Indoor WLAN localization using group sparsity optimization technique

Abstract

This paper proposes a novel approach for Wireless Local Area Networks (WLAN) indoor fingerprinting localization using a novel interpretation and optimization solution. The user's location is introduced as a sparse vector which can be estimated in a single minimization problem. The coarse localization is embedded in the fine localization to prevent the cases where the user's position is searched in wrong subset of reference points (RPs). First, the RPs are clustered in layers using the similarity between the online measurement and RP fingerprints via an AP coverage vector. Then, the localization approach is performed via a joint minimization of convex least squares minimization of residuals, l1-norm of the entire position vector, and weighted l2-norm of groups of RPs. The method has been evaluated though Monte Carlo simulation runs and the results showed a great positioning accuracy.