the multi-hop communications. However, the two systems have

several significant differences. The typical and most important difference is that

WSNs is subject to energy constraint, which is not always the case in MANETs,

where the communication devices handled by humans can often be replaced or65

recharged. Due to the unattended operation of several weeks or months for mostapplications of WSNs, it is very important to manage energy resources of sensor

nodes efficiently. Therefore, many routing schemes of MANETs [16] are inappro-

priate to WSNs. And the design of energy-efficient routing algorithm for WSNs

has become the research focus.70

At present there is a great amount of research about the design and develop-

ment of routing protocols in WSNs [17, 18, 19, 20]. These protocols are designed

for different applications and different architecture of networks. However, they

3


7/35

Page 6 of 34

A c c e p t e d

M a n u s

c r i p t

all have one thing in common: they have considered the energy efficiency of sen-

sor nodes, which directly affects the extension of the network lifetime. In this75

section, a brief overview on applying heuristic and meta-heuristic algorithms to

design routing protocols for WSNs is presented. A detailed and complete refer-

ence on the motif can be seen from [21, 22, 23].

Camilo et al. proposed an Energy Efficient Ant Based Routing (EEABR) Al-

gorithm, which is based on Ant Colony Optimization (ACO), to improve the en-80

ergy efficiency of WSNs and maximize the network lifetime [24]. In this routing

algorithm, the paths between the source nodes and the sink node are found by

forward ants, which select next hop according to the amount of pheromone trail

stored in current nodes routing table and residual energy of neighbors. When

computing the number of pheromone trail that a backward ant will left during its85 journey, both the energy levels of all nodes in path and the path length are taken

into consideration, which have a great contribution to balancing the energy con-

sumption of nodes and reducing the time delay. Meanwhile, a parameter called

travelled distance, i.e., the number of visited nodes, is introduced in this phase,

which makes those nodes closer to sink node have more pheromone trail, so that90

forward ants could reach the sink node more efficiently. However, during the

early period of transmitting packets, the forward ants cannot find the optimal or

near optimal paths due to the little difference of the amount of pheromone trail in

routing table.

Multipath Routing Protocol (MRP) is proposed by Jing et al. to reduce en-95

ergy consumption and maximize the network lifetime [25], which is based on dy-

namic clustering and ant colony optimization. MRP contains three phases: cluster

formation, multipath construction and data transmission. In the first phase, the

cluster head is selected from those nodes located in the event area according to

their residual energy. Furthermore, with dynamic clustering, the efficiency of data100

aggregation has been improved dramatically, so that the energy consumption of

network has been decreased effectively. In the phase of constructing multipath,

an improved ACO algorithm is used to calculate the multiple paths between the

cluster head and the sink node, which has balanced the energy consumption of

nodes. In the last phase, the cluster head dynamically select one path to transmit105

data to the sink node according to a load balancing function.

Adaptive energy-efficient and lifetime-aware routing protocol (QELAR), based

on Q-learning technique, is proposed by Hu et al. [26]. It has introduced Q-

learning technique to the routing protocol to balance nodes workload for prolong-

ing network lifetime. Whats more, QELAR can improve energy efficiency of net-110

work by decreasing networking overhead, and better adapt to dynamic networks

by efficiently learning the environment.

Routing using ant colony optimization router chip (ACORC), proposed by Ok-

dem and Karaboga, is a multi-path routing protocol, which can provides reliable

4


8/35


9/35

Page 8 of 34

A c c e p t e d

M a n u s

c r i p t

tency and energy efficiency.

Lu et al. have proposed FRMOO routing algorithm based on fuzzy randomoptimization and multi-objective optimization, which aims to prolong network

lifetime and achieve better and wider performances of delay, latency jitter, relia-

bility, communication interference, energy and balanced energy distribution [31].160

This routing algorithm utilizes fuzzy random variables to depict the fuzziness and

randomness of link delay, link reliability and residual energy of nodes. The rout-

ing model is based on fuzzy random expected value and standard deviation model.

And a multi-objective genetic algorithm with fuzzy random simulation is used to

calculate Pareto optimal solution. However, the selection mechanism of one path165

used to transmit packets from the Pareto optimal solution has not been provided.

Fuzzy Ant Colony Optimization Routing (FACOR) algorithm, proposed byAmiri et al. [32], has combined the foraging behavior of ants with fuzzy logic to

achieve optimal path. The experimental results showed that FACOR has a better

performance than AODV in terms of routing setup time, the number of routing170

request packets, energy consumption, delay and network lifetime.

From the above, we can find that these meta-heuristics based routing algo-

rithms are developed for different applications, and the energy efficiency of net-

work need to be improved in a greater degree. HS algorithm is firstly developed by

Zong Woo Geem [33], which is inspired by the improvisation process of musician175

looking for a perfect state of harmony. According to reference [33], HS algo-

rithm is very efficient in finding the global optima in both continues and discreteoptimization problems. Until now, HS algorithm has been successfully applied

to address various optimization problems including traveling salesman problem

[33], assembly sequence planning [34], multicast routing of high-speed networks180

[35], design of water distribution network [36], clustering of the web documents

[37], design of pipeline network [38] and vehicle routing [39], etc. Reference

[35] mainly uses HS algorithm to solve the bandwidth-delay-constrained least-

cost multicast routing problem of high-speed networks. The encoding scheme of

multicast tree, i.e., harmony, is called Node Parent Index (NPI) encoding. This185

representation is implemented in an array with size 2 N elements, where N is the

number of nodes in networks. For each node v in the tree, two consecutive posi-

tions in the array are allocated to which the first position (with odd index number

in the array) is the node label and the second position (with even index number

in the array) is the position index of parent of v. However, the problem be ad-190

dressed in this paper is unicast routing problem in WSNs, which is different from

the multicast routing problem in reference [35]. The number of sensor nodes in

transmitting path is variable for unicast routing problem in WSNs. Therefore, the

encoding scheme of harmony in reference [35] is not suitable to design unicast

routing algorithm for WSNs. Nevertheless, It is still able to utilize HS algorithm195

to design unicast routing approach for WSNs in principle. However, at present,

6


10/35

Page 9 of 34

A c c e p t e d

M a n u s

c r i p t

the research regarding the application of applying HS algorithm to design unicast

routing approach for WSNs are few. Our anterior research of how to implementHS algorithm to route for SWSNs was presented in [40]. In this paper, HS algo-

rithm is used to solve the routing problem WSNs for maximizing network lifetime.200

