On Data Fusion for Parametric-Model-Based Wireless Localization

In this paper, we present a data fusion framework for parametric-model-based wireless localization where the mobile station location is treated as a deterministic unknown vector. Three types of fusion schemes are presented: measurement fusion, estimate fusion and mixed fusion. Theoretical performance comparison among these schemes in terms of the estimation root mean square error via the weighted least square estimator (WLSE) is conducted. Such a performance metric coincides with the Cramer-Rao lower bound (CRLB) in the case of Gaussian noise. We show that, if the raw measurement vectors are correlated, then measurement fusion achieves the best performance, mixed fusion follows and estimate fusion is the worst. If the raw measurement vectors are uncorrelated, then these different fusion schemes achieve the same performance. Benefits that can be earned from data fusion for wireless localization are also investigated and numerical examples are presented to validate our theoretical analysis.