Introduction
As the telecommunications industry transitions toward 5G and beyond, Cloud Radio Access Networks (Cloud-RAN) have emerged as a pivotal architecture for managing network interference. By coordinating multiple base stations (BSs) through a central processor, Cloud-RAN enhances spectral efficiency. Simultaneously, there is a growing emphasis on "green" networking, utilizing renewable energy sources like solar and wind, alongside Simultaneous Wireless Information and Power Transfer (SWIPT). SWIPT allows the same radio frequency (RF) signal to deliver data to a receiver while providing energy to a harvester.
However, implementing this in a sustainable Cloud-RAN introduces a "double uncertainty": the intermittent nature of renewable energy arrivals at base stations and the inherent tradeoff between optimizing a signal for data throughput versus energy delivery. This paper addresses these challenges by developing a joint optimization framework for beamforming in a cluster of renewable-powered base stations.
System Model and Problem Formulation
The study considers a Cloud-RAN cluster where multiple base stations, each equipped with energy harvesting capabilities, serve a single data receiver (DR) and an energy receiver (ER). A central processor coordinates the base stations via backhaul links to optimize transmission strategies.
The objective is to maximize the total throughput of the DR over a finite time horizon while satisfying two critical constraints: ensuring the ER receives a minimum required amount of RF energy ($q$) and maintaining the energy neutrality of the base stations (i.e., they cannot consume more energy than they have harvested).
Optimization Methodology
The authors formulate this as an offline optimization problem. Due to the quadratic nature of beamforming, the initial problem is non-convex and computationally difficult. To solve it, the researchers employ several key technical strategies:
- Semidefinite Relaxation (SDR): The problem is relaxed into a Semidefinite Program (SDP), a class of convex optimization that is easier to solve.
- Rank-One Proof: A significant theoretical contribution of this work is the proof that the optimal solution to the relaxed SDP is always "rank-one." This confirms that the relaxation is tight, meaning the optimal beamforming vectors can be extracted directly without any loss of accuracy.
- Direct Water-Filling (DWT) Adaptation: The multi-slot problem is simplified by dividing the time horizon into intervals based on energy arrival events, allowing for more manageable power allocation.
- Online Heuristic: Recognizing that real-world systems lack perfect knowledge of future energy arrivals, the authors propose a practical online algorithm. This heuristic allocates power based on current residual energy levels and historical average harvesting rates.
Performance Evaluation
Numerical simulations were conducted to validate the proposed strategies, using a 3-BS cluster and Rayleigh fading channel models. The results highlight a fundamental Energy-Throughput (E-T) region, representing the Pareto optimal boundary where data rate must be sacrificed to increase energy delivery.
As shown in Figure 2, the proposed joint beamforming strategy significantly outperforms "separate beamforming," where energy and data are handled in distinct time slots. Furthermore, the practical online algorithm performs remarkably close to the theoretical offline optimal, particularly at higher energy harvesting rates. The simulations quantify the tradeoff: for example, increasing the required charging rate from 0.2 to 1.4 $\mu W$ typically results in a throughput drop of 1-2 Mbps.
Key Insights and Future Work
Several critical insights emerged from the study:
- Channel Correlation: When the channels to the DR and ER are highly correlated (e.g., the devices are physically close), the E-T region expands, as a single beam can efficiently serve both purposes.
- Proportional Powering: To maximize energy transfer, base stations should ideally consume power in proportion to their harvested energy ratios.
- Optimality of Single-Stream Beamforming: The rank-one proof confirms that a single data stream is sufficient for optimal performance in this SWIPT scenario.
Future research will look to extend this framework to more complex environments, including multi-user interference channels and massive MIMO systems, where increased antenna counts could offer even greater beamforming precision and efficiency.
Summary updated