Controversy CornerGenetic algorithm for energy-efficient clustering and routing in wireless sensor networks
Introduction
Wireless sensor networks (WSNs) integrate sensor technology, embedded computing technology, distributed information processing technology, and communication technology. WSNs have been employed widely in various fields, including military, national defence, environmental monitoring, traffic management, health care, manufacturing, and disaster prevention applications (Leu et al., 2015). In a WSN, the sensor nodes collaboratively collect and process environmental and physical information from the area covered by the network and send the information to observers (Akyildiz et al., 2002). A monitoring area often requires the deployment of a large number of sensor nodes, but the sensor nodes are limited in terms of their computational, storage, and communication capacities in order to reduce costs. For instance, Micaz (Ali et al., 2012) developed by CrossBow is a representative type of sensor node, which is equipped with an Atmegal28L microprocessor, CC2420 chip, 128 KB Flash, and 4KB RAM. The sensor node is supplied with limited battery power and it is difficult to provide secondary energy to the nodes. Thus, network failure occurs after more than a certain percentage of the nodes die. Therefore, reducing the energy consumption of sensor nodes and prolonging the network life-cycle is the key challenge for WSNs.
Previous studies have shown that the energy consumption required for transferring 1-bit is much more than that for processing 1-bit data (Abbasi and Younis, 2007). Thus, reducing the transmitted or received data sizes for sensor nodes and optimizing data transmission routing between the nodes can effectively reduce the energy consumed by the network. Clustering algorithms (Singh et al., 2016a, Singh et al., 2016b) divide the network into multiple independent clusters, where each cluster comprises a cluster head (CH) node and multiple cluster member (CM) nodes. The CH node is responsible for receiving data from the CM nodes. By using effective data aggregation algorithms, the CH can remove redundant or incorrect data so the large amounts of collected data are merged into a small amount of meaningful information. Therefore, efficient clustering algorithms can reduce the data traffic and optimize the topology, thereby improving the energy efficiency of WSNs.
Cluster-based WSNs usually comprise two types: (i) those with temporary CHs (Heinzelman et al., 2000b) and (ii) those with permanent CHs (Mohapatra and Behera, 2012). In the first type, the sensor nodes have a relatively fair energy supply and equal status, and all the nodes have a chance of being selected as the CH. For instance, at moment t1, node A is selected to become the CH. However, at moment t2, node A may only act as the CM. In the second type, the CH nodes and CM nodes are permanent. The CH nodes are also called gateway nodes or relay nodes, which have a higher energy supply. During the operation of the network, the permanent CHs manage all of the sensor nodes in their corresponding clusters. The status of the CHs is stable, but the relationships between each CH and its CMs may change over time. For instance, at moment t1, the CH of CM a is node A. However, at moment t2, node a is likely to join CH B. We need to employ clustering algorithms with different mechanisms according to the two different types of cluster-based WSNs. In this study, we consider a clustering algorithm for the second type: WSNs with permanent CHs. Most previous studies of clustering algorithms have considered the first type, and the second type has rarely been investigated. The second type of WSNs are important for the research community because of the following reasons: (i) they are more energy efficient because CH nodes with a higher computing capacity can efficiently operate complex data aggregation algorithms; (ii) they are more secure because the CHs possess more storage and they have a higher capacity, and hence, complex encryption algorithms can be executed and some trusted hardware modules can be equipped (Wang et al., 2016); and (iii) the ordinary nodes only need to send the collected data to their corresponding head nodes without conducting the work of CHs, and thus the network life cycle can be prolonged.
It is important to note that devising an effective clustering algorithm with high energy efficiency and load balancing for the second type of large-scale WSNs is an NP-hard problem. If the network has a CH nodes and b ordinary nodes, then there will be ab clustering schemes. If routing is considered and the average number of neighbouring nodes of each CH is c, then there will be ca routing schemes. Therefore, for large-scale WSNs, calculating the optimal clustering and routing scheme has high time complexity. Metaheuristic algorithms (Yang, 2008) such as genetic algorithms (GAs) (Goldberg, 1989) can solve this problem quickly and efficiently.
In this study, we propose a GA-based energy-efficient clustering and routing algorithm (GECR), which employs a GA to obtain the optimal solution. The main contributions of this study are as follows:
- •
The network time is divided into multiple rounds. The sink node needs to run the GA separately in each round. The optimization objective for the clustering and routing schedule is related to the distances between the nodes. The locations of the nodes are fixed, so the optimal solution for a certain network round is related to the optimal solution for the previous round. In contrast to some traditional GA-based clustering and routing algorithms (Elhoseny et al., 2015, Lai et al., 2007, Rao and Banka, 2015), we add the optimal solution from the previous network round to the initial population in the current network round to improve the search efficiency.
