Back to search

Virginia Commonwealth University

ENHANCING IRS LOCALIZATION VIA ML-DRIVEN AOA AND DISTANCE ESTIMATION

Abstract

dc:description.abstract

Intelligent Reflecting Surfaces (IRSs) are transforming next-generation wireless networks by enabling dynamic control of the propagation environment for improved localization and security. This thesis advances Angle-of-Arrival (AoA) and distance estimation in IRS-assisted systems using a hybrid architecture with active sensing and passive reflecting elements. A modified MUSIC algorithm, optimized for near-field localization, decouples 2‑D AoA and 1‑D range estimation using a grid-and-peak pseudo-spectrum scan, achieving sub-degree AoA (0.003 rad) and sub-meter range (0.05 m) accuracy with over 100× lower complexity than classical 3D MUSIC. The study further applies the algorithm to the DeepMIMO dataset (O1_28 scenario, 3.5 GHz) using hybrid IRSs with varying active elements (M = 10×10, 15×15, 18×18), achieving improved accuracy (0.06 rad AoA, ≃0.03 m range) and 90% signal classification. Phase shifts of the 25×25 passive arrays are optimized to maximize SINR (20 dB), enhancing secure communication. To refine estimates, two Convolutional Neural Networks (CNNs) are developed: RefineNet, which improves MUSIC-based AoA and range estimation to achieve centimeter-level accuracy, and ElementNet, which optimizes the number and placement of active IRS elements (M = 4), reducing hardware costs while maintaining high accuracy. These methods enable 100% classification of legitimate versus malicious signals using the Hungarian algorithm. In summary, this work validates the modified MUSIC algorithm, extends its application to realistic channels, and integrates deep learning to achieve high-precision localization, secure communication, and cost-efficient IRS configurations.

Degree

thesis:*
Name thesis:degree_name
Master of Science
Level thesis:degree_level
Thesis
Discipline thesis:degree_discipline
Electrical & Computer Engineering

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Kotlapara Surendran, Nikhila
Contributors dc:contributor
  • Dr. Yanxiao Zhao
  • Dr. Ruixin Niu
  • Dr. Changqing Luo

Subjects

dc:subject × 7

Rights

dc:rights
Statement dc:rights
  • © The Author

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Virginia Commonwealth University
Base URL
scholarscompass.vcu.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Kotlapara Surendran, Nikhila. ENHANCING IRS LOCALIZATION VIA ML-DRIVEN AOA AND DISTANCE ESTIMATION. Thesis thesis, https://doi.org/10.25772/CV5W-7C45