1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017
Genetic based Balanced Clustering Protocol for Mobile Wireless Sensor Network Dr. Zainab Tawfeeq Alisa (Assist. Prof.)
Hussein A. Nassrullah
Electrical Engineering Dept. University of Baghdad Baghdad, Iraq
[email protected]
Computer Engineering Dept. University of Baghdad Baghdad, Iraq
[email protected]
Abstract— During the last decades, the popularity of Wireless Sensor network (WSN) has been rapidly growth. Many applications of WSNs didn’t work perfectly unless use mobile WSN. In mobile WSN that use clustering routing protocol, the location of the sensor nodes may be changed during the runtime which may lead to lose out the connection between cluster head (CH) and cluster members. Clustering protocol in mobile WSN should handle the challenges of static WSN like energy limitation and network life time as well as the challenges of mobile WSN like reliability and success data rate. In this paper, a protocol has been proposed for mobile WSNs performs clustering with dynamic number of clusters depending on the instant network parameters. Modified genetic algorithm has been used to select the optimum number of clusters and elect the suitable cluster head for each cluster. The target of fitness function that used in genetic algorithm is to minimize the consumed energy in the round and reduce the distance between cluster members and its cluster head. Balancing the residual energy in the sensor node is important factor to prolong the network lifetime. Balance filter has been used to ensure the energy balance between the nodes by blocking low energy nodes from working as a cluster head. The simulation result shows that the proposed Mobile Genetic Balanced Clustering protocol (GBC-M) outperforms the common mobile WSN clustering protocols in almost evaluation metrics for different network environments. Keywords— Mobile WSN ,MWSN, clustering, genetic algorithm, load balance, dynamic number of CHs.
I. INTRODUCTION Sensing is the technique used to gather information about the state of physical conditions like temperature, animations, sounds, lights …etc. The device used to perform the sensing task is called “sensor”. Each sensor node is equipped with a processor and memory to process and store the sensed data, transceiver to send the processed data to central sink point called “Base Station” (BS) and an energy source (battery). Sometimes the sensor node equipped with additional component -depends on the application- like “Mobilizer” to add mobility feature to the node. After the nodes deployment, it is impossible to recharge the battery. Efficient energy management is an important factor to prolong the network lifetime. Clustering model is an efficient and popular solution to minimize transmission energy [1]. This model divides the nodes into clusters. Each cluster has one head that collects the data from the member nodes and send it to the Base Station (BS). To prolong the network lifetime and minimize the power
consumption in Mobile wireless sensor network, a Clustering protocol based on a Genetic algorithm with some modification in addition to a filter to Balance the energy of the nodes been presented in this paper (GBC-M). The GBC-M protocol selects the dynamic number of clusters and elects suitable Clusters Heads. The following sections deal with the Related Work (II), Proposed Algorithm (III), Assumptions and Simulation Results (IV) and Conclusions (V). II. RELATED WORK There are many researches in WSN fields especially in minimizing the required power in data transmission. Low Energy adaptive Clustering Hierarchy (LEACH) [2] [3] protocol is one of the common WSNs routing protocols. LEACH protocol use clustering technique and suggested random cluster head selection method based on the desired percentage of cluster heads. The main drawback in LEACH protocol is the use of evenly distribution method in CH selection based on priori probability regardless of the network structure and the actual nodes energy. D. Kim et al. [4] enhances LEACH protocol and proposed “LEACH-M protocol to handle the challenges of mobile WSN. It assumes that the radio transmitter has limited radio transmission range and considers threshold power for successful packet receives. Although that LEACH-M protocol adds overhead to confirm communication link between cluster members and cluster heads, it outperform LEACH protocol in term of successful data packet. G. Kumar et al. [5] enhanced LEACH-M protocol by proposing LEACH-M Enhancement protocol (LEACHME). Cluster head has been elected in LEACH-ME based on the remoteness of the node. The sensor node with minimum remoteness metric is elected as a cluster head. LEACH-ME protocol adds more overhead compared with LEAH-M protocol but it ensures high success rate. J. Long et al. [6] proposed “LEACH-GA. It was based on LEACH protocol and use the genetic algorithm to specify the optimum cluster head probability. LEACH-GA suggests adding propagation phase before the setup phase in the first round only. In the propagation phase the BS applies genetic algorithm to search for the optimum probability of cluster heads and broadcast it to all nodes. The sensor nodes use this optimal probability in clustering formation during the next rounds as it in LEACH protocol. LEACH-GA is useful in determining the optimum percentage of CHs for different BS positions but it didn’t
978-1-5386-0802-9/17/$31.00 ©2017 IEEE
1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017 support dynamic number of CHs at the runtime. M. Chourasia et al. [7] suggested a routing protocol for static and mobile WSNs by mixing between direct communications and clustering technique. The node which is near to the BS communicates directly with the BS and the nodes which far away from the BS use the clustering technique to communicate with the BS.
