RIS-Aided Wireless Fingerprinting Localization Based on Multilayer Graph Representations

Stefania Sardellitti, Paolo Di Lorenzo, Sergio BarbarossaIEEE Communications Letters 2024 · DOI

Abstract

This paper introduces a novel framework for high-precision indoor and outdoor localization using Reconfigurable Intelligent Surfaces (RIS). By integrating Graph Signal Processing (GSP) with multi-layer graph theory, the proposed method reconstructs high-fidelity radio maps from sparse physical measurements. The approach exploits spatial correlations and signal evolution across varying RIS configurations. Experimental results demonstrate that the multi-layer graph interpolation significantly outperforms traditional techniques, achieving localization accuracy near the theoretical limit of a full-density radio map even in the presence of signal blockages.

Introduction

Wireless Fingerprinting Localization (WFL) is a common technique that matches Received Signal Strength Indicators (RSSI) against a pre-recorded radio map. However, WFL faces significant challenges: the "curse of measurement," where creating dense maps is labor-intensive, and signal blockage in high-frequency bands like 6G/mmWave.

Reconfigurable Intelligent Surfaces (RIS) technology can mitigate blockages by creating alternative propagation paths, but it introduces a new dimension of complexity: multiple RSSI values per location based on RIS phase configurations. This work addresses the efficient interpolation of these multi-dimensional maps from minimal samples by leveraging the underlying structural correlations of the signal field.

Methods

The core innovation is the Multi-layer Graph Representation. The environment is modeled as a graph where nodes represent grid locations. Each layer corresponds to a specific RIS phase configuration.

  • Intra-layer edges connect nodes within the same layer based on spatial proximity and signal similarity.
  • Inter-layer edges connect the same physical location across different layers, capturing how the signal evolves as the RIS configuration changes.
Multi-layer graph representation of the environment across different RIS configurations.

RSSI values are treated as "signals" on this graph. Since these signals are spatially and structurally correlated, they are "bandlimited" in the Graph Fourier Transform (GFT) domain. The authors employ a convex optimization approach (Basis Pursuit) to find the sparsest representation of the radio map, enabling accurate interpolation of missing data points from a tiny subset of physical samples. Once the map is reconstructed, a $k$-Nearest Neighbor ($k$-NN) approach is used to estimate the user's position.

Key Results

The method was evaluated in a simulated 2D scenario (20m x 10m) featuring a central obstacle providing 50dB of attenuation. The area was discretized into 150 points with 100 distinct RIS configurations.

The multi-layer approach demonstrated superior reconstruction performance, particularly with low sample counts. In terms of localization, the proposed method achieved an average error of approximately 2.8 to 3 meters in noisy environments. As shown in the performance analysis, this accuracy is remarkably close to the "True Map" benchmark—the theoretical limit achievable with exhaustive measurements. Furthermore, the RIS-aided approach significantly outperformed MISO systems without RIS, as the RIS effectively "illuminated" shadowed regions behind the central obstacle that traditional beamforming could not reach.

Average localization error as a function of the number of RIS configurations compared to various benchmarks.

Conclusion

By jointly exploiting spatial and configuration-domain correlations, the multi-layer graph framework provides a robust solution for RIS-aided localization. While computational complexity for large-scale environments remains a consideration, the ability to achieve high accuracy with sparse sampling offers a significant path forward for 6G localization systems. Future work will focus on extending the model to MIMO systems and developing techniques to learn graph topologies directly from data.

Summary updated

BibTeX
@article{sardellitti2024ris,
  author = {Stefania Sardellitti and Paolo Di Lorenzo and Sergio Barbarossa},
  title = {{RIS-Aided Wireless Fingerprinting Localization Based on Multilayer Graph Representations}},
  journal = {IEEE Communications Letters},
  year = {2024},
  doi = {10.1109/lcomm.2024.3380741},
}
Figure