WSN201201孟加拉国

无线传感器网络wireless sensor network

Wireless Sensor Network, 2012, 4, 18-24

doi:10.4236/wsn.2012.41003 Published Online January 2012 (doc.docsou.com)

Multiple Parameter Based Clustering (MPC): Prospective Analysis for Effective Clustering in Wireless Sensor

Network (WSN) Using K-Means Algorithm

Asif Khan, Israfil Tamim, Emdad Ahmed, Muhammad Abdul Awal

Department of EECS , North South University, Dhaka, Bangladesh

Email: alaxwest@yahoo.com, israfiltamim@hotmail.com, {emdad, abdulawal}@northsouth.edu

Received October 15, 2011; revised November 24, 2011; accepted December 21, 2011

ABSTRACT

In wireless sensor network cluster architecture is useful because of its inherent suitability for data fusion. In this paper we represent a new approach called Multiple Parameter based Clustering (MPC) embedded with the traditional k-means algorithm which takes different parameters (Node energy level, Euclidian distance from the base station, RSSI, Latency of data to reach base station) into consideration to form clusters. Then the effectiveness of the clusters are evaluated based on the uniformity of the node distribution, Node range per cluster, Intra and Inter cluster distance and required energy level of each centroid. Our result shows that by varying multiple parameters we can create clusters with more uniformly distributed nodes, minimize intra and maximize inter cluster distance and elect less power consuming cen-troid.

Keywords: K-Means Algorithm; Energy Efficient; Uniform Distribution; RSSI; Latency

1. Introduction

Wireless sensor networks (WSN) are highly distributed networks of autonomous small, lightweight sensors (nodes) in large numbers to monitor physical or environmental conditions by the measurement of, temperature, sound, vibration, pressure, motion or pollutants and to coopera- tively pass their data through the network to a main loca- tion (often called a sink).

It has already made its way in military object, habitat monitoring [1] and object tracking because of the char-acteristics such as feasibility of rapid deployment, self- organization (different from Infrastructure Mode or ad hoc network [2]) and fault tolerance. But limited compu- tation capability, limited power and small memory size has made designing the WSNs is very difficult [3]. The energy consumption is the most important factor among these three factors, because the battery is not changeable if once the sensor nodes are deployed. The energy is also the major consideration in designing the routing of the WSNs. Hierarchical protocols reduce energy consump- tion in the networks by clustering. Clustering algorithms partition data objects (patterns, entities, instances, obser- vances, units) into a certain number of clusters (groups, subsets, or categories).

Several available operational definitions [4] summa- rized by Everitt (1980) are as follows:

“A cluster is a set of entities which are alike, and enti-ties from different clusters are not alike.”

A cluster is “an aggregate of points in the test space such that the distance between any two points in the cluster is less than the distance between any point in the cluster and any point not in it.”

“Clusters may be described as continuous regions of this space (d-dimensional feature space) containing a rel- atively high density of points, separated from other such regions by regions containing a relatively low density of points.”

In these protocols, nodes are divided into some clus-ters and some nodes based on some parameter are the selected as cluster heads (CH). These cluster heads ex-change data with the Base station (BS) which costs the most energy of the nodes. Even though this concept has similarity with the Basic Service set (BSS) of Infrastruc-ture mode where there is an Access point (AP) [5] (here the CH) and few cells communicate via this access point, the method in WSN is much more dynamic and energy efficient. Due to these advantages, sensor nodes can re- markably save their own energy.

In this paper by we have proposed a new method called MPC (Multiple Parameter based Clustering) using k-means clustering algorithm and variation of parameter and thus made four contributions:

1) We can have control over the random node distribu-

第1页下一页