First of all, a basic implementation of the routing process is presented. Secondly,

the encoding of harmony memory and improvisation of a new harmony have been

improved based on the characteristics of routing in WSNs. Thirdly, the adjustment

process of HS algorithm has been discarded to make the proposed routing algo-

rithm concise and only contains one algorithm-specific parameter, i.e., HMCR.205

Meanwhile, the dynamic adaptation is introduced for the parameter HMCR to

avoid the prematurity in early generations and strengthen its local search ability in

late generations. Finally, an effective local search strategy is proposed to improvethe convergence speed and the accuracy of algorithm. In addition, an objective

function model of algorithm that has taken both the energy consumption and the210

length of path into consideration is developed, which contributes to the balance of

energy consumption and the maximization of the network lifetime. Then an Im-

proved Harmony Search Based Energy Efficient Routing (IHSBEER) Algorithm

for WSNs is proposed.

3. Improved harmony search based energy-efficient routing algorithm215

Because of the strict energy constraint that the sensor nodes, deployed in real

environment and powered by battery, may not have energy replenishment capabil-

ities, it is important to balance the energy consumption of network for the routing

algorithm, so as to prolong the network lifetime as far as possible. In this section

we propose a new energy-efficient routing algorithm for WSNs, the IHSBEER220

algorithm, which is based on improved HS algorithm and Local Search strategy.

First, a basic implementation of the routing process of proposed routing algo-

rithm for WSNs is presented. Next, the Energy Efficient Harmony Search Based

Routing (EEHSBR) algorithm is explained. Finally, the proposed IHSBEER al-

gorithm is presented.225

3.1. Basic implementation of the routing process

First of all, the sink node was designed with uninterrupted power supply, big-

ger transmission power range, stronger computing power and more storage capac-

ity to execute the proposed routing algorithm. All the sensor nodes were assigned

unique IDs, and the position of all nodes were obtained by a localization algorithm230

with GPS system. All nodes are capable of detecting their remaining power and

communicate with each other using slotted Carrier Sense Multiple Access with

Collision Avoidance (CSMA-CA) [41] protocol. A basic implementation of the

routing process is as follows.

7


11/35

Page 10 of 34

A c c e p t e d

M a n u s

c r i p t

1) At the beginning, the sink node sets the same transmission power with sensor235

nodes, and broadcasts a packet containing number of hops, initialized to 0,to sensor nodes within its communication range. Receiving nodes record the

minimal hops to sink node, and ignore the packet containing bigger hops to

sink node. Then they let the value of hops plus 1, and broadcast the packet

to their neighbor nodes. In this way, all nodes can obtain their minimal hops240

to sink node. Meanwhile, they record the IDs of their neighbor nodes. Then,

each sensor node sends the packet containing its residual energy, hops and IDs

of neighbor nodes to the sink node. And the sink node executes the proposed

routing algorithm to find best forwarding paths for all nodes. After that, the

sink node sets a strong enough transmission power to send packet containing245

each nodes best forwarding path to each node in single hop.2) When a sensor node has detected useful data from surrounding, it processes

them to packet. And the forwarding path is stored in memory and its residual

energy is added to the packet. Then the sensor node transmits the packet to

next hop according to the forwarding path.250

3) When the next hop has received the packet, it adds its residual energy to the

packet, and deletes the node, which has been visited, from the forwarding path

in packet. Then sending the packet to next hop according to the forwarding

path stored in the packet. Going on with this step until the packet has been

received by the sink node.255

4) When the sink node has received the packet, it processes the packet and sendsthe data detected from surrounding to observer via internet or GPRS, then,

updates the residual energy of those nodes in packet. At regular intervals, the

sink node transmits the packet containing each nodes forwarding path to each

sensor node in single hop. If the new forwarding path of a node calculated by260

the sink node is the same as that stored in the node, the sink node does not send

packet to the node.

3.2. Energy-efficient harmony search based routing algorithm

3.2.1. Harmony search algorithm

HS algorithm has strong global search ability and also has a very simple con-265cept and few parameters. The main steps of classical HS algorithm are as follows

[42]: (1) initialize the optimization problem and algorithm parameters, (2) ini-

tialize the Harmony Memory (HM), (3) improvise a new harmony, (4) update the

HM, (5) repeat step (3) and step (4) until the termination criterion is satisfied.

The traditional HS algorithm usually used for addressing continuous optimiza-270

tion problem requests the dimension of solution vectors, i.e., harmony in HM, to

be equal. However, routing for WSNs is an unusual discrete optimization prob-

lem, in which the number of sensor nodes in path is uncertain. That is to say, the

8


12/35

Page 11 of 34

A c c e p t e d

M a n u s

c r i p t

dimension of harmony in HM may be different. Therefore, the classical HS algo-

rithm cant be applied to address routing problem of WSNs directly. And several275

improvements for traditional HS algorithm have to be put forward according to

the characteristics of WSNs routing problem.

3.2.2. Energy-efficient harmony search based routing algorithm

As explained in previous section, the encoding of harmony memory and gen-

eration of a new path are the two steps we should improve based on the character-280

istics of routing in WSNs. By improving the two key steps and proposing an ob-

jective function model taking both the energy consumption and the length of path

into consideration, Energy Efficient Harmony Search Based Routing (EEHSBR)

Algorithm is presented in this section. Fig.1 depicts the flowchart of EEHSBRalgorithm.285

As we can see from Fig.1, the initial HM is established with roulette wheel

selection method. This section describes the encoding of HM first, then illustrates

the initialization of HM with roulette wheel selection method and the generation

of a new path respectively, and finally explains the new objective function model.

1) Encoding of harmony memory290

For classical HS algorithm, the dimension of each harmony in HM must be

equal [33]. In our study, a harmony represents a forwarding path that consists

of some sensor nodes, in which the first node is source node and the last node issink node. Therefore, the dimension of each harmony in HM can be different, as

shown in Eq.1.295

HM =

X 1 X 2

...

X i...

X HMS

=

s, x1,2, · · · , x1,l, · · · , · · · , d

,

(s, x2,2, · · · , x2,m, · · · , d ) ,...

s, xi,2, · · · , xi, j, · · · , · · · , d ,

...

