On Performance of Node Placement Approaches for Hierarchical Heterogeneous Sensor Networks

  • Authors:
  • Santhosh Pandey;Shaoqiang Dong;Prathima Agrawal;Krishna M. Sivalingam

  • Affiliations:
  • Department of ECE, Auburn University, Auburn, USA 36849 and Cisco Systems, San Jose, USA;Department of ECE, Auburn University, Auburn, USA 36849;Department of ECE, Auburn University, Auburn, USA 36849;CSEE Department, University of Maryland, Baltimore County (UMBC), Baltimore, USA 21250 and Indian Institute of Technology Madras, Chennai, India

  • Venue:
  • Mobile Networks and Applications
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper considers a two-tier hierarchical heterogeneous wireless sensor network using the concept of clustering. The network has two type of nodes: regular sensor nodes (litenodes or LN) with limited communications, storage, energy, and computation power; and high-end sophisticated nodes (SNs), or clusterheads, with significantly additional resources. The litenodes communicate their data to the SNs and the SNs forward all collected data to a central gateway node called the base station (BS). Our network architecture allows the LNs to reach a SN via multiple hops through other LNs. We investigate the problem of optimally placing a minimum number of sophisticated nodes to handle the traffic generated by the lite nodes, while ensuring that the SNs form a connected network using their wireless links. This placement problem is formulated and solved as multi-constraint optimization problem using well known approaches: Binary Integer Linear Programming (BILP) approach, Greedy approach (GREEDY) and Genetic Algorithm (GA) approach. It was found through simulations that BILP performed best for regular grid topologies, while GA performed better for random LN deployment. Furthermore, the effects of various parameters on the solution are also presented. The paper also proposes a HYBRID approach that uses the solutions provided by GREEDY and/or BILP as the initial solution to the GA. Using HYBRID, results comparable to original GA could be obtained in only 11.46% of the time required for the original GA.