III.
PROPOSED ALGORITHM
In this paper, a Clustering protocol based on a Genetic algorithm with some modification and a Balance filter for Mobile wireless sensor network has been proposed to minimize power consumption of the sensor nodes and prolong network lifetime. The operations of the proposed (like other clustering protocols [8]) are broken to rounds. Each round consists of two phases: setup phase and steady state phase. In the setup phase, the network is divided to clusters and the sensor nodes are classified to cluster heads and cluster members. In steady state phase, the sensor nodes began to transmit its data to the Cluster heads.
data from all the nodes in the cluster and these operations would add more overload on cluster head energy source. Therefore it’s so important to periodically change the cluster heads list in the network to balance the load between the nodes. The main objective of the balance filter is dividing the load between the nodes in order to prolong the network stable time. Network stable time can be declared as the time from the beginning of the network operations until the dead of the first node in the network. The Proposed balance filter [9] in mobile WSN based mainly on three parameters: Residual node energy, estimation of the remaining rounds and the behavior history of the nodes in the network. Residual energy is send from the node to base station at the beginning of each round. It is clear that the residual energy is important to prevent the nodes with low residual energy from being cluster head. Estimation of the remaining rounds is important to test whether if the specific node has enough power to stay alive for the remaining rounds or not. Remaining rounds can be calculating in the BS as
Re mRounds =
ETotal E min Round
(1)
A. Setup phase:
The following steps explain the operation of this phase: • Each sensor nodes send small message to the BS contains its location and the value of residual energy. • The BS block the low energy nodes from being a cluster head and select other nodes as a candidate Cluster Head by using a Balance Filter as explained bellow. • A genetic algorithm with some modification is applied to choose the optimum number of clusters and elect the final cluster heads from the candidate CHs list. • The base station broadcast the routing information to inform each node with its function rule, cluster head or cluster member.
Where ETotal represent the total residual energy of all nodes in the current round and EminRound represents the minimum energy dissipated from all nodes during single round. The history of the node here means the amount of energy dissipated by the node in each previous rounds. The history of the node is useful to estimate whether the node has the ability to stay alive for the remaining round or not. At the beginning of each round the base station receive the residual energy of all nodes therefore the base station has the history of the all nodes. The proposed balance filter is used to prevent low energy nodes from being cluster head and it work as follows: Candidate CH
• Every cluster head will broadcast advertising message to all sensor nodes. • Cluster Members receive advertising message from different CHs and send back join request to the head with the highest received signal strength Indication (RSSI) which is normally the nearest CH. • CH build TDMA schedule according to the number of join request then send the TDMA schedule to its members. When the TDMA schedule is received from the nodes, the setup phase finished and the steady state phase is ready to start. In clustering protocols, cluster members consume low energy in data transmission compared with cluster heads. This is because that the cluster members send its sensed data to the nearest cluster head (normally with small distance ) while the cluster head send its data, in a single hop technique, directly to the base station (normally with large distance). Furthermore, the base station is responsible for receiving and aggregating
if equation (3) true
(2) Filter(S(i)) = Member Node
if equation (3) false
[S(i).E >= ETotal/ N ] OR [ S(i).E >(RemRounds *S(i).EMaxMN)+ S(i).EMaxCH] (3) Where: S(i).E: amount of energy in sensor node (i). ETotal: Total network energy in all nodes in the current round N : The Number of sensor nodes in the network S(i).EMaxMN : Maximum energy consumed from node i when work as a Member node. S(i).EMAXCH : energy consumed from node i in Last times working as a cluster head.
1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017 Since the nodes density in mobile WSN may be changed in any period of network operation time due to node mobility and this will effect on the value of the consumed energy. EMaxMN, EMaxCH are used instead of ELastMN, ELastCH to cover the worst case of power consumption in different network structure.
