Localization with Cellular Signal RSRP Fingerprint of Multiband and Multicell

Precisely predicting the location of the user in a Global-Navigation-Satellite-System-degraded environment is a highly challenging task. Localization based on cellular signal fingerprints is one of the promising solutions to this problem and has attracted increasing attention. Long Term Evolution (LTE) signal is popularly utilized for localization due to its global usage, extensive urban coverage, and favorable signal properties. This paper proposes a new multiband multicell Reference Signal Received Power (MBMC-R) fingerprint, which properly fuses LTE signals’ carrier band information, the physical cell identifier information, and RSRP values. Next, a sequential block-matching weight
K
nearest neighbor algorithm with a cosine similarity criterion is specially designed for performing the pattern-matching localization with the MBMC-R fingerprint. The proposed method also includes the derivation of the Cramer-Rao lower bound, which reveals the impact of various factors on the lower bound of position error. Simulation and on-field experiments prove the performance superiority over other fingerprint localization algorithms reported in the literature.