Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CSI Prediction Frameworks for Enhanced 5G Link Adaptation: Performance-Complexity Trade-offs

Francisco Díaz-Ruiz, Francisco J. Martín-Vega, Jose A. Cortés, Gerardo Gómez, Mari Carmen Aguayo

TL;DR

The paper addresses the challenge of timely and accurate CSI for 5G link adaptation in the presence of channel aging and feedback delays by forecasting future effective SINR values in the SINR domain, enabling low-complexity, CQI-compatible predictions. It compares a classical Wiener filter against AI-based predictors (GRU, LSTM, DNN) across TDD and FDD modes, and analyzes input window length, Doppler effects, and target strategies (Best-CQI vs By-CQI). The results show GRU-based predictors offer superior generalization to diverse channels, while the Wiener filter provides competitive accuracy with far lower complexity; optimal configurations include a small input window and, depending on deployment, different predictor choices for BS versus UE. Practically, the study provides deployment guidance for 5G LA: use learning-based predictors at TDD BS to maximize throughput under mobility, and prefer the Wiener filter at FDD UE for low-power, low-complexity operation, while maintaining compatibility with standard EESM-based CQI selection.

Abstract

Accurate and timely channel state information (CSI) is fundamental for efficient link adaptation. However, challenges such as channel aging, user mobility, and feedback delays significantly impact the performance of adaptive modulation and coding (AMC). This paper proposes and evaluates two CSI prediction frameworks applicable to both time division duplexing (TDD) and frequency division duplexing (FDD) systems. The proposed methods operate in the effective signal to interference plus noise ratio (SINR) domain to reduce complexity while preserving predictive accuracy. A comparative analysis is conducted between a classical Wiener filter and state-of-the-art deep learning frameworks based on gated recurrent units (GRUs), long short-term memory (LSTM) networks, and a delayed deep neural network (DNN). The evaluation considers the accuracy of the prediction in terms of mean squared error (MSE), the performance of the system, and the complexity of the implementation regarding floating point operations (FLOPs). Furthermore, we investigate the generalizability of both approaches under various propagation conditions. The simulation results show that the Wiener filter performs close to GRU in terms of MSE and throughput with lower computational complexity, provided that the second-order statistics of the channel are available. However, the GRU model exhibits enhanced generalization across different channel scenarios. These findings suggest that while learningbased solutions are well-suited for TDD systems where the base station (BS) handles the computation, the lower complexity of classical methods makes them a preferable choice for FDD setups, where prediction occurs at the power-constrained user equipment (UE).

CSI Prediction Frameworks for Enhanced 5G Link Adaptation: Performance-Complexity Trade-offs

TL;DR

The paper addresses the challenge of timely and accurate CSI for 5G link adaptation in the presence of channel aging and feedback delays by forecasting future effective SINR values in the SINR domain, enabling low-complexity, CQI-compatible predictions. It compares a classical Wiener filter against AI-based predictors (GRU, LSTM, DNN) across TDD and FDD modes, and analyzes input window length, Doppler effects, and target strategies (Best-CQI vs By-CQI). The results show GRU-based predictors offer superior generalization to diverse channels, while the Wiener filter provides competitive accuracy with far lower complexity; optimal configurations include a small input window and, depending on deployment, different predictor choices for BS versus UE. Practically, the study provides deployment guidance for 5G LA: use learning-based predictors at TDD BS to maximize throughput under mobility, and prefer the Wiener filter at FDD UE for low-power, low-complexity operation, while maintaining compatibility with standard EESM-based CQI selection.

Abstract

Accurate and timely channel state information (CSI) is fundamental for efficient link adaptation. However, challenges such as channel aging, user mobility, and feedback delays significantly impact the performance of adaptive modulation and coding (AMC). This paper proposes and evaluates two CSI prediction frameworks applicable to both time division duplexing (TDD) and frequency division duplexing (FDD) systems. The proposed methods operate in the effective signal to interference plus noise ratio (SINR) domain to reduce complexity while preserving predictive accuracy. A comparative analysis is conducted between a classical Wiener filter and state-of-the-art deep learning frameworks based on gated recurrent units (GRUs), long short-term memory (LSTM) networks, and a delayed deep neural network (DNN). The evaluation considers the accuracy of the prediction in terms of mean squared error (MSE), the performance of the system, and the complexity of the implementation regarding floating point operations (FLOPs). Furthermore, we investigate the generalizability of both approaches under various propagation conditions. The simulation results show that the Wiener filter performs close to GRU in terms of MSE and throughput with lower computational complexity, provided that the second-order statistics of the channel are available. However, the GRU model exhibits enhanced generalization across different channel scenarios. These findings suggest that while learningbased solutions are well-suited for TDD systems where the base station (BS) handles the computation, the lower complexity of classical methods makes them a preferable choice for FDD setups, where prediction occurs at the power-constrained user equipment (UE).

Paper Structure

This paper contains 23 sections, 22 equations, 17 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Motivation and Scope
  3. Related Work
  4. Main Contributions
  5. System Model
  6. CSI acquisition framework in 5G
  7. Metrics
  8. Proposed CSI Prediction Frameworks
  9. Wiener Filter Prediction
  10. AI-Based Prediction Networks
  11. DNN-Based Prediction Network
  12. LSTM-Based Prediction Network
  13. GRU-Based Prediction Network
  14. Numerical Results
  15. Model design
  16. ...and 8 more sections

Figures (17)

  • Figure 1: CSI acquisition process in a TDD system.
  • Figure 2: CSI acquisition process in an FDD system.
  • Figure 3: Timeline of the CSI acquisition and downlink data transmission process, illustrating the $k$-th transmission interval.
  • Figure 4: Proposed CSI prediction framework.
  • Figure 5: Unified architecture of the proposed AI-based prediction networks.
  • ...and 12 more figures

Theorems & Definitions (4)

  • Remark 1: Optimal Model Complexity
  • Remark 2: Dependence on Doppler and CSI reporting period
  • Remark 3: Input sampling
  • Remark 4