(s, xHMS,2, · · · , xHMS,n, · · · , d )

(1)

where s is the source node, d is the sink node. X i , i.e., (s, xi,2, · · ·, xi, j, · · ·, d ),is the i-th harmony, which represents the i-th forwarding path between the source

node and the sink node.

This improvement is the first step and has a tremendous contribution to apply-

ing HS algorithm to design routing approach for WSNs.300

2) Initialization of harmony memory

9


13/35

Page 12 of 34

A c c e p t e d

M a n u s

c r i p t

improvise a new path

evaluate the fitness ofthe new path

is the new path better than the worst

path in HM ?

use the new path to replace theworst path in HM

yes

t = 0

evaluate the fitness of eachinitial path in HM

end

no

no

t < t max

yes

start

t =t +1

initialize the HM with roulettewheel selection method

Fig. 1. Flowchart of EEHSBR algorithm

10


14/35

Page 13 of 34

A c c e p t e d

M a n u s

c r i p t

Each harmony in HM is generated with roulette wheel selection method. The

probability P(i, j) of node i selecting node j as the next hop is calculated as shownin Eq.2.

P (i, j) =

∑k ∈ N i

(hk )−h j

(C ( N i)−1)∗ ∑k ∈ N i

(hk ) if ( j ∈ N i) ,

0 else.

(2)

where hk denotes the hop count of node k to the sink node; similarly, h j denotes305

the hop count of node j to the sink node; N i is the the set of neighboring nodes of

node i, excepting the nodes already exist in current forwarding path; C ( N i) is the

capacity of set N i.It can be seen from Eq.2, the node which is within the communication range of

current node i and gets closer to the sink node is more likely to be next hop. With310

this method, it is able to generate some good initial harmony, i.e., initial forward-

ing path, which will contribute to the convergence speed of routing algorithm.

The detailed procedure of utilizing roulette wheel selection method to select

the next hop of node i is as follows: First of all, a random number R between 0 and

1 is generated. The R is used to subtract the probability P of the first node in N i.315

If the value of subtraction is negative, choose the first node in N i as the next hop

of node i. Otherwise, the value of subtraction is used to subtract the probability P

of the second node in N i. Then determine whether the value of second subtraction

is positive or negative. If it is negative, choose the second node in N i as the next

hop of node i. Otherwise, the subtraction operation goes on, until the value of 320

subtraction is negative. And the corresponding node in N i is selected as the next

hop of node i.

As we can see the initialization of the first path in HM from Fig.2, the source

node s is set as the first node. When selecting the next hop of source node, execut-

ing the above roulette wheel selection method. After subtracting the probability325

P of node 27, the subtraction value is negative, so the next hop of source node is

set as node 27. Going on with the roulette wheel selection until the sink node is

selected into the forwarding path to complete the initialization of the first path.

Going on with the process of initialization until the last path is initialized to com-plete the initialization of HM.330

3) Improvisation of a new harmony (generation of a new path)

This step is the crucial step of applying HS algorithm to design routing ap-

proach for WSNs. Supposing X ′=(s, x

′

2, · · ·, x′

i, · · ·, d ) is a new harmony, the

element x′

i is selected by memory considerations and pitch adjustments according

to the requirements of routing for WSN, as shown in Eq.3 and Eq.4.335

11


15/35

Page 14 of 34

A c c e p t e d

M a n u s

c r i p t

s d

70

89

38

79

74 73 50

86

4988

51

52

84

90

6755

54

53

85

14

32

10

7

568

13

9

17

15

16

20

29 48

21

23

27

18

11

12

19

28

30

26

31

47

57

56

66

7868

77

83

6437

41

40

60

34

82

71

80

69

59

24

39

6142

58

62

75

32

87

44

76

63

36

4381

65

4645

25 22

62

35

33

sensor field sensor node

sink nodesource node

(a) Sketch map of a WSNs

(b) initial harmony memory

31,

HM =

(s,

(s,

(s,

(s,

(s, 27,

37,

26,

32,

64,

23,

64,

56,

47,

83,

41,

71,

46,

66,

80,

70,

82,

79,

68,

89,

74,

70,

45,

55,

53,

73,

49,

55,

67,

51,

50,

88,

67,

85,

d ),

86,

86,

84,

52,

d ),

d ),

52,

d )

d ),

Fig. 2. Sketch map of initialization of the harmony memory

12


16/35

Page 15 of 34

A c c e p t e d

M a n u s

c r i p t x

′

i ←

x′

i ∈

x1,i, x2,i, · · · , xHMS,i if (P1 < HMCR) && N x′i−1

∩

x1,i, x2,i, · · · , xHMS,i

= / 0,

x′

i ∈ N ( x′

i−1) otherwise.

(3)

where HMCR is Harmony Memory Considering Rate varying from 0 to 1. N x

′i−1

is the set of neighbor nodes of node x′

i−1, excepting the nodes already exist in

current forwarding path. And P1 is a random number between 0 and 1. It can be

seen from Eq.3, if P1 is smaller than HMCR and the intersection of N x′i−1and { x1,i,

x2,i, · · ·, xHMS,i} is not empty, x′i is selected from the intersection N x′i−1

∩ { x1,i, x2,i,340

· · ·, xHMS,i} randomly. Otherwise, x′

i is randomly chosen from N x′i−1, as shown in

Fig.3.

if ( P 1 < HMCR) :

A new harmony: (s, 31, 33, 71, 80, 38, d )49, 88, 51,

if ( P 1 HMCR) :

N 31 N 80 N 88

31,

HM =

(s,

(s,

(s,

(s,

(s, 27,

37,

26,

32,

64,

23,

64,

56,

47,

83,

41,

71,

46,

66,

80,

70,

82,

79,

68,

89,

74,

70,

45,

55,

53,

73,

49,

55,

67,

51,

50,

88,

67,

85,

d ),

86,

86,

84,

52,

d ),

d ),

52,

d ),

d ),

Fig. 3. Sketch map of improvisation of a new harmony

As we can see from Fig.3, the source node s is set as the head node of the

new path. When choosing the next hop of s, if P1 is smaller than HMCR, the next

hop is randomly selected from the second column, i.e., {27, 31, 37, 26, 32}, here,345node 31 is selected. When choosing the next hop of node 31, if P1 is not smaller

than HMCR, a node randomly selected from the neighbors of node 31 is set as the

next hop, here node 33 is selected. Going on with the progress until the sink node

d is selected into the path to complete the generation of a new harmony.