- •
In order to guarantee the energy efficiency, the fitness functions of some algorithms (Gupta and Jana, 2015, Gupta et al., 2013, Shokouhifar and Jalali, 2015) are based on the total transmission distances between the nodes. The energy consumption is related to but not absolutely equal to the distance. Algorithms that construct the fitness function based on the total distance can only obtain the final solution with the shortest total distance. In the proposed method, we combine the clustering and routing scheme into a single chromosome and calculate the total amount of energy consumed for clustering and routing together. The fitness function is constructed directly based on the energy consumption of the whole network. In this manner, we can finally obtain the solution with the lowest energy consumption, and thus the final energy efficiency can be improved.
- •
Load balancing is an important criterion for evaluating clustering and routing algorithms. Many GA-based algorithms (Gupta et al., 2013, Liu and Ravishankar, 2011, Peiravi et al., 2013, Shokouhifar and Jalali, 2015) do not consider load balancing. In addition, although some algorithms (Gupta and Jana, 2015, Kuila et al., 2013, Liao et al., 2013) do consider load balancing, they only count the number of CMs when calculating the loads on CHs. However, the CH needs to transmit the data from its previous hop nodes in addition to the data from its CMs. In the proposed method, we add the previous hops to the loads on each CH, thereby improving the accuracy of the load calculations.
- •
Simulations indicated that the performance of our proposed algorithm was better than that of some existing algorithms (Gupta and Jana, 2015, Gupta et al., 2013, Kuila et al., 2013, Liu and Ravishankar, 2011, Shokouhifar and Jalali, 2015) in terms of load balancing, network life cycle, and energy consumption. Thus, the proposed GECR always had the lowest variances in the loads on the CHs under different scenarios. In terms of the network life cycle, GECR had the most living nodes at most times. In addition, GECR consumed the smallest amount of energy in all of the network rounds.
The remainder of this paper is organized as follows. In Section 2, we discuss related research. In Section 3, we present the network model and some terminology. In Section 4, we give some preliminary details used in this study. In Sections 5 and 6, we present the proposed algorithm and the experimental results. In Section 7, we give our conclusions.
Section snippets
WSNs with temporary CHs
Many studies (Al-Karaki and Kamal, 2004, Boyinbode et al., 2010, Kumar et al., 2011, Raja et al., 2017, Rostami et al., 2017) have investigated clustering and routing algorithms for WSNs with temporary CHs. The low-energy adaptive clustering hierarchy (LEACH) algorithm (Heinzelman et al., 2000b) is one of the best known hierarchical routing protocols based on clustering, where it divides the network time into multiple rounds. In each round, all of the sensor nodes start to calculate a
Network model and terminology
In this section, we first introduce the network topology and energy model for the proposed GECR algorithm, before providing the terminology used in this study.
Overview of GA
A GA is a type of metaheuristic that searches for an optimal solution by simulating the natural process of evolution. Fig. 3 illustrates the general process employed by a GA. First, some solutions are initialized randomly to form an initial population. These solutions are called individuals and each solution comprises one or more chromosomes made of a set of characters or strings. In an individual chromosome, each unit (a character or a string) is called a gene. After generating the initial
Proposed algorithm
Next, we present the detailed design of the proposed GECR algorithm. The chromosome representation (Section 5.1) is introduced first. We then present the population initialization method (Section 5.2) and the fitness function (Section 5.3). Finally, the crossover operator and mutation operator are explained in Section 5.4.
Experimental results
In experiments, we compare the proposed GECR algorithm with five clustering and routing algorithms: GACR (Gupta and Jana, 2015), GAR (Gupta et al., 2013), ASLPR (Shokouhifar and Jalali, 2015), LEACH-GA (Liu and Ravishankar, 2011) and GALBCA (Kuila et al., 2013). These algorithms are all based on GA.
Conclusions
In this study, we proposed a GECR algorithm to calculate globally the total energy consumed by all sensor nodes where the algorithm encodes the clustering scheme and routing scheme together in the same chromosome. GECR treats the total energy consumed by all nodes as a parameter in the fitness function. In cluster-based WSNs, CHs need to transmit the data from the previous hop nodes in addition to the data from the CMs. In the proposed method, we add the previous hops to the loads on each CH,
Tianshu Wang, was born in Jiangsu Province, China, in 1989. She received her Ph.D. candidate of School of Computer Science and Engineering in Nanjing University of Science and Technology in China. She is currently a lecturer of School of Information Technology in Nanjing University of Chinese Medicine. Her main research interests include embedded system, wireless sensor networks, clustering and routing, and trusted computing.