B. Genetic Algorithm Genetic Algorithm (GA) is a modern optimization method that mimics natural selection process based on the concept of survival to fittest. It consists of main three operations: Selection, crossover and mutation. The population in GA consists of number of chromosomes and each chromosome represents one possible solution. Selection step is responsible of preferring the best chromosomes by using objective function. In this work the problem of selecting an optimum network structure in MWSN has been solved based on genetic algorithm. The encoding representation of chromosomes is implemented in two common method integer and binary encoding method. This work assumes successful threshold distance 𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 so if the distance between the cluster head and the cluster member greater than 𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 the connection will be lost. In such cases, routing protocol should use multi objective function optimization technique. The proposed objective function should prefer the network structure that makes the connection distance between the nodes less than 𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 to increase success data rate as well as minimizing energy consumption. Evaluation of each chromosome is done by using a special fitness function that simulate the work of WSN and calculate the energy dissipated form all nodes in a single round as in equation (4) Fitness function =
𝟏 (𝐑𝐍𝐍𝐮𝐧𝐍𝐍𝐄𝐧𝐍𝐍𝐅𝐅𝐠𝐲 )
(4)
Where RoundEnergy is the energy required from all nodes to operate the sensor network for a single round. In the proposed mobile protocol impact distance are used instead of real distance in calculation of RoundEnergy to increase success data rate. Equation (5) explains the proposed impact distance calculation. 𝒅𝒊𝒎𝒑𝒂𝒄𝒕 = 𝒅𝒓𝒆𝒂𝒍 + 𝟎. 𝟓 ∗ 𝑴𝒊𝒏(𝒅𝒓𝒆𝒂𝒍 , 𝒅𝒓𝒆𝒂𝒍 ∗ � Where :
𝒅𝒓𝒆𝒂𝒍
𝒅𝒔𝒖𝒄𝒄𝒆𝒔𝒔
𝟑
� ) (5)
𝑑𝑖𝑚𝑝𝑎𝑐𝑡 will be approximately equal to 𝑑𝑟𝑒𝑎𝑙 but when the
value of 𝑑𝑟𝑒𝑎𝑙 near to 𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 the value of 𝑑𝑖𝑚𝑝𝑎𝑐𝑡 become close to 1.5 𝑑𝑟𝑒𝑎𝑙 .
By using equation (5) in the fitness function calculation, the genetic algorithm will prefer chromosomes with small distance transmission over that with large distance and the network will become more reliable against nodes mobility. Special crossover operation has been used in proposed protocol[9] which swaps single gene in the parents’ chromosomes at random position, then -in case of integer encoding- replacing the duplicated genes in offspring chromosomes with zero. Zero genes considered as member nodes and it use to allow proposed protocol to dynamically select the optimum number of CHs. Fig.1 demonstrate single gene crossover operation. C. Steady state phase The operations of steady state phase in clustering WSN are broken to frames. In each frame all the nodes send their sensed data to the cluster head with a specific time slot allocated in the TDMA schedule. In the proposed mobile clustering protocol, data transmissions begin with CH request. Cluster head send data-request message to each member nodes at the beginning time of its time slot. If the mobile node received the datarequest message, the node will replay to it by data-response message which contain the sensed data. If the mobile node does not receive the data-request message, the node will not transmit its data at this frame and it should wait the datarequest message on the next frame .If the mobile node does not receive the request message also in the next frame, the mobile node will leave its CH and broadcast cluster join-request message to join with another cluster in the next frame. On the other hand, if the cluster head doesn’t receive any replay on its data-request message to a specific node for two times, the cluster head will consider that node becomes out of cluster region and delete it from TDMA schedule then generate and broadcast new TDMA schedule. Fig.2 demonstrates the Data transmission flowchart of the GBC-M steady state phase.
Parent
0
35
0
0
22
84
0
0
0
7
Parent
0
12
84
0
6
55
0
0
52
0
Min : Minimum function 𝑑𝑟𝑒𝑎𝑙 : Distance between two nodes
𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 : Threshold successful distance When the value of 𝑑𝑟𝑒𝑎𝑙 less than 𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠 with high
difference the ratio �
𝑑𝑟𝑒𝑎𝑙
𝑑𝑠𝑢𝑐𝑐𝑒𝑠𝑠
3
� become very small, the value of
Offspring
0
35
0
0
22
55
0
0
0
7
Offspring
0
12
84
0
6
0
0
0
52
0
Fig. 1. cross over operation
1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017 IV.
