Decision Feedback Differential Detection for Reconfigurable Intelligent Surfaces

Technical Summary: Decision Feedback Differential Detection for Reconfigurable Intelligent Surfaces

Problem Statement
This work addresses the performance limitations of Differential Reflecting Modulation (DRM) schemes in Reconfigurable Intelligent Surface (RIS) systems operating over time-varying fading channels. While DRM offers the advantage of noncoherent detection—eliminating the need for Channel State Information (CSI) acquisition, which is complex in RIS systems due to the lack of baseband processing at the RIS units—conventional differential detection (CDD) receivers suffer from significant performance degradation in dynamic environments. Specifically, CDD-based systems experience high error floors as the channel varies rapidly, limiting their reliability in time-varying Rayleigh fading scenarios.

Methodology
The authors propose a Decision Feedback Differential Detection (DFDD) technique tailored for DRM RIS systems. The core methodology involves:

• System Model: The system utilizes an RIS with $K$ reflecting units and $M$-ary PSK modulation. Information is encoded via permutation matrices and phase shifts, with differential encoding performed across time blocks. The channel is modeled as a time-varying Rayleigh fading process based on Jakes' model, assuming spatially uncorrelated fading.
• DFDD Mechanism: Unlike CDD, which uses only the immediately preceding symbol as a reference, the DFDD receiver utilizes a linear prediction filter. It combines multiple previously decoded symbols ($Y[t-v]$) weighted by predictor coefficients ($p_v$) to construct a reference signal ($\hat{Y}[t]$). This reference assists in detecting the current symbol, effectively smoothing out channel variations.
• Parameter Optimization: The study investigates the impact of the prediction order ($V$), the normalized Doppler frequency ($f_D T_s$), and the number of reflecting units ($K$) on system performance. Predictor coefficients are derived by minimizing the mean-square error (MSE) between the predicted channel state and the actual state, assuming previous decisions are correct.
• Comparative Analysis: The proposed DFDD scheme is evaluated against uncoded DRM and Differential Space-Time Modulation (DSTM) coded DRM, both employing CDD receivers. Performance is assessed via extensive Monte-Carlo simulations using BPSK and QPSK modulations.

Key Contributions and Results

• Error Floor Reduction: The primary finding is that DFDD significantly lowers the error floor compared to CDD-based schemes in time-varying channels. While CDD systems encounter high error floors (e.g., $10^{-4}$ to $10^{-3}$) as SNR increases, the DFDD scheme continues to improve, reaching error floors as low as $10^{-8}$.
• Optimal Prediction Order: Simulation results indicate that a prediction order of $V=2$ consistently provides the best performance across various scenarios (different $K$, $f_D T_s$, and modulation types). Higher orders (e.g., $V=3$) often lead to performance degradation at high SNR due to error propagation from incorrect past decisions.
• SNR-Dependent Performance: At low SNR, the DFDD scheme performs comparably to uncoded DRM with CDD, as decision feedback errors are more prevalent. However, as SNR increases, DFDD outperforms both uncoded DRM and DSTM-coded DRM. For instance, at high SNR, the DFDD scheme achieves a significantly lower error floor than DSTM-coded schemes, which eventually plateau at higher error rates.
• Robustness to Doppler: The system demonstrates sensitivity to the normalized Doppler frequency; as $f_D T_s$ increases, the error floor rises for all configurations. However, the relative advantage of DFDD over CDD remains consistent.
• Complexity Analysis: The paper analyzes computational complexity, noting that while DFDD introduces a linear increase in complexity proportional to the prediction order $V$ (specifically $O(2^r V K^3)$ for demodulation), this is a modest trade-off for the substantial gains in error rate performance. The DSTM-coded scheme is noted to have higher complexity due to the size of the candidate set ($2^{r'}$).

Significance and Claims
The paper claims that the chief merit of employing DFDD receivers in RIS-DRM systems is the provision of significantly lower error floors over time-varying fading channels, a critical improvement for reliable communication without CSI. The authors assert that while DSTM coding also reduces error floors compared to uncoded CDD, the uncoded DRM with DFDD achieves the lowest error floor among the three schemes evaluated. The work highlights that proper parameter selection (specifically $V=2$) is essential to maximize performance. The authors conclude that DFDD is a viable and effective technique for RIS systems, offering a balance between performance enhancement and manageable complexity, and suggest future research directions involving DFDD for coded DRM-DSTM systems and alternative prediction coefficient designs.