Like the classical HS algorithm, every pitch selected from HM needS to be350

determined whether it should be adjusted. For the new harmony, X ′=(s, x

′

2, · · ·, x′

i,

· · ·, d ), if x′

i is selected from HM, the adjusting decision-making for x′

i is done as

follows:

13


17/35

Page 16 of 34

A c c e p t e d

M a n u s

c r i p t

Pitch adjusting decision for x′i ←

YesNo

if (P2


18/35

Page 17 of 34

A c c e p t e d

M a n u s

c r i p t

3.3. Improved harmony search based energy-efficient routing algorithm

When optimizing the continuous problems using HS algorithm, the pitch ad-

justing bandwidth is bw, which is different from the case in discrete optimization

problem. There are two algorithm-specific parameters, i.e., HMCR and PAR, in

EEHSBR algorithm presented above. From Eq.3 and Eq.4 we can see that, when380

P1 is not smaller than HMCR or N x′i−1∩ { x1,i, x2,i, · · ·, xHMS,i} = / 0, x

′

i is se-

lected from N x

′i−1

, which is as same as the case when P2 is smaller than PAR.

Obviously, PAR is not so important to EEHSBR. And because WSNs routing al-

gorithms should be concise and contain few parameters, we have discarded the

adjustment process of EEHSBR. As a result, only one algorithm-specific param-385

eter (i.e., HMCR) is left. Whats more, to avoid the prematurity and enhance the

global exploratory capability of the proposed routing algorithm in early genera-

tions and strengthen its local search ability in late generations, we have introduced

dynamic adaptation for the parameter HMCR in improvisation step. In addition,

we have proposed an effective local search strategy to improve the convergence390

speed and the accuracy of the algorithm. Thus a new routing method called Im-

proved Harmony Search Based Energy-Efficient Routing (IHSBEER) Algorithm

is proposed.

The flowchart of IHSBEER algorithm is depicted in Fig.4.

3.3.1. Adaptive HMCR395

There is only one algorithm-specific parameter (i.e., HMCR) in the proposedrouting algorithm IHSBEER. If HMCR keeps a small static value with the in-

creasing of iterations, the algorithm would loss local search ability and cant find

global optimum. And if HMCR keeps a big static value, the algorithm would be

easily trapped in local optimum. Therefore, it is important for HMCR to take400

small value to enhance the global exploratory capability in early generations, and

take big value to strengthen its local search ability in late generations, so as to find

the global optimum.

According to a large number of experimental results, we find that the proposed

routing algorithm performs best when HMCR increases with generation number405

by the exponential rule as shown in Fig.5 and expressed as follow:

HMCR (gn) = HMCRmin × exp

gn × ln(HMCRmax

HMCRmin

)

NI

(9)

where HMCR(gn) denotes the value of HMCR at gn-th generation. HMCRmaxand HMCRmin are the maximum HMCR value and minimum HMCR value re-

spectively. N I is the maximum number of iterations (improvisation) and gn is the

generation number.410

15


19/35

Page 18 of 34

A c c e p t e d

M a n u s

c r i p t

improvise a new path from HM

evaluate the fitness of the new path

is the new path

better than the worst path in HM ?

use the new path to replace the worst path in HM

yes

execute local search for the (r+1)th path

is there anyadjacent path for the

(r+1)th path ?

evaluate the fitness of the path generatedwith local search

yes

is the path generated with

local search better than the (r+1)th

path in HM ?

use the path generated with local search to replace the (r+1)th path in HM

yes

t = 0

evaluate the fitness of eachinitial path in HM

yes

end

no

t < t max

no

yes

no

no

r = 0

t < t max

r=r +1

r < HMS

yes

no

start

no

t =t +1

t =t +1

initialize the HM with roulettewheel selection method

Fig. 4. Flowchart of IHSBEER algorithm

16


20/35

Page 19 of 34

A c c e p t e d

M a n u s

c r i p t

Fig. 5. Sketch map of adaptive HMCR

3.3.2. Local search

To make the local search strategy of the proposed routing algorithm more un-

derstandable, we provide the following definition.

Definition 1. For a path(s, xi,2 , · · · , xi, j−1 , xi, j , xi, j+1 , · · · ,d) in HM, assuming that a node v belongs to the intersection between N xi, j−1 and N xi, j+1 , so define the path415

(s, xi,2 , · · · , xi, j−1 , v, xi, j+1 , · · · , d) as an adjacent path of path (s, xi,2 , · · · , xi, j−1 , xi, j , xi, j+1 , · · · , d).

The local search strategy for WSNs routing is illustrated in detail as follows.

For example, for a path, X i, i.e., (s, xi,2, · · ·, xi, j, · · ·, d ) stored in HM, if xi, j isselected, the process of local search is as follow:420

xi, j ← xi, j ∈ ( N xi, j−1 ∩ N xi, j+1 ) xi, j

if ( N xi, j−1 ∩ N xi, j+1 = / 0),

otherwise.

(10)

Eq.10 conveys that, if the intersection of N xi, j−1 and N xi, j+1 is non-empty, xi, j is

randomly selected from the intersection as shown in Fig.6. Otherwise, it remains

unchanged.

As we can see from Fig.6, for example, the node 71 in path {s, 31, 33, 71, 80,38, 49, 88, 51, d } is selected for local search. The intersection of the neighbors425of node 33 and the neighbors of node 80 is {63, 64, 65, 71, 82}. So an element israndomly selected from {63, 64, 65, 71, 82} to replace node 71 to complete localsearching, here, node 63 is selected.

