Introduction
Relay networks are essential for enhancing throughput and reliability in wireless communications. While traditional schemes like Amplify-and-Forward (AF) and Decode-and-Forward (DF) are common, they suffer from noise amplification and decoding failures at the relay, respectively. Quantize-Map-and-Forward (QMF) relaying offers a robust alternative that approaches theoretical capacity limits, particularly at high signal-to-noise ratios (SNR). However, the practical implementation of QMF has been hindered by the complexity of joint decoding at the destination. This paper presents a practical, low-complexity coding framework that bridges this gap using a unified factor graph approach.
System Architecture and Coding Design
The proposed system employs a specific LDPC-LDGM architecture to achieve efficient QMF relaying. The source node encodes information using a standard binary Low-Density Parity-Check (LDPC) code. At the relay, instead of attempting to decode the source message, the received signal is scalar-quantized. These quantized bits are then mapped using a Low-Density Generator Matrix (LDGM) code.
The core innovation lies in the representation of the entire network as a single joint factor graph. This graph links the source's LDPC subgraph with the relay's LDGM subgraph through "Quantize" (Q) nodes, which model the probabilistic relationship between source bits and relay observations.
To further simplify the destination's decoding task, the authors utilize a Diagonal Bell Labs Space-Time (DBLAST) architecture. This approach effectively orthogonalizes the signals from the source and relay, eliminating complex multiple-access interference and allowing for a decoupled factor graph structure.
Key Insights: LDPC-LDGM Synergy
The choice of LDPC for the source and LDGM for the relay is strategic. While LDPC is optimal for channel coding, LDGM is exceptionally well-suited for lossy compression tasks at the relay. By integrating them, the destination can perform joint decoding using standard Belief Propagation (BP). A significant finding is that for 1-bit scalar quantization, the complex Q nodes can be decomposed into simple parity-check (CHK) nodes and dummy variable nodes, making the system compatible with existing LDPC hardware.
Performance Evaluation
The framework was evaluated using synthetic simulations on Binary Input Additive White Gaussian Noise (BIAWGN) and MIMO channels (specifically a $1 \times 1 \times 2$ relay network). With block lengths ranging from $10^4$ to $10^5$ bits, the results demonstrate that the proposed scheme performs within 0.5–1 dB of the information-theoretic QMF threshold. In high-SNR scenarios with high-order modulation (e.g., 64-QAM), the QMF design outperformed DF by approximately 1.5 dB. Furthermore, all operations scale linearly with block length, ensuring feasibility for real-time applications.
Conclusion and Future Directions
This work provides a practical path for implementing QMF relaying in real-world wireless systems. While the current focus is on single-relay scenarios with perfect synchronization, future research will explore extensions to multi-relay "diamond" networks and the impact of channel estimation errors. Additionally, integrating Hybrid ARQ could enhance performance in fading environments.
Summary updated