TL;DR
This paper introduces an optimized "$\alpha$-duplex" scheme that allows for partial spectrum overlap between uplink and downlink transmissions in cellular networks. By dynamically tuning the overlap fraction and transmission powers, the proposed method significantly outperforms traditional half-duplex and full-duplex systems, specifically protecting the vulnerable uplink from being overwhelmed by high-power downlink interference.
What Problem Does This Paper Address?
In the quest for higher spectral efficiency, In-Band Full Duplex (FD) communication is a holy grail, theoretically doubling capacity by allowing simultaneous transmission and reception on the same frequency. However, in a cellular environment, this creates a "loud neighbor" problem.
Base stations (BS) transmit at much higher power levels than mobile users (UE). In a full-duplex setup, the downlink signal from a base station can cause massive "cross-mode" interference for the uplink signals being received at the same or nearby base stations. While self-interference (a radio hearing its own transmission) can be cancelled, this cross-mode interference often cripples uplink performance. The authors address this by moving away from the binary choice of Half-Duplex (0% overlap) or Full-Duplex (100% overlap), proposing a flexible $\alpha$-duplex model where the overlap is a tunable parameter.
Methods
The authors model a single-tier cellular network where each cell can choose its own spectrum overlap fraction ($\alpha$) and transmission power levels.
Key technical components include:
- Pulse-Shaping Analysis: Unlike many studies that assume idealized rectangular pulses, this paper uses more realistic Sinc and Sinc² pulse shapes. This allows them to calculate an "effective cross-mode interference factor" based on how much the signals actually bleed into each other's frequency space.
- Optimization Framework: They formulate a non-convex optimization problem to maximize network utility. They test two objective functions: Sum-Rate (maximizing total throughput) and Sum-Log-Rate (prioritizing fairness between uplink and downlink).
- Algorithmic Approach: To solve the non-convex problem, they employ the Interior Point Method using a logarithmic barrier function. This allows them to iteratively find optimal power levels and overlap fractions ($\alpha$) for every base station in the network.
Datasets
The paper does not use a real-world dataset; instead, it utilizes a simulation environment based on 3GPP urban macro-cell standards.
- Setup: 9 base stations with an inter-site distance of 500 meters.
- Parameters: 20 MHz channels, pathloss modeled as $l(d) = 22 \log_{10}(d) + 28 + 20 \log_{10}(f_c)$.
- Benchmarks: The proposed scheme is compared against traditional Half-Duplex (HD), standard Full-Duplex (FD), and a previously proposed $\alpha$-duplex scheme that used fixed (non-optimized) parameters.
Key Insights and Ideas
- The Uplink Vulnerability: The results confirm that the uplink is the "weak link." In pure FD mode, the uplink rate can drop by as much as 75% compared to HD because it is drowned out by downlink power.
- $\alpha$ as a Control Knob: Partial overlap acts as a middle ground. By choosing an $\alpha$ between 0 and 1, the network can harvest some of the bandwidth gains of FD without completely sacrificing the uplink.
- Utility Choice is Critical: If you optimize for "Sum-Rate," the algorithm tends to favor the downlink because it’s easier to get high rates there. To truly protect the uplink, a "Sum-Log-Rate" utility is necessary, as it penalizes cases where one link (the uplink) performs very poorly.
- Per-Cell Optimization: The paper shows that a "one-size-fits-all" overlap fraction for the whole network is inefficient. Allowing each cell to tune its $\alpha$ based on local interference conditions provides a significant performance boost.
Key Results
- Throughput Gains: The optimized $\alpha$-duplex scheme achieved a 43% sum-rate gain over traditional Half-Duplex and a 10% gain over standard Full-Duplex.
- Uplink Protection: Compared to standard FD, the proposed scheme improved uplink rates by 42%.
- Downlink Improvement: Downlink rates saw improvements of up to 80% compared to statistical models used in prior literature.
- Power Disparity Threshold: The authors identified a critical threshold: if the uplink transmit power is less than ~1.2% of the base station power, standard FD actually performs worse than HD for the network as a whole. The $\alpha$-duplex scheme remains robust even below this threshold.
Limitations and Future Work
- Perfect SIC Assumption: The study assumes perfect self-interference cancellation (SIC). In real-world hardware, residual self-interference remains a major hurdle that could degrade these theoretical gains.
- Static Pulse Shapes: While the authors used Sinc pulses, they treated the pulse shapes as fixed. They note that optimal pulse shaping—designing the signal filter specifically to minimize cross-mode interference—is a promising direction for future research.
- Computational Complexity: The interior point method is effective but can be computationally expensive for very large-scale, real-time deployments. Future work might look into decentralized or lower-complexity heuristic algorithms.
Summary updated