17


21/35

Page 20 of 34

A c c e p t e d

M a n u s

c r i p t

s d

70

89

38

79

7473

50

86

4988

51

52

84

90

6755

54

53

85

14

32

10

7

568

13

9

17

15

16

20

29 48

21

23

27

18

11

12

19

28

30

26

31

47

57

56

66

7868

77

83

6437

41

40

60

34

82

71

80

69

59

24

39

6142

58

62

75

32

87

44

76

63

36

4381

65

4645

2522

62

35

33

Fig. 6. Sketch map of local search

4. Simulations and results

This section presents the results of simulation experiments. The simulation430

program was developed by C++ programming language on a PC with 2.3 GHz

Intel core i7 3610QM processor and 8 GB RAM to analyze the energy cost of

sensor nodes and the network lifetime. According to the effective communi-

cation radius, 10 small-scale scenarios were randomly generated. The number

of nodes of these scenarios varied from 10 to 100, with increments of 10. And435

the simulated area also varied from 200×200 m2 (10 nodes), 300× 300 m2 (20nodes) to 1100×1100 m2 (100 nodes) as shown in Fig.7 in which the squaresand stars represent the sensor nodes, and the inverted solid triangles represent the

sink nodes. Two well-known ACO based WSN routing algorithm, i.e., EEABR

[24]and ACORC [27], and EEHSBR were used to make comparisons, so as to440verify the success of IHSBEER algorithm. For each scenario four metrics were

used to compare the performance of the algorithms:

1) The Average Residual Energy, denotes the average residual energy of all nodes.

2) The Standard Deviation of Residual Energy, denotes the standard deviation of

residual energy levels on all nodes.445

3) The Minimum Residual Energy, denotes the residual energy of the node with

lowest residual energy among all nodes.

18


22/35

Page 21 of 34

A c c e p t e d

M a n u s

c r i p t

4) The Network Lifetime, denotes the number of rounds that the networks have

sustained until any node runs out of energy.

0 50 100 150 2000

50

100

150

0 100 200 3000

100

200

0 100 200 300 4000

100

200

300

400

0 125 250 375 5000

125

250

375

500

0 150 300 450 6000

150

300

450

600

(a) 10 nodes (b) 20 nodes (c) 30 nodes (d) 40 nodes (e) 50 nodes

0 175 350 525 7000

175

350

525

0 200 400 600 8000

200

400

600

0 300 600 9000

00

00

0 250 500 750 10000

250

500

750

0 275 550 825 11000

275

550

825

(f) 60 nodes (g) 70 nodes (h) 80 nodes (i) 90 nodes (j) 100 nodes

Fig. 7. 10 scenarios

To make comprehensive comparison between the proposed routing algorithms450

and EEABR algorithm, this paper has performed extensive experiments on the

following three cases:

1) Case 1: the energy level of all nodes in 10 scenarios is set to 10J. For all sce-

narios, the packets are all transmitted from the source nodes which are denoted

by the stars as shown in Fig.7.455

2) Case 2: the energy level of all nodes is set to 10J. For each scenario, each

sensor node periodically sends packets to the sink node.

3) Case 3: three energy levels were used: 10J, 20J and 30J. These levels were

uniformly distributed over the nodes. And each sensor node periodically sends

packets to the sink node.460

Related parameters are shown in Table 1. And the experiments results were

obtained by taking an average of 10 simulation runs.

4.1. case 1

Fig.8(a)-8(c) respectively show the value of the former three metrics (i.e., Av-

erage Residual Energy, Minimum Residual Energy and Standard Deviation of 465

Residual Energy) after 600 packets are received by the sink node for different

WSNs having various number of nodes. And Fig.8(d) shows the number of rounds

that the networks have sustained until any node exhausted in different scenarios.

19


23/35

Page 22 of 34

A c c e p t e d

M a n u s

c r i p t

Higher residual energy means less energy cost. From Fig.8(a) we can find

that IHSBEER algorithm gives the best results in the vast majority of the scenar-470

ios, in particular performs far better than ACORC and EEABR. It indicates that

IHSBEER algorithm can save much more energy than EEHSBR, ACORC and

EEABR do, for transmitting the same size packet to the sink node.

The balance of energy consumption has a significant impact on the network

lifetime, which is a very important factor in energy efficiency in WSNs. Large475

standard deviation of nodes residual energy means the energy is not averagely

consumed whereas smaller standard deviation indicates the balanced energy con-

sumption. Fig.8(b) conveys that the three algorithms have almost equal perfor-

mance in balancing the energy consumption of networks for case 1. That is be-

cause all the eventS happen in the same place and all the packets are sent out from480the same source node for the three algorithms in each scenario.

Minimum residual energy directly affects the network lifetime: the residual

energy of the node with minimum residual energy is greater, and the network

lifetime will get longer. As we can see from Fig.8(c), the results obtained by

IHSBEER and EEHSBR are better than those obtained by ACORC and EEABR485

in the vast majority of the scenarios. As a result, IHSBEER and EEHSBR will

perform better in prolonging the network lifetime than ACORC and EEABR does

in the vast majority of the scenarios, as shown in Fig.8(d).

And we have also recorded the time of running IHSBEER 10 times to calculate

the optimal forwarding paths from the source node to the sink node for the last490

scenario shown in Fig.7. The total time of 10 runs is 39 seconds, which means

that the base station only needs an average of 3.9 seconds to calculate the optimal

forwarding path from the source node to the sink node. So the proposed routing

algorithm has a satisfactory performance.

Table 1 EXPERIMENTAL PARAMETERS

Item Parameters

Packet size 4098 bits

communication radius 150 m

HMS 5

HMCR in IHSBEER HMCRmin=0.2, HMCRmax=0.9

HMCR, PAR in EEHSBR HMCR=0.7, PAR=0.02

evaluation times 500

α ,β α =1,β =5

ρ ρ=0.0Initial pheromone 1.0

20


24/35

Page 23 of 34

A c c e p t e d

M a n u s

c r i p t

(a) Average Residual Energy

(b) Standard Deviation of Residual Energy

(c) Minimum Residual Energy

(d) Network Lifetime

Fig. 8. Performance in WSNs with case 1

21


25/35

Page 24 of 34

A c c e p t e d

M a n u s

c r i p t

4.2. case 2495

Fig.9(a)-9(c) respectively show the value of the former three metrics after 1000

packets are received by the sink node for different WSNs scenarios. And the num-

ber of rounds that the networks have sustained until any node run out of energy

for different scenarios is shown in Fig.9(d).

