Quantum Noise-Aware RIS-Aided Wireless Networks Using Variational Encoding and Signal Stabilization

TL;DR

The paper tackles blockage prediction in RIS-enabled wireless networks under realistic quantum hardware noise. It introduces a quantum-enabled architecture with a Quantum Base Station (QBS), a QRIS, and a mobile Quantum User Node (QUN), using a hybrid input $|\Psi_u\rangle = |\tilde{\psi}_X\rangle \otimes |\phi_u\rangle$ that combines amplitude-encoded visual data $|\psi_X\rangle$ and rotation-encoded channel data, processed by a variational quantum circuit to output a ternary label $l_u \in \{-1,0,1\}$. The method explicitly models depolarizing and dephasing noise on direct and QRIS paths ($\mathcal{E}_{dep}$, $\mathcal{E}_{phase}$) and employs amplitude damping during encoding with a fidelity-aware loss $\mathcal{L}_{total} = \mathcal{L}_{CE} + \lambda (1 - F(\rho_{ideal}, \rho_{noisy}))$ to stabilize training, optimized via the parameter-shift rule. Empirical results on a quantum-adapted ViWi dataset show that the hybrid input with fidelity-aware training achieves higher accuracy and quantum fidelity under noise than classical baselines and single-modality variants, highlighting the practical viability of quantum-enhanced wireless inference.

Abstract

This paper presents a noise-aware quantum-assisted framework for blockage prediction in reconfigurable intelligent surface (RIS)-enabled wireless networks. The proposed architecture integrates a Quantum Base Station (QBS), a Quantum RIS (QRIS), and a mobile Quantum User Node (QUN). Visual information captured by an onboard RGB camera is amplitude-encoded into quantum states, while channel state observations are mapped into quantum rotation-encoded features. These hybrid inputs are processed through variational quantum circuits, enabling ternary classification of the link status. To address the inherent imperfections of noisy intermediate-scale quantum (NISQ) hardware, the system explicitly models depolarizing and dephasing channels along direct and QRIS-assisted paths. A fidelity-aware training objective is employed to jointly minimize classification loss and quantum state degradation, with amplitude damping and synthetic noise injection enhancing robustness. Simulation results on a quantum-adapted version of the ViWi dataset demonstrate that the proposed hybrid quantum model achieves superior accuracy and stability under realistic noise conditions, outperforming baseline and single-modality approaches.

Figures (2)

• Figure 1: System model of the proposed quantum RIS-assisted blockage prediction framework. The QBS encodes visual input $|\psi_X\rangle$ and channel parameter $\theta_u$ into a hybrid quantum state. The signal propagates via a direct channel $\mathcal{E}_{\text{BQ}}$ or is reflected through the QRIS, resulting in an effective received state $\rho_{\text{eff}}^u$. The VQC receives both quantum input and noisy channel state, and performs ternary blockage prediction $l_u \in \{-1, 0, 1\}$.
• Figure 2: Illustrative performance analysis of the proposed noise-aware quantum RIS-assisted blockage prediction framework.