Near-Field RIS-Aided Localization Under Deliberate Model Misspecification: Bounds and Algorithms

Artikel i vetenskaplig tidskrift, 2026

This paper investigates the optimization of reconfigurable intelligent surfaces (RIS) for near-field user equipment (UE) localization in the presence of channel spatial non-stationarity (SNS) across elements and the involved deliberate model misspecification. Traditional far-field localization techniques struggle to maintain accuracy in near-field scenarios characterized by spherical wavefronts (SWFs) while algorithms relying on a full-fledged near-field model with SNS and SWF necessitate long transmission durations to estimate a large number of parameters. To address these challenges, we propose a novel low-complexity localization algorithm based on a misspecified model that reduces the number of unknown channel parameters with limited impact on accuracy. A misspecified Cramér-Rao bound (MCRB) analysis is also performed to evaluate theoretical performance degradation due to the involved model misspecification. Additionally, we propose a codebook design that leverages coarse UE location information to enhance localization accuracy. Numerical simulations validate the effectiveness of the proposed estimation method and highlight the advantages of using the proposed codebook.