It can be seen from Fig.9(a) that the IHSBEER algorithm gives the best results500

in the vast majority of the scenarios, in particular performs far better than ACORC

and EEABR. Therefore, the IHSBEER algorithm can save much more energy in

than EEHSBR, ACORC and EEABR do, for transmitting the same size packet

to the sink node. From Fig.9(b) we can find that IHSBEER and EEHSBR have

almost equal performance in balancing the energy consumption of networks for505

case 2, and obviously, they perform far better than ACORC and EEABR. As wecan see from Fig.9(c), IHSBEER gives the best results in the vast majority of the

scenarios, in particular performs far better than ACORC and EEABR. So IHS-

BEER will performs far better in prolonging the network lifetime than EEHSBR,

ACORC and EEABR do, as shown in Fig.9(d). For 10 scenarios, the rounds ob-510

tained by IHSBEER have been increased by 73.5%, 120.3%, 121.1%, 137.4%,

141.8%, 139.1%, 130%, 121%, 100% and 106.9%, respectively, compared with

those obtained by ACORC,, and increased by 72.3%, 134.6%, 130%, 158%,

148.7%, 130.8%, 148.6%, 140.2%, 163.3% and 125.3%, respectively, compared

with those obtained by EEABR. Thus, it can be concluded that IHSBEER and515

EEHSBR performs far better than ACORC and EEABR in terms of case 2.It can be seen from Fig.9, IHSBEER and EEHSBR perform far better than

ACORC and EEABR in both saving energy and balancing the energy consumption

of networks, as well as extending the network lifetime. That is because sensor

nodes always transmit packets to the sink node along optimal or near optimal520

paths by utilizing IHSBEER or EEHSBR, whereas sensor nodes using ACORC

or EEABR are not able to obtain good paths to the sink node during the early

period of transmitting packets, for difference between the amount of pheromone

trail in routing table is too small.

4.3. case 3525

Fig.10(a)-10(c) respectively show the value of the former three metrics after

1000 packets are received by the sink node for different WSNs. And Fig.10(d)

shows the number of rounds that the networks have sustained until any node ex-

hausted for different scenarios.

As we can see from Fig.10(a), the IHSBEER algorithm gives the best results in530

all the scenarios, in particular performs far better than ACORC and EEABR. Thus,

the IHSBEER algorithm can save much more energy than EEHSBR, ACORC

and EEABR do, for transmitting the same size packet to the sink node in case

22


26/35

Page 25 of 34

A c c e p t e d

M a n u s

c r i p t






23


27/35

Page 26 of 34

A c c e p t e d

M a n u s

c r i p t

3. And IHSBEER also has best performance in balancing the energy consump-

tion of networks for case 3, for IHSBEER gives the best results in all the sce-535

narios as shown in Fig.10(b). Fig.10(c) shows that IHSBEER gives the best re-

sults in the vast majority of the scenarios, in particular performs far better than

ACORC and EEABR. So IHSBEER will perform far better in prolonging the

network lifetime than EEHSBR, ACORC and EEABR do for case 3, as shown

in Fig.10(d). For 10 scenarios, the rounds obtained by IHSBEER have been in-540

creased by 138.1%, 191.5%, 199.1%, 301.9%, 307.3%, 501.9%, 317.6%, 304.1%,

282.6%, and 299.1%, respectively, compared with those obtained by ACORC,

and increased by 124.9%, 201.8%, 222.1%, 362.1%, 339.1%, 388.5%, 342.6%,

295.2%, 393.9% and 342.7%, respectively, compared with those obtained by

EEABR.545






To further prove the superiority of IHSBEER, we simulated on the scenario

contains 100 nodes additionally. In the experiments, the average residual energy,

standard deviation of residual energy and minimum residual energy of nodes were

calculated respectively with different algorithms when the number of packets re-

ceived by the sink node varied from 300, 400 to 1200. Figs.11-13 illustrate the550

24


28/35

Page 27 of 34

A c c e p t e d

M a n u s

c r i p t

comparison results. It can be seen from Fig.11-13 that IHSBEER also presents

better results than EEHSBR, ACORC and EEABR do, which conveys that IHS-BEER performs better in both saving energy and balancing the energy consump-

tion of networks than EEHSBR, ACORC and EEABR do for SWSNs. Especially

it can be seen from Fig.12, the standard deviation of residual energy obtained555

by ACORC and EEABR gets larger and larger with the increasing of number of

packets received by the sink node, while the standard deviation obtained by IHS-

BEER and EEHSBR decrease with the increasing of number of packets received

by the sink node, which indicates that IHSBEER and EEHSBR perform far better

in balancing the energy consumption of networks than ACORC and EEABR do.560

Fig. 11. Average Residual Energy

Fig. 12. Standard Deviation of Residual Energy

25


29/35

Page 28 of 34

A c c e p t e d

M a n u s

c r i p t

Fig. 13. Minimum Residual Energy

From the experiments results of the three cases, we can find that the IHSBEER

algorithm almost has presented the best results in all metrics. Finally, it can be

concluded that IHSBEER can perform far better in balancing the energy consump-

tion of networks and extending the network lifetime than EEHSBR, ACORC and

EEABR do.565

5. Discussion

IHSBEER proposed in this paper is very efficient in balancing the energy con-

sumption of networks and extending the network lifetime as shown in the experi-

ments results of section 4. That is because the forwarding paths of sensor nodes,

calculated by the sink node with IHSBEER algorithm, are always optimal or near570

optimal. The outstanding results of IHSBEER algorithm depends on four aspects:

1) a novel encoding of harmony memory; 2) an effective generation method of a

new path; 3) an effective local search strategy; 4) an excellent objective function

model. However, there are two main drawbacks of the proposed routing algorithm.

One is that the proposed approach requires a static configuration phase to calculate575

the hops of sensor nodes and to record the IDs of neighbor nodes of each node dur-ing deployment, which will takes away scalability and ease of deployment. And

the static configuration phase is essential to IHSBEER algorithm. Because of this,

it is reasonable to establish hierarchical networks with cluster heads have more

battery capacity than member sensor nodes to improve the scalability and the ease580

of deployment of networks. In the hierarchical networks, IHSBEER algorithm

is performed for the multi-hop communication between the cluster heads and the

base station. And each cluster head take charge of the joining and exiting of the

normal sensor nodes nearby. Therefore, the scalability of networks will be im-

26


30/35

Page 29 of 34

A c c e p t e d

M a n u s

c r i p t

proved on the premise of keeping the networks energy efficiency. The other one585

is that the data forwarding paths of sensor nodes are received from the sink node,which is unsatisfactory to a very large-scale wireless sensor networks. However,

large-scale wireless sensor networks can be hierarchical networks comprised of

some SWSNs. IHSBEER can be executed by base station to calculate optimal

forwarding path for each cluster head, or be executed by cluster heads to calculate590

optimal forwarding path for each member sensor nodes. And for most of cur-

rent applications, such as smart homes, industrial and manufacturing automation,

etc, the number of nodes is not so big. Thus the proposed approach have great

application prospects.

