Abstract
Ultra-Reliable and Low Latency Communications (URLLC) is a cornerstone of 5G, enabling mission-critical applications like autonomous driving and industrial automation. Traditional coding schemes, optimized for long data blocks, struggle with the short-packet transmissions characteristic of URLLC. This paper introduces Sparse Vector Coding (SVC), a novel scheme that encodes information into the positions of non-zero elements within a sparse vector. By leveraging compressed sensing for recovery, SVC achieves significant reliability gains and lower latency compared to existing 4G and 5G standards for short-packet scenarios.
Introduction
The evolution of 5G extends beyond high-speed mobile broadband to include "mission-critical" services. These services, categorized under URLLC, demand extreme reliability (packet error rates as low as $10^{-5}$) and ultra-low latency (within 1ms). Standard wireless coding techniques, such as Turbo or Polar codes, are designed to maximize coding gain over large data blocks. However, URLLC packets are typically small—often just a few bytes of sensor data or control commands. In these "short-packet" regimes, the overhead of traditional codes becomes prohibitive, and their performance degrades. There is a clear need for a transmission strategy specifically tailored for small, high-stakes data bursts.
Proposed Methodology: Sparse Vector Coding (SVC)
The proposed Sparse Vector Coding (SVC) shifts the paradigm from traditional channel coding to the principles of Compressed Sensing (CS). The core idea is to represent information not just through signal values, but through the sparsity pattern of a transmitted vector.
1. Sparse Mapping
In the first stage, information bits are mapped to the indices of a few non-zero elements in a large vector $\mathbf{s}$. For a vector of size $N$ with $K$ active elements, there are $\binom{N}{K}$ possible combinations. This mapping allows for efficient representation of short bit sequences.
The algorithm above illustrates how an information vector $\mathbf{w}$ is converted into a sparse vector $\mathbf{s}$ by selecting specific bit positions.
2. Multi-code Spreading
The sparse vector $\mathbf{s}$ is then multiplied by a "sensing matrix" or codebook $\mathbf{C}$. This process spreads the sparse information across the available frequency resources. The choice of $\mathbf{C}$ is critical; the authors suggest using random Bernoulli matrices or deterministic Second-Order Reed-Muller (SORM) sequences to maintain low correlation between codewords, which is essential for accurate recovery.
3. Support Identification and Decoding
At the receiver, the task is "support identification"—determining which positions in the original vector were active. Rather than recovering exact values, the receiver uses greedy algorithms like Orthogonal Matching Pursuit (OMP) or more sophisticated Multipath Match Pursuit (MMP). MMP performs a tree search to find the most likely set of active indices, balancing complexity and accuracy.
Performance Evaluation
The performance of SVC was evaluated using standard industry channel models, including Additive White Gaussian Noise (AWGN) and realistic multipath fading models like Extended Typical Urban (ETU) and Extended Pedestrian-A (EPA). Benchmarks included 3GPP LTE-Advanced Rel.13 (4G PDCCH) and 5G Polar codes.
Key Findings
- Reliability Gains: In AWGN environments, SVC demonstrated a 4 dB gain over 4G LTE PDCCH and a 1.1 dB gain over 5G Polar codes at a target Block Error Rate (BLER) of $10^{-5}$.
- Power Efficiency: By concentrating energy into only a few active symbols, SVC achieves a higher effective Signal-to-Noise Ratio (SNR) per symbol than traditional systems that distribute power uniformly.
- Implicit Error Detection: The residual energy in the compressed sensing recovery process serves as a built-in reliability metric. This can potentially eliminate the need for extra Cyclic Redundancy Check (CRC) bits, further reducing overhead.
- Latency and Interference: SVC consistently met the 1ms latency requirement. Furthermore, it showed superior robustness in small-cell scenarios with high inter-cell interference compared to traditional 4G schemes.
Conclusion and Future Directions
Sparse Vector Coding offers a compelling solution for the unique demands of URLLC. By encoding information in the "where" rather than just the "what," it provides high reliability and low latency for short-packet transmissions. While currently optimized for very sparse vectors due to decoding complexity, future work could explore hybrid approaches combining SVC with traditional error correction or applying it to massive Machine Type Communications (mMTC) for IoT.
Summary updated