Adaptive Membership Selection Criteria using Genetic Algorithms for Fuzzy Centroid Localizations in Wireless Sensor Network

Abstract

This paper investigates the effect of fuzzy inputs, i.e., signal strength, of various known nodes, to fuzzy logic systems in order to derive a proper weight for Centroid, properly used to approximate the location in wireless sensor networks with its key advantage on simplicity but with precision trade-off. Due to a fluctuation behavior of location estimation precisions with respect to a diversity of various inputs, here, we propose the use of heuristic approach applying genetic algorithms with mutation and cross-over steps to adaptively seek the optimal solution – a proper number of membership functions for fuzzy logic systems in weighted Centroid – to achieve higher location estimation accuracy. The performance of our methodology is effectively confirmed by the intensive evaluation on a large scale simulation in various topologies and node densities against fixed membership function scenarios including a traditional Centroid