• No results found

In this study, a WSN of Y stationery, identical nodes is considered. The nodes are organised into a group of K Clusters, each with one CH coordinating cluster operations. A cluster is formed up of the CH itself, FFD and RFD nodes working as end devices. In order to enhance reliability and availability of the network, the CH operations are rotated among strategically deployed FF nodes. The choice of the CH is based on node energy levels, and other metrics deemed appropriate

[Hashmi et al.,2010], [Chiasserini and Garetto,2004], and [Li,2011]. To conserve

next CH. For this purpose, use of the best energy-saving protocols like UHEED

[Ever et al., 2012] is assumed. The system is also assumed to have redundant

sensor nodes deployed at inception but kept inactive until the need to replace a

failing node arises [Munir and Gordon-Ross, 2011]. It is further assumed that all

nodes are equipped with omnidirectional antennas with same radius (d) and can communicate directly with the CH based on the IEEE802.15.4/ZigBee standards. In order to reduce the energy consumption further, nodes are capable of choosing an arbitrary transmission power level as long as the radius d is not exceeded. Information sensed at the nodes is forwarded to the CH, which finalises clus- ter data aggregation. The CHs may also generate data packets based on their observations. The total information is then transmitted by the CH to the sink directly or through another intermediary CH. It is assumed that at least one

path exists towards the sink [Chiasserini and Garetto, 2004]. Like other com-

munication networks, this system is subject to failures, which may result from

hardware, software and channel link errors. Figure2.4 shows the system scenario

in consideration.

Figure 2.4: Network topology of the reference scenario

A closer look at the functions of the CH indicates several possible operative states, including active and sleep modes implemented in a couple of protocols in order to conserve the limited power available for node operations.

2.5.1

WSN Clustering

In most cases, the formation of WSNs is by deployment of large magnitudes of small sensor nodes in the habitats. Once deployed, the sensor nodes that are pre- loaded with necessary software then discover their neighbours and self-configure themselves into a desired network. Due to inability to cope with energy and re- source constraints of WSNs, hierarchical networks are preferred over flat networks. Lots of clustering algorithms have been developed in order to address the con-

strains of WSNs [Anker et al.,2008], [Mamun,2012], [Liu,2012], [Jadidoleslamy,

2013]. Clustering is preferred in WSN deployment because of its ability to achieve network scalability, energy efficiency, prolonged life time, reduced communication

overheads and ease of management in large scale WSNs [Jadidoleslamy, 2013].

This arrangement allows all the cluster nodes to have an opportunity to operate

as a CH in a rotational arrangement [Ameer Ahmed Abbasi,2007]. The rotation

has also been enhanced by the introduction of BCHs used for creating redun-

dancy in the event of failure before rotation time is reached [Hashmi et al.,2010],

[Ameer Ahmed Abbasi, 2007], [Gupta and Younis, 2003]. In every cluster, the

nodes communicate directly with the CH. They operate as end devices collecting data and transmitting to the CH, making the CH become the central point for data aggregation. The CH is also responsible for transmission of all cluster traffic to the base station directly or hopping through some other intermediary CHs. Some of the benefits of using clustering include route localization, which reduces

routing table stored at individual nodes [Akkaya and Younis,2005], conservation

of communication bandwidth by limiting inter-cluster interactions to CHs and avoiding redundant exchange of messages among sensor nodes [Ameer Ahmed Ab- basi,2007]. Clustering also benefits from the ability to stabilize topology at sen- sors’ level thereby reducing topology maintenance overhead. Under this, sensors would only concern themselves with connection to the CH since changes at the

inter-CH never affect them. [Ameer Ahmed Abbasi, 2007], [Hou et al., 2005].

Noting that the major concern is to guarantee reduced energy consumption and

longer life time, research by [Vlajic and Xia, 2006] indicated that a good choice

of a clustering scheme is necessary. In this paper, authors also recommend a maximum of five hops a CH can be deployed away from the sink. However in

every cluster, two hops is recommended for deployment of nodes away from the CH.

In [Ameer Ahmed Abbasi, 2007], a clear outline of clustering objectives is given.

These include increasing connectivity and reducing delay, which ensures a good path for connecting all CHs to the sink at reduced delay, minimum cluster count, which enables limiting number of hop counts to the sink and maximizing network longevity, which determines the lifetime of the network. Together with taxon-

omy of clustering attributes, Figure 2.4, this work depicts WSN clusters as a key

medium to the provision of WSN network QoS [Bhuyan et al., 2010] and con-

centrates on modelling integrated performance and availability of these networks when subjected to a number of metrics.

Clustering hierarchy continues to be used enormously because of the advantages it offers in the resource management and scalability of the network among others

[Jadidoleslamy, 2013]. However, QoS handling, mobility effects and redundancy

management for ensuring network reliability remains an important trade-off for

improving performance in WSNs [Liu, 2012].

2.5.2

Cluster Head Selection

The process of choosing the CH involves consideration of a number of metrics, which vary between protocols. Commonly used metrics include the initial and residual energies. During first deployment, same power level is assumed for all the nodes. Using the algorithms that vary from one protocol to another, a CH is selected among strategically placed contending nodes that are preferably 1-hop apart in the neighbourhood. Examples of the protocols used for clustering include

LEACH [Heinzelman et al., 2000], where CHs are rotationally selected from the

cluster nodes in order to distribute communication energy within the cluster to all the nodes. LEACH has since been improved a lot in order to enhance its

performance. The LEACH family include TL-LEACH [Loscri et al., 2005], E-

LEACH & M-LEACH [Xiangning and Yulin, 2007], V-LEACH [Yassein et al.,

Another clustering protocol is the HEED [Younis and Fahmy,2004] which chooses the CH based on the nodes residual energy and the intra-cluster communication cost. In addition, HEED ensures even distribution of the CH throughout the network. Like in the case of LEACH, HEED also does CH election periodically. As an improvement to HEED, Distributed Weight-based Hierarchical Clustering Protocol (DWEHC) was developed. DWEHC improves on HEED by building balanced cluster sizes through creation of a multi-level structure for intra-cluster communication while at the same time limiting the number of sensor nodes that can be attached to a parent node. Other clustering protocols include Position- based Aggregation Node Election protocol (PANEL), Unequal Clustering Size (UCS), Energy Efficient Clustering Scheme (EECS) and others have been devel-

oped. A detailed summary of the clustering protocols is presented in [Liu,2012].

Further details on clustering is presented in [Anker et al., 2008], [Mamun, 2012]

and [Jadidoleslamy, 2013].