References (59)
- et al.
Survey on clustering algorithms for wireless sensor networks
Comput. Commun.
(2007) - et al.
Wireless sensor networks: a survey
Comput. Netw.
(2002) - et al.
Two-tier particle swarm optimization protocol for clustering and routing in wireless sensor network
J. Netw. Comput. Appl.
(2015) - et al.
An adaptive clustering approach to dynamic load balancing and energy efficiency in wireless sensor networks
Energy
(2016) - et al.
A novel evolutionary approach for load balanced clustering problem for wireless sensor networks
Swarm Evol. Comput.
(2013) - et al.
Energy efficient load-balanced clustering algorithm for wireless sensor networks
Procedia Technol.
(2012) - et al.
Energy efficient clustering and routing algorithms for wireless sensor networks: particle swarm optimization approach
Eng. Appl. Artif. Intell.
(2014) - et al.
Efficient load-balanced clustering algorithms for wireless sensor networks
Comput. Commun.
(2008) - et al.
A new evolutionary based application specific routing protocol for clustered wireless sensor networks
AEU - Int. J. Electron. Commun.
(2015) - et al.
Deployment of MICAz mote for wireless sensor network applications
Routing techniques in wireless sensor networks: a survey
IEEE Wirel. Commun.
LEACH-MAC: a new cluster head selection algorithm for wireless sensor networks
Wirel. Netw.
A survey on clustering algorithms for wireless sensor networks
A distributed clustering scheme with self nomination: proposal and application to critical monitoring
Wirel. Netw.
Ant colony system: a cooperative learning approach to the traveling salesman problem
IEEE Trans. Evol. Comput.
Developed distributed energy-efficient clustering (DDEEC) for heterogeneous wireless sensor networks
Balancing energy consumption in heterogeneous wireless sensor networks using genetic algorithm
IEEE Commun. Lett.
Characterizing the topology of an urban wireless sensor network for road traffic management
IEEE Trans. Veh. Technol.
Load-balanced clustering of wireless sensor networks
Energy efficient clustering and routing algorithms for wireless sensor networks: GA based approach
Wirel. Personal Commun.
GAR: an energy efficient GA-based routing for wireless sensor networks
Low energy adaptive clustering hierarchy with deterministic cluster-head selection
An application-specific protocol architecture for wireless microsensor networks
IEEE Trans. Wireless Commun.
Energy-efficient communication protocol for wireless microsensor networks//System sciences, 2000
Directed diffusion for wireless sensor networking
IEEE/ACM Trans. Netw. (ToN)
Performance analysis of hierarchical and flat network routing protocols in wireless sensor network using Ns-2
Int. J. Model. Optim.
Particle swarm optimization
Cited by (104)
A multi-objective evolutionary approach towards automated online controlled experiments
2023, Journal of Systems and SoftwareAn evolutionary routing protocol for load balancing and QoS enhancement in IoT enabled heterogeneous WSNs
2023, Simulation Modelling Practice and TheoryA literature study and performance gaps on Centralized server-based load balancing and routing strategies under cloud-IoT-WSN
2024, Australian Journal of Electrical and Electronics Engineering
Tianshu Wang, was born in Jiangsu Province, China, in 1989. She received her Ph.D. candidate of School of Computer Science and Engineering in Nanjing University of Science and Technology in China. She is currently a lecturer of School of Information Technology in Nanjing University of Chinese Medicine. Her main research interests include embedded system, wireless sensor networks, clustering and routing, and trusted computing.
Gongxuan Zhang, was born in Jiangxi Province, China, in 1961. He received his Ph.D. in School of Computer Science and Engineering from Nanjing University of Science and Technology, Nanjing, China. He is currently a professor in the School of Computer Science and Engineering at the Nanjing University of Science and Technology. His research interests include cloud computing, trusted computing, wireless sensor networks.
Xichen Yang, was born in Jiangsu Province, China, in 1989. He received his Ph.D. in School of Computer Science and Engineering in Nanjing University of Science and Technology in China. He is currently a lecturer of School of Computer Science and Technology in Nanjing Normal University in China. His main research interests include Image quality assessment, Image processing and computer vision.
Ahmadreza Vajdi, was born in Iran, in 1988. He is a Ph.D. candidate of School of Computer Science and Engineering in Nanjing University of Science and Technology in China. His main research interests include clustering and routing, wireless sensor networks.