1. Introduction
As the telecommunications industry pivots toward 6G, the "Internet of Everything" (IoE) presents a significant challenge: managing the simultaneous connection of millions of devices. To address this, grant-free access has emerged as a critical paradigm, allowing devices to transmit data without the overhead of a formal handshake. In these systems, only a small fraction of devices are active at any given time, creating a "sparse" signal environment. Traditionally, Compressed Sensing (CS) algorithms such as Approximate Message Passing (AMP) have been employed to identify active users and estimate channel states. However, these classical approaches often struggle when device activities are correlated—a common occurrence in localized sensor networks.
2. The Correlation Problem in Compressed Sensing
Standard CS algorithms typically rely on element-wise denoisers that assume device activity is independent and identically distributed (i.i.d.). In practical 6G scenarios, devices in proximity often exhibit correlated behavior, such as a cluster of sensors triggered by the same environmental event. When these correlations exist, the i.i.d. assumption fails, leading to suboptimal performance in user detection and channel estimation. There is a pressing need for denoisers capable of learning and exploiting these complex joint probability distributions.
3. The VQC-CS Framework
This paper introduces VQC-CS, a hybrid classical-quantum framework designed to bridge the gap between quantum computing and wireless communications. The architecture integrates a Variational Quantum Circuit (VQC) into the iterative loop of the Orthogonal Approximate Message Passing (OAMP) algorithm.
3.1 Hybrid Iterative Loop
The VQC-CS algorithm alternates between two primary stages:
- Linear Estimation (LE): A classical step that provides an initial estimate of the signal based on the received symbols and the pilot matrix.
- Quantum Non-Linear Estimation (NLE): A quantum step where the VQC acts as a learnable denoiser.
3.2 VQC as a Learnable Denoiser
Unlike fixed mathematical denoisers, the VQC is trained to recognize patterns in the signal. The LE estimate and the current residual error are embedded into the rotation angles of quantum gates ($R_X, R_Y, R_Z$). The circuit utilizes a multi-layer "data re-uploading" ansatz, which allows the quantum processor to function as a universal function approximator. This structure is particularly effective at capturing the dependencies between multiple qubits, representing the correlated states of different devices.
3.3 Optimization
The parameters of the VQC are optimized using classical back-propagation via the parameter-shift rule. The objective is to minimize the Mean Square Error (MSE) of the signal estimation over multiple iterations. To further enhance performance, a small classical Multi-Layer Perceptron (MLP) is employed as a post-processing step to refine the final binary decision regarding device activity.
4. Methodology and Experimental Setup
The proposed framework was evaluated using synthetic data simulating a realistic wireless environment:
- Channel Model: Rayleigh fading coefficients.
- Activity Pattern: A sparse vector with an activity ratio of $\rho = 0.2$.
- Correlation Model: An autoregressive model where the correlation between devices $i$ and $j$ is defined by $\gamma^{|i-j|}$, with $\gamma = 0.6$.
- System Dimensions: A system with $N=10$ devices and $M=6$ or $7$ received symbols, representing a highly underdetermined system.
5. Results and Discussion
The VQC-CS framework was compared against established classical algorithms, including Iterative Soft Thresholding Algorithm (ISTA), Fast ISTA (FISTA), and standard OAMP.
5.1 Channel Estimation Performance
VQC-CS demonstrated superior convergence and accuracy. It achieved the lowest MSE after only three iterations, whereas classical OAMP saturated at a significantly higher error floor. This suggests that the quantum denoiser effectively utilizes the correlation information that classical denoisers ignore.
5.2 User Identification Accuracy
The performance was also measured using the Area Under the Curve (AUC) for the Receiver Operating Characteristic (ROC):
- VQC-CS (Direct): 0.976 AUC
- VQC-CS (with DNN Post-processing): 0.986 AUC
- ISTA/FISTA: 0.952 AUC
- OAMP: 0.988 AUC
While OAMP remains highly competitive in specific metrics, VQC-CS showed remarkable robustness in "severe" channel conditions where the number of received symbols was minimal.
6. Conclusion and Future Work
The VQC-CS framework demonstrates the potential of hybrid quantum-classical algorithms to solve complex problems in 6G communications. By using VQCs to learn joint probability distributions, the system achieves high-fidelity channel estimation and user identification in the presence of correlated activity.
However, several challenges remain. The current study focused on a small-scale system ($N=10$). Scaling to hundreds of devices will require more qubits or innovative "tiling" strategies. Additionally, while the simulations are promising, future research must validate these results on physical Noisy Intermediate-Scale Quantum (NISQ) hardware. The integration of the post-processing DNN also highlights that maintaining strict orthogonality in learned denoisers is a complex task that warrants further investigation.
Summary updated