6. Conclusions and future works595

In this paper, we have proposed an energy efficient routing algorithm for

WSNs, called IHSBEER, which is based on improved HS algorithm. First of

all, the encoding of harmony memory has been improved based on the charac-

teristics of routing in WSNs, and the roulette wheel selection method is used to

initialize the HM, which contributes to the convergence speed of routing algo-600

rithm. Secondly, the improvisation of a new harmony has also been improved.

In this procedure, the adjustment process of HS algorithm has been discarded to

make the proposed routing algorithm concise and only contains one algorithm-

specific parameter, i.e., HMCR. Meanwhile, we have introduced dynamic adap-

tation for the parameter HMCR to avoid the prematurity in early generations and605

strengthen its local search ability in late generations. Thirdly, an effective local

search strategy is proposed to enhance the local search ability, which contributes

to improve the convergence speed and the accuracy of routing algorithm. In ad-

dition, an objective function model that has taken both the energy consumption

and the length of path into consideration is developed, which has a great con-610

tribution to the maximization of network lifetime. The IHSBEER algorithm has

been compared to two well-known ant based routing algorithms, i.e., ACORC and

EEABR, and EEHSBR by several simulation experiments. The experimental re-

sults showed that the IHSBEER algorithm has a far better performance in both

balancing the energy consumption, saving energy and extending the lifetime of 615

networks for different sized WSNs.

The main contributions of this paper are:

1) An improved HS algorithm is applied to develop energy efficient routing algo-

rithm for WSNs, to prolong network lifetime.

2) An effective objective function model which has a great contribution to the620

maximization of network lifetime is proposed.

27


31/35

Page 30 of 34

A c c e p t e d

M a n u s

c r i p t

3) The improved HS algorithm for WSNs routing in this paper can be used to op-

timize those problems with uncertain variable number similar to WSNs routingproblem.

In the future, we will focus on the following aspects:625

1) Based on the lr-wpan module and the energy module containing lithium ion

battery model in NS3, simulating a hierarchical WSNs with cluster heads have

more battery capacity than member sensor nodes to validate the scalability of

IHSBEER algorithm.

2) Implementing IHSBEER algorithm on real wireless sensor networks to vali-630

date its effectiveness and to test its QoS metrics.

Acknowledgements

This research was supported by NSFC (60933012), the National Key Technol-

ogy Support Program (2015BAF01B04) and the Fundamental Research Funds for

the Central Universities (2013QN138).635

References

[1] K. Yang, Wireless Sensor Networks, John Wiley & Sons, 2014.

[2] A. Thakkar, K. Kotecha, A new bollinger band based energy efficient routing

for clustered wireless sensor network, Applied Soft Computing 32 (2015)

144–153.640

[3] P. Kuila, P. K. Jana, A novel differential evolution based clustering algorithm

for wireless sensor networks, Applied Soft Computing 24 (2014) 414–425.

[4] S. A. Haque, M. Rahman, S. M. Aziz, Sensor anomaly detection in wireless

sensor networks for healthcare, Sensors 15 (4) (2015) 8764–8786.

[5] T. D. Nguyen, T. T. Thanh, L. L. Nguyen, H. T. Huynh, On the design of 645

energy efficient environment monitoring station and data collection network

based on ubiquitous wireless sensor networks, in: Computing & Communi-

cation Technologies-Research, Innovation, and Vision for the Future (RIVF),

2015 IEEE RIVF International Conference on, IEEE, Kottayam, 2015, pp.

163–168.650

[6] D. Sahin, V. C. Gungor, T. Kocak, G. Tuna, Quality-of-service differentia-

tion in single-path and multi-path routing for wireless sensor network-based

smart grid applications, Ad Hoc Networks 22 (2014) 43–60.

28


32/35

Page 31 of 34

A c c e p t e d

M a n u s

c r i p t

[7] N. Harris, G. Merrett, H. Zia, Catchment-scale water quality monitoring,

control and management framework using collaborative wireless sensor net-655

works, in: 11th International Conference on Hydroinformatics, New York,

2014.

[8] I. F. Akyildiz, M. C. Vuran, Wireless sensor networks, John Wiley & Sons,

2010.

[9] G. Xu, W. Shen, X. Wang, Applications of wireless sensor networks in ma-660

rine environment monitoring: A survey, Sensors 14 (9) (2014) 16932–16954.

[10] D. Ding, R. A. Cooper, P. F. Pasquina, L. Fici Pasquina, Sensor technology

for smart homes, Maturitas 69 (2011) 131–136.

[11] G. Hackmann, W. Guo, G. Yan, Z. Sun, C. Lu, S. Dyke, Cyber-physical

codesign of distributed structural health monitoring with wireless sensor net-665

works, Parallel and Distributed Systems, IEEE Transactions on 25 (1) (2014)

63–72.

[12] C. J. Huang, Y. W. Wang, H. H. Liao, C. F. Lin, K. W. Hu, T. Y. Chang,

A power-efficient routing protocol for underwater wireless sensor networks,

Applied Soft Computing 11 (2) (2011) 2348–2355.670

[13] K. Chi, Y. h. Zhu, X. Jiang, V. Leung, Energy-efficient prefix-free codes forwireless nano-sensor networks using ook modulation, wireless communica-

tions, IEEE transactions on 13 (5) (2014) 2670–2682.

[14] J. Park, S. Sahni, An online heuristic for maximum lifetime routing in wire-

less sensor networks, Computers, IEEE Transactions on 55 (8) (2006) 1048–675

1056.

[15] W. Ye, J. Heidemann, Medium access control in wireless sensor networks,

Springer, 2004.

[16] G. Mohandas, S. Silas, S. Sam, Survey on routing protocols on mobile adhocnetworks, in: Automation, Computing, Communication, Control and Com-680

pressed Sensing (iMac4s), 2013 International Multi-Conference on, IEEE,

Kottayam, 2013, pp. 514–517.