ASSUMPTION AND SIMULATION RESULT
A. Radio and Mobility Model To simulate the mobile WSN, the sensor nodes should be moved according to a specific mobility model. The simulation of sensor node movement should mimic the movements in real mobile WSN. There are different mobility models that can be used in mobile node simulation [ [10]]. In this research, Random Way Point Mobility Model has been adopted. In this model each node stays on its location for a random period (pause time) then it choose random position (x,y) inside the work area as a destination and random speed. The node move to its destination according to random position and random speed selected. If the node reach to its destination, it wait for new random pause time and the operation restart again. Fig.3 give an example of node movement according to random way point model. A radio model is used in this paper with the following equations [11]:
𝐸𝑇𝑥 (𝑘, 𝑑) = (a)
CH algorithm
𝐸𝑒𝑙𝑒𝑐 ∗ 𝑘 + 𝐸𝑓𝑠 ∗ 𝑘 ∗ 𝑑2
𝑑 < 𝑑0
𝐸𝑒𝑙𝑒𝑐 ∗ 𝑘 + 𝐸𝑚𝑝 ∗ 𝑘 ∗ 𝑑4
d > 𝑑0
𝐸𝑅𝑥 (𝑘) = 𝐸𝑒𝑙𝑒𝑐 ∗ 𝑘
(6)
(7)
Where 𝐸𝑒𝑙𝑒𝑐 : dissipated energy to run the electronic circuit for receiver and transmitter k: Packet size d: Distance between receiver and transmitter 𝐸𝑓𝑠 : Free space constant energy 𝐸𝑚𝑝 : Multipath constant energy 2
f
d0: the distance threshold expressed as: 𝑑0 = � s m
p
B. Simulation results This section presents the performance comparison between the proposed GBC-M protocols and LEACH-M and LEACHME protocols. The simulation is implemented by MATLAB program and uses the parameters of table (1).
(b)
Member node algorithm
Fig. 2. Data transmission flowchart for proposed mobile WSN protocol
Fig. 3. Movement in Random waypoint model [10]
1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017 `TABLE 1: PARAMETERS USED IN THE SIMULATION 300*300 meters
Field Dimensions
(150,150)
Base Station location Number of nodes (N)
100
packet size
4000 0.5 Joule
Initial energy (Eo)
50 nJ/bit
Eelec
10 pJ/bit/m2
Efs
0.0013 pJ/bit/m4
Emp Energy of data aggregation EDA
5 nJ/bit/packet
Genetic Initial population
20
Genetic crossover rate(Pc)
0.3
Genetic mutation rate (Pe)
0.1
Number of iteration
100
Mobility model
It’s clear from table 2 that the results of two types of encoding are convergent and this small difference may be due to random values used inside the genetic algorithm. The simulation result shows that Proposed mobile protocol (GBC-M) are outperform other tested protocols on several performance metrics. Fig.4 shows the number of live nodes per rounds where proposed Protocols has a greater network stable time. Network stable time is the time from starting the network process until the time of first node dies. It’s required to increase stability time as long as possible. Fig.5 shows the total number of successful send packets per rounds. The proposed protocol outperforms other tested protocols in this term.
Random waypoint model
Pause time
Random: [0-10] second Random : [0, 2π]
Sensor direction Sensor Speed
Random : [0, 10] m/sec
No. of frames per round
25 (150,150)
Base Station location
Proposed routing protocol is implemented using two types of genetic algorithm encoding integer and binary. To compare between these two types, random node distribution have been used as a design variable of algorithm and the values of the objective function are used to compare between the two encoding method as shown in table 2. TABLE 2: GENETIC ENCODING SIMULATION RESULT Network
Objective Function ( Binary)
Objective Function ( Integer)
Network 1
17.3921
17.3027
Network 2
15.1531
14.3134
Network 3
14.6763
14.7627
Network 4
17.2504
16.6851
Network 5
16.3453
16.7946
Network 6
17.3259
17.1494
Network 7
17.1801
15.5802
Network 8
16.0291
13.4403
Network 9
15.7578
16.0331
Network 10
17.4315
17.2753
Average
16.4542
15.9337
Fig. 4. Number of live nodes per rounds
Fig.. 5. Number of successful sent packets per rounds.
1st International Conference on Recent Trends of Engineering Sciences and Sustainability 17-18 May / 2017 Fig.6 shows the number of member node transitions. Node transition occurs when the member node lost the connection with its CH and join with new CH. Node transitions add an overhead to network which increase power consumption and reduce network live time. It is required to minimize node transition as minimum as possible. The simulation result shows that the proposed protocol has better performance than LEACH-ME and LEACH-M respectively. Fig.7 shows the number of failure send packets. It is required to minimize the number of failure send packets as minimum as possible.
