Abstract
This paper addresses the optimization of simultaneous wireless information and power transfer (SWIPT) within full-duplex (FD) multi-user MIMO networks. By developing novel precoding algorithms based on a "path-following" mathematical framework, the research solves complex, non-convex optimization problems inherent in these systems. The study demonstrates that full-duplex architectures using time-splitting (TS) significantly outperform traditional half-duplex (HD) and power-splitting (PS) alternatives in both data throughput and energy harvesting (EH) efficiency.
Introduction
As wireless communication evolves toward 5G and beyond, the integration of Full-Duplex (FD) communication and Wireless Energy Harvesting (EH) has become a critical area of research. FD technology allows for simultaneous transmission and reception on the same frequency, potentially doubling spectral efficiency. However, it introduces significant interference challenges, including Self-Interference (SI) at the base station and inter-user interference. Balancing these interference components while optimizing for both information decoding and energy scavenging requires sophisticated precoding strategies.
System Model and Problem Formulation
The research considers a multi-cell network environment where base stations operate in full-duplex mode to serve both Downlink Users (DLUs) and Uplink Users (ULUs). The primary challenge lies in managing the massive amounts of interference generated in such a dense environment.
As shown in the figure above, the network topology involves complex interactions:
- Self-Interference (SI): The leakage from a base station's transmitter to its own receiver.
- Inter-Tier Interference (ITI): Interference between different cells and between uplink and downlink transmissions.
The authors specifically focus on the mathematical difficulty of Time Splitting (TS). While TS is hardware-efficient compared to Power Splitting (PS), it is harder to optimize because the time-allocation variable is coupled with throughput and interference equations across the network.
Proposed Methodology: Path-Following Algorithms
To solve the non-convex optimization problems, the authors propose a framework based on Successive Convex Quadratic Programming, referred to as path-following algorithms.
- Minorant Functions: The non-convex objective functions (such as sum throughput) are approximated using "minorant" functions—concave quadratic approximations that provide a lower bound for the original function.
- Iterative Optimization: The system solves a sequence of these simpler convex quadratic programs. Each iteration refines the solution, ensuring convergence to a local optimum.
- Variable Transformation: To manage the non-convexity introduced by the Time Splitting variable ($\alpha$), the authors employ a change of variables ($\rho = 1/\alpha$). This allows non-convex constraints to be approximated by affine functions.
Performance Evaluation
The algorithms were evaluated using extensive Monte Carlo simulations employing Rician fading models and far-field uniform linear antenna arrays.
Key Insights
- Superiority of Time Splitting: TS often outperforms PS because it allows the base station to utilize distinct precoding matrices for the energy and data phases, better exploiting spatial diversity.
- The FD Bottleneck: FD gain is limited by intracell interference (e.g., ULU signals interfering with DLU reception), even with perfect SI cancellation.
- Antenna Scaling: The benefits of FD are most significant when User Equipment (UE) possesses multiple antennas, providing the spatial degrees of freedom necessary to mitigate interference.
Key Results
- Throughput Gains: In single-cell scenarios with dual-antenna UEs, FD achieved a ~26% throughput increase over HD for both PS and TS mechanisms.
- Energy Harvesting: In multi-cell tests, FD networks harvested significantly more energy (e.g., 10.09 dBm) compared to HD networks (8.5 dBm) by scavenging from both base stations and nearby ULUs simultaneously.
- Convergence: The proposed algorithms are computationally efficient, typically reaching 90% of the optimal value within 10 to 40 iterations.
Conclusion and Future Work
This research provides a robust mathematical framework for optimizing FD-MIMO networks with energy harvesting. While the results show clear advantages for FD-TS systems, the current model assumes perfect Channel State Information (CSI). Future work will focus on developing robust precoder designs that account for channel estimation errors in real-world deployments.
Summary updated