Fingerprint-Based Localization Performance Analysis: From the Perspectives of Signal Measurement and Positioning Algorithm

This article analyzes the localization performance of fingerprinting positioning system from the perspectives of the signal measurement and positioning algorithm. Unlike the existing works which have not taken the influence of grid size into account, first of all, we construct a novel derivation model involving the grid size information. Then, from the signal measurement's perspective, the localization performance is analyzed based on our new model under two cases: with specific and nonspecific signal distributions. For the first case, we utilize the traditional knowledge of Cramér-Rao lower bound (CRLB) to rededuce it. For the second case, a Gaussian-Markov theorem method is introduced to conduct derivation. Furthermore, from the latter perspective, we first analyze the localization performance of the mostly used k-nearest neighbors (KNN) algorithm leveraging the probability density function (PDF) of these nearest neighbors. Then, a novel adaptive KNN algorithm is designed based on the derivation result, which has improved the location accuracy by about 20%. Finally, extensive simulations and real experiments are conducted to show the effectiveness of our claims.