TABLE 4. RESULT SUMMARY WHEN 50% OF NODES ARE MOBILE First Node Dies
Successful Failure Node Send Send E. C. R. Transition Packets Packets
LEACH-M LEACHME
19
83751
19981
1137
1.5023
𝑬𝑬. 𝑪𝑪. 𝑹𝑹. 𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅 𝐍𝐍𝐍𝐍𝐍𝐍𝐍𝐍 𝐃𝐃𝐃𝐃𝐃𝐃𝐃𝐃
21
90142
2432
571
1.5006
0.071457
GBC-M
32
95131
2135
541
1.3092
0.040913
Protocol
0.079068
V. CONCLUSION This paper proposed Mobile Genetic Balanced Clustering protocol (GBC-M). Simulation results shows that the propsed protocol outperforms LEACH-M and LEACH-ME protocols in terms of network stable time, energy consumption, successful send packets and number of node transission for different network environments. Simulation results show there is no significant difference between integer and binary genetic encoding method. Furthermore, it shows that the proposed protocol has efficient and stable performance with different percentage of nodes mobility unlike other tested protocols. REFERENCES [1] O. Younis, M. Krunz, and S. Ramasubramanian, "Node clustering in wireless sensor networks: recent developments and deployment challenges," network, IEEE, vol. 20, no. 3, pp. 20-25, May-June 2006. [2] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, "Energy efficient communication protocol for wireless microsensor networks," in 33rd Annual Hawaii International Conference on System Sciences, Vol. 2, Jan. 2000.
Fig. 6. Number of Node transitions per rounds
[3] Wendi B. Heinzelman, Anantha P. Chandrakasan, and Hari Balakrishnan, "An Application-Specific Protocol Architecture for Wireless Microsensor Networks," IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, vol. 1, no. 4, pp. 660-670, OCTOBER 2002. [4] D. S. Kim and Y. J. Chung, "Self-Organization Routing Protocol Supporting Mobile Nodes for Wireless Sensor Network," in Computer and Computational Sciences, 2006. IMSCCS '06. First International Multi-Symposiums, Hanzhou, Zhejiang, 2006, pp. 622-626. [5] G. S. Kumar, P. M. V. Vinu, and K. P. Jacob, "Mobility Metric based LEACH-Mobile Protocol," in 16th International Conference on Advanced Computing and Communications, Chennai, 2008, pp. 248253. [6] J. L. Liu and C. V. Ravishankar, "LEACH-GA: Genetic AlgorithmBased Energy-Efficient Adaptive Clustering Protocol for Wireless Sensor Networks," International Journal of Machine Learning and Computing, vol. 1, no. 1, April 2011. [7] M. K. Chourasia, M. Panchal, and A. Shrivastav, "Energy Efficient Protocol for Mobile Wireless Sensor Networks," in International Conference on Communication, Control and Intelligent Systems (CCIS), 2015, pp. 79-84.
Fig. 7. Number of failure send packets per rounds
The figures above are calculated if 100% of nodes are mobile. Tables 3 and 4 summarize the simulation result when 100% and 50 % of nodes are mobile respectively where E.C.R is the energy consumption rate which represents the average consumed energy in single round assuming that all protocols transmit the same amount of packets to the nearest CHs. TABLE 3. RESULT SUMMARY WHEN 100% OF NODES ARE MOBILE Protocol LEACH-M LEACHME GBC-M
First Node Dies
Successful Failure Node Send Send E. C. R. Transition Packets Packets
26
88791
16658
1552
1.4180
𝑬𝑬. 𝑪𝑪. 𝑹𝑹. 𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅𝐅 𝐍𝐍𝐍𝐍𝐍𝐍𝐍𝐍 𝐃𝐃𝐃𝐃𝐃𝐃𝐃𝐃
32
96034
7256
1449
1.3460
0.042063
34
100649
2191
872
1.2581
0.037003
0.054538
[8] N. Pradhan, K. Sharma, and V. K. Singh, "A Survey on Hierarchical Clustering Algorithm for Wireless Sensor Networks," International Journal of Computer Applications, vol. 134, no. 4, pp. 30-35, 2016. [9] Z. T. Alisa and H. A. Nassrullah, "Minimizing energy consumption in wireless sensor networks using modified genetic algorithm and an energy balance filter," in Multidisciplinary in IT and Communication Science and Applications (AIC-MITCSA), Al-Sadeq International Conference on. IEEE, Baghdad, 2016, pp. 1-6. [10] J. Rezazadeh, M. Moradi, and A. S. Ismail, "Mobile Wireless Sensor Networks Overview," JCCN International Journal of Computer Communications and Networks, vol. 2, no. 1, pp. 17-22, 2012. [11] T. , Rappaport, Wireless Communications: Principles & Practice. Englewood Cliff, 1996.
