Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Network-Assisted Full-Duplex Cell-Free mmWave Networks: Hybrid MIMO Processing and Multi-Agent DRL-Based Power Allocation

Qingrui Fan, Yu Zhang, Jiamin Li, Dongming Wang, Hongbiao Zhang, Xiaohu You

TL;DR

This work addresses cross-link interference in network-assisted full-duplex cell-free mmWave networks by developing a hybrid analog-digital MIMO processing framework and a collaborative MATD3-based bidirectional power allocation. It derives explicit uplink/downlink lower-bound spectral efficiencies under realistic channel estimation and interference models, and formulates a non-convex joint power optimization problem solved via MATD3, a TD3-based multi-agent reinforcement learning approach. The proposed method uses MMSE-based inter-AP and user-AP channel estimation, RF beamforming tied to spatial covariances, and centralized training with distributed execution to mitigate overestimation and reduce computational burden. Simulation results demonstrate accurate channel estimation, MATD3’s superior performance over MADDPG and conventional schemes, and notable gains in bidirectional spectral efficiency, suggesting practical scalability for dense mmWave cell-free deployments.

Abstract

This paper investigates the network-assisted full-duplex (NAFD) cell-free millimeter-wave (mmWave) networks, where the distribution of the transmitting access points (T-APs) and receiving access points (R-APs) across distinct geographical locations mitigates cross-link interference, facilitating the attainment of a truly flexible duplex mode. To curtail deployment expenses and power consumption for mmWave band operations, each AP incorporates a hybrid digital-analog structure encompassing precoder/combiner functions. However, this incorporation introduces processing intricacies within channel estimation and precoding/combining design. In this paper, we first present a hybrid multiple-input multiple-output (MIMO) processing framework and derive explicit expressions for both uplink and downlink achievable rates. Then we formulate a power allocation problem to maximize the weighted bidirectional sum rates. To tackle this non-convex problem, we develop a collaborative multi-agent deep reinforcement learning (MADRL) algorithm called multi-agent twin delayed deep deterministic policy gradient (MATD3) for NAFD cell-free mmWave networks. Specifically, given the tightly coupled nature of both uplink and downlink power coefficients in NAFD cell-free mmWave networks, the MATD3 algorithm resolves such coupled conflicts through an interactive learning process between agents and the environment. Finally, the simulation results validate the effectiveness of the proposed channel estimation methods within our hybrid MIMO processing paradigm, and demonstrate that our MATD3 algorithm outperforms both multi-agent deep deterministic policy gradient (MADDPG) and conventional power allocation strategies.

Network-Assisted Full-Duplex Cell-Free mmWave Networks: Hybrid MIMO Processing and Multi-Agent DRL-Based Power Allocation

TL;DR

This work addresses cross-link interference in network-assisted full-duplex cell-free mmWave networks by developing a hybrid analog-digital MIMO processing framework and a collaborative MATD3-based bidirectional power allocation. It derives explicit uplink/downlink lower-bound spectral efficiencies under realistic channel estimation and interference models, and formulates a non-convex joint power optimization problem solved via MATD3, a TD3-based multi-agent reinforcement learning approach. The proposed method uses MMSE-based inter-AP and user-AP channel estimation, RF beamforming tied to spatial covariances, and centralized training with distributed execution to mitigate overestimation and reduce computational burden. Simulation results demonstrate accurate channel estimation, MATD3’s superior performance over MADDPG and conventional schemes, and notable gains in bidirectional spectral efficiency, suggesting practical scalability for dense mmWave cell-free deployments.

Abstract

This paper investigates the network-assisted full-duplex (NAFD) cell-free millimeter-wave (mmWave) networks, where the distribution of the transmitting access points (T-APs) and receiving access points (R-APs) across distinct geographical locations mitigates cross-link interference, facilitating the attainment of a truly flexible duplex mode. To curtail deployment expenses and power consumption for mmWave band operations, each AP incorporates a hybrid digital-analog structure encompassing precoder/combiner functions. However, this incorporation introduces processing intricacies within channel estimation and precoding/combining design. In this paper, we first present a hybrid multiple-input multiple-output (MIMO) processing framework and derive explicit expressions for both uplink and downlink achievable rates. Then we formulate a power allocation problem to maximize the weighted bidirectional sum rates. To tackle this non-convex problem, we develop a collaborative multi-agent deep reinforcement learning (MADRL) algorithm called multi-agent twin delayed deep deterministic policy gradient (MATD3) for NAFD cell-free mmWave networks. Specifically, given the tightly coupled nature of both uplink and downlink power coefficients in NAFD cell-free mmWave networks, the MATD3 algorithm resolves such coupled conflicts through an interactive learning process between agents and the environment. Finally, the simulation results validate the effectiveness of the proposed channel estimation methods within our hybrid MIMO processing paradigm, and demonstrate that our MATD3 algorithm outperforms both multi-agent deep deterministic policy gradient (MADDPG) and conventional power allocation strategies.
Paper Structure (23 sections, 3 theorems, 54 equations, 9 figures, 1 table, 1 algorithm)

This paper contains 23 sections, 3 theorems, 54 equations, 9 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. Hybrid MIMO Processing for NAFD Cell-Free mmWave Networks
  4. Inter-AP Interference Channel Estimation
  5. User-AP Channel Estimation
  6. Downlink Channel Estimation
  7. Uplink Channel Estimation
  8. Data Transmission
  9. Downlink Data Transmission
  10. Uplink Data Transmission
  11. Spectral Efficiency Analysis
  12. Downlink Spectral Efficiency
  13. Uplink Spectral Efficiency
  14. MADRL-based Bidirectional Power Allocation
  15. Problem Formulation
  16. ...and 8 more sections

Key Result

Theorem 1

The MMSE channel estimate of $\mathrm{vec}(\mathbf{H}_{m, z}^\mathrm{AP})$ is given by where the optimal coupling matrix $\mathbf{A}$ is designed as where $\mathbf{\Sigma}_A$ is obtained by solving a typical water-filling problem and $\mathbf{U}_R$ is calculated from the eigenvalue decomposition $\mathbf{R}_{m,z}^\mathrm{AP} = \mathbf{U}_R \mathbf{\Lambda}_R \mathbf{U}_R^\mathrm{H}$.

Figures (9)

  • Figure 1: The framework of NAFD cell-free mmWave networks.
  • Figure 2: The MATD3 framework for the power allocation of NAFD cell-free mmWave networks.
  • Figure 3: NMSE for different RF chains and full digital case with $\mathrm{N_{AP}}=32$.
  • Figure 4: NMSE for different RF chains and full digital case with $\mathrm{N_{AP}}=128$.
  • Figure 5: The average reward with different $\gamma$ based on the MATD3.
  • ...and 4 more figures

Theorems & Definitions (5)

  • Theorem 1
  • Remark 1
  • Remark 2
  • Theorem 2
  • Theorem 3