[17] N. A. Pantazis, S. A. Nikolidakis, D. D. Vergados, Energy-efficient routing

protocols in wireless sensor networks: A survey, Communications Surveys

& Tutorials, IEEE 15 (2) (2013) 551–591.685

29


33/35

Page 32 of 34

A c c e p t e d

M a n u s

c r i p t

[18] S. Kavi Priya, T. Revathi, K. Muneeswaran, K. Vijayalakshmi, Heuristic

routing with bandwidth and energy constraints in sensor networks, AppliedSoft Computing 29 (2015) 12–25.

[19] N. Magaia, N. Horta, R. Neves, P. R. Pereira, M. Correia, A multi-objective

routing algorithm for wireless multimedia sensor networks, Applied Soft690

Computing 30 (2015) 104–112.

[20] B. A. Attea, E. A. Khalil, A new evolutionary based routing protocol for

clustered heterogeneous wireless sensor networks, Applied Soft Computing

12 (7) (2012) 1950–1957.

[21] M. Saleem, G. A. Di Caro, M. Farooq, Swarm intelligence based routing695

protocol for wireless sensor networks: Survey and future directions, Infor-

mation Sciences 181 (20) (2011) 4597–4624.

[22] A. M. Zungeru, L. M. Ang, K. P. Seng, Classical and swarm intelligence

based routing protocols for wireless sensor networks: A survey and compar-

ison, Journal of Network and Computer Applications 35 (5) (2012) 1508–700

1536.

[23] W. Guo, W. Zhang, A survey on intelligent routing protocols in wireless

sensor networks, Journal of Network and Computer Applications 38 (2014)

185–201.[24] T. Camilo, C. Carreto, J. Silva, F. Boavida, An energy-efficient ant-based705

routing algorithm for wireless sensor networks, in: M. Dorigo, L. Gam-

bardella, M. Birattari, A. Martinoli, R. Poli, T. Sttzle (Eds.), Ant Colony

Optimization and Swarm Intelligence, Vol. 4150, Springer Berlin Heidel-

berg, 2006, pp. 49–59.

[25] J. Yang, M. Xu, W. Zhao, B. Xu, A multipath routing protocol based on710

clustering and ant colony optimization for wireless sensor networks, Sensors

10 (5) (2010) 4521–4540.

[26] T. Hu, Y. Fei, Qelar: a machine-learning-based adaptive routing protocol forenergy-efficient and lifetime-extended underwater sensor networks, Mobile

Computing, IEEE Transactions on 9 (6) (2010) 796–809.715

[27] S. Okdem, D. Karaboga, Routing in wireless sensor networks using an ant

colony optimization (aco) router chip, Sensors 9 (2) (2009) 909–921.

[28] M. Sousa, W. Lopes, F. Madeiro, M. S. Alencar, Cognitive lf-ant: A novel

protocol for healthcare wireless sensor networks, Sensors (Basel) 12 (8)

(2012) 10463–10486.720

30


34/35

Page 33 of 34

A c c e p t e d

M a n u s

c r i p t

[29] M. Liu, S. Xu, S. Sun, An agent-assisted qos-based routing algorithm for

wireless sensor networks, Journal of Network and Computer Applications35 (1) (2012) 29–36.

[30] M. Saleem, I. Ullah, M. Farooq, Beesensor: An energy-efficient and scal-

able routing protocol for wireless sensor networks, Information Sciences 200725

(2012) 38–56.

[31] J. Lu, X. Wang, L. Zhang, X. Zhao, Fuzzy random multi-objective opti-

mization based routing for wireless sensor networks, Soft Computing 18 (5)

(2014) 981–994.

[32] E. Amiri, H. Keshavarz, M. Alizadeh, M. Zamani, T. Khodadadi, Energy730

efficient routing in wireless sensor networks based on fuzzy ant colony opti-

mization, International Journal of Distributed Sensor Networks 2014 (2014)

1–17.

[33] Z. W. Geem, J. H. Kim, G. Loganathan, A new heuristic optimization algo-

rithm: harmony search, Simulation 76 (28) (2001) 60–68.735

[34] L. Wang, Y. Hou, X. Li, S. Sun, An enhanced harmony search algorithm for

assembly sequence planning, International Journal of Modelling, Identifica-

tion and Control 18 (1) (2013) 18–25.

[35] R. Forsati, A. T. Haghighat, M. Mahdavi, Harmony search based algo-

rithms for bandwidth-delay-constrained least-cost multicast routing, Com-740

puter Communications 31 (10) (2008) 2505–2519.

[36] Z. W. Geem, Optimal cost design of water distribution networks using har-

mony search, Engineering Optimization 38 (3) (2006) 259–277.

[37] M. Mahdavi, M. H. Chehreghani, H. Abolhassani, R. Forsati, Novel meta-

heuristic algorithms for clustering web documents, Applied Mathematics745

and Computation 201 (1) (2008) 441–451.

[38] Z. Geem, J. Kim, G. Loganathan, Harmony search optimization: applica-

tion to pipe network design, International journal of modelling & simulation

22 (2) (2002) 125–133.

[39] Z. W. Geem, K. S. Lee, Y. Park, Application of harmony search to vehicle750

routing, American Journal of Applied Sciences 2 (12) (2005) 1552.

[40] B. Zeng, Y. Dong, An energy efficient harmony search based routing algo-

rithm for small-scale wireless sensor networks, in: Computational Science

31


35/35

A c c e p t e d

M a n u s

c r i p t

and Engineering (CSE), 2014 IEEE 17th International Conference on, IEEE,

2014, pp. 362–367.755

[41] P. Baronti, P. Pillai, V. W. C. Chook, S. Chessa, A. Gotta, Y. F. Hu, Wireless

sensor networks: A survey on the state of the art and the 802.15.4 and zigbee

standards, Computer Communications 30 (7) (2007) 1655–1695.

[42] K. S. Lee, Z. W. Geem, A new meta-heuristic algorithm for continuous engi-

neering optimization: harmony search theory and practice, Computer meth-760

ods in applied mechanics and engineering 194 (36) (2005) 3902–3933.

32

elsevier 2015 modified hs wsn

Documents