Hybrid ARQ in Multiple-Antenna Slow Fading Channels: Performance Limits and Optimal Linear Dispersion Code Design

Cong Shen, Michael P. Fitz

TL;DR

This paper investigates the fundamental performance limits and optimal design of Hybrid Automatic Repeat reQuest (HARQ) protocols in multiple-input multiple-output (MIMO) systems operating over slow fading channels. It introduces a general framework for optimizing average rates and proposes new design criteria for Linear Dispersion Codes (LDCs) based on both information-theoretic capacity and error probability analysis, offering a unified approach to HARQ design.

What Problem Does This Paper Address?

Traditional Hybrid ARQ (HARQ) protocols, designed for single-antenna systems, face challenges when extended to MIMO environments. While existing research has explored MIMO-HARQ, many problems remain unanswered. Specifically, the Diversity-Multiplexing Tradeoff (DMT) framework, often used for high Signal-to-Noise Ratio (SNR) analysis, doesn't fully address the practical need for fixed-rate codes operating well at finite SNRs. Furthermore, there's a lack of a unified framework to guide the design of optimal space-time codes (STCs) for MIMO-HARQ, leading to ad-hoc designs for retransmission strategies. This paper aims to bridge this gap by providing a comprehensive analysis of performance limits and developing optimal design criteria for LDCs in MIMO-HARQ systems.

Methods

The paper employs a two-pronged approach to analyze and design MIMO-HARQ protocols:

  1. Information-Theoretic Analysis:
    • System Model: A single-user MIMO system with $L_t$ transmit and $L_r$ receive antennas is considered, assuming an error-free and delay-free ARQ feedback link. A slow fading channel model is used, where the channel remains constant across all retransmissions.
  2. Performance Metric: The primary metric is the long-term average rate, which represents the average number of successfully accepted bits per unit time.
  3. General Rate Expression: A general expression for the average rate of HARQ protocols is derived, which serves as a foundation for optimization.
  4. Capacity-Based Design Criterion (Criterion 1): This criterion aims to make each LDC transmission "capacity lossless" by ensuring the equivalent MIMO channel capacity matches the capacity of the physical MIMO channel. This is achieved by optimizing the rate and time allocation for each ARQ round.
  5. Linear Dispersion Codes (LDCs): The paper utilizes the LDC framework, a general STC design, to model and analyze MIMO-HARQ protocols.
  6. Error Probability Analysis:
    • Pairwise Error Probability (PWEP): The analysis extends to studying the n-th pairwise error probability (n-PWEP) for decoding errors across multiple ARQ rounds.
  7. Diversity Order: The concept of diversity order is analyzed, and a criterion is derived to maximize it.
  8. Error-Probability-Based Design Criterion (Criterion 2): This criterion focuses on ensuring that the optimum detector achieves a specific diversity order after each ARQ round, aiming to improve error performance.

The paper then applies these frameworks to analyze traditional HARQ protocols like Incremental Redundancy (IR) and Chase Combining (CC), and evaluates various existing LDCs.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and numerical simulations based on synthetic channel models.

Key Insights and Ideas

  • Unified Framework for HARQ Design: The paper provides a general framework for optimizing HARQ protocols in MIMO systems, unifying information-theoretic and error-probability-based design approaches.
  • Capacity-Lossless LDCs: The concept of "capacity-lossless" LDCs (Criterion 1) is introduced, aiming to maximize the information rate at each transmission stage. This is a key insight for designing efficient HARQ protocols.
  • Diversity as a Key Design Goal: Criterion 2 highlights the importance of maximizing diversity order, especially in slow fading channels where time diversity is unavailable. This is crucial for robust communication.
  • Trade-offs in HARQ: The paper implicitly shows trade-offs between rate, complexity, and error performance. For instance, IR offers flexibility in rate assignment but at the cost of complex packetization, while CC is simpler but less efficient.
  • LDC Advantage over CC: LDC-based protocols can approach MIMO channel capacity more effectively than simple repetition (Chase Combining) schemes, especially in slow fading channels, by intelligently spreading symbols across antennas.
  • Importance of ARQ Feedback: The paper demonstrates that ARQ feedback can significantly improve average rates, particularly in uncoded scenarios, by allowing for retransmissions and adapting to channel conditions.

Key Results

  • Optimal Average Rate Expressions: General expressions for the average rate of HARQ protocols are derived, applicable to both IR and CC, and LDC-based schemes.
  • Design Criteria: Two key design criteria are proposed: a capacity-based criterion (Criterion 1) and an error-probability-based criterion (Criterion 2) focusing on diversity order.
  • Performance Evaluation of LDCs: Various existing LDCs are evaluated using these criteria. In MISO systems, the Alamouti code is shown to be optimal under certain conditions. For more complex MIMO systems, the paper presents numerical results showing that while some LDCs perform close to optimal, others are significantly suboptimal.
  • IR vs. LDC: The paper suggests that for scenarios requiring constant-length retransmissions, LDC-based protocols can achieve the same optimal average rates as IR but with potentially lower receiver complexity.
  • ARQ Feedback Gains: Numerical results (Fig. 9) show that HARQ protocols (Alamouti and CC) provide significant average rate gains over no-ARQ transmission, especially for uncoded transmissions. The packet error rate gain is also observed, though it diminishes with strong channel coding.

Limitations and Future Work

  • Focus on Slow Fading: The analysis is primarily for slow fading channels where the channel remains constant across retransmissions. The paper acknowledges that for time-varying channels, protocols exploiting both time and spatial diversity would be necessary.
  • Isotropic Gaussian Input Assumption: For analytical tractability, the paper often assumes isotropic Gaussian input distribution. While this simplifies analysis, it might not represent all practical scenarios.
  • Receiver Complexity: While the paper touches upon receiver complexity (e.g., Alamouti's advantage), a more explicit incorporation of receiver complexity into the design criteria is suggested as future work.
  • Search for Novel LDCs: For certain configurations (e.g., $L_t=4, L_r=2, N=4$), none of the existing LDCs approach optimal performance, indicating a need for searching or designing novel codes.
  • Fast-Decodable Property: The paper suggests that future LDC designs should not only satisfy the presented criteria but also possess a "fast-decodable" property for each ARQ round.

Summary updated

BibTeX
@article{shen2011hybrid,
  author = {Cong Shen and Michael P. Fitz},
  title = {{Hybrid ARQ in Multiple-Antenna Slow Fading Channels: Performance Limits and Optimal Linear Dispersion Code Design}},
  journal = {IEEE Trans. Information Theory},
  year = {2011},
  doi = {10.1109/tit.2011.2161949},
}
Figure