An Improved Localization Framework Based on Maximum Likelihood for Blind WSN Nodes

Abstract

Localization is a way to find the blind node's position through some anchor nodes whose positions are known before, which is the essential issue of wireless sensor networks (WSN). Among all the localization algorithms, maximum likelihood localization (ML) algorithm based on received signal strength (RSS) is more accurate than most of other algorithms. However, ML algorithm needs to compute conjugate gradient in multiple iterations to maximize a likelihood equation, which slows down the localization process. To speed up the location process based on ML without losing accuracy, a localization framework which combines ML with early estimation method is proposed in this paper. Weighted centroid localization (WCL) method cost far less time than most other algorithms and is suitable to do early estimation. In our framework, a modified weighted centroid localization (MWCL) method is proposed to do early estimation. The simulations demonstrate that this localization framework outperforms classic ML method in terms of localization speed. Moreover, The accuracy of the localization framework in terms of mean square error (MSE) of estimation is close to the theoretical lower bound, i.e. Cramer-Rao low bound (CRLB).