Harvesting full-duplex rate gains in cellular networks with half-duplex user terminals

Ahmad AlAmmouri, Hesham ElSawy, Mohamed-Slim Alouini

TL;DR

You don’t need expensive Full-Duplex (FD) hardware in every smartphone to reap the benefits of next-generation cellular speeds. This paper demonstrates that by upgrading only the base stations to FD and using a "3-node topology"—where one base station serves two different half-duplex users simultaneously—networks can achieve nearly the same performance gains as a system where every device is fully duplex.

What Problem Does This Paper Address?

In traditional wireless networks, devices operate in Half-Duplex (HD) mode: they either talk (uplink) or listen (downlink), but never both at the exact same time on the same frequency. Full-Duplex (FD) communication promises to double the capacity by allowing simultaneous transmission and reception.

However, FD comes with two massive headaches:

  1. Self-Interference (SI): A device’s own transmission is so loud it "deafens" its own receiver. Canceling this requires complex, expensive hardware.
  2. Network Interference: In a large-scale network, having twice as many active transmitters at any given time creates a chaotic interference environment that can degrade performance.

While base stations can afford the bulky hardware for SI cancellation, putting it into every consumer smartphone is cost-prohibitive. The authors investigate whether we can "harvest" FD gains by keeping the users on HD hardware and using the base station as a bridge between an uplink user and a downlink user (the 3-node topology or 3NT).

Methods

The researchers developed a unified mathematical framework using Stochastic Geometry. They modeled the locations of base stations and users as a Poisson Point Process (PPP), which allows for a realistic analysis of interference in a large, random network.

Key novel components include:

  • $\alpha$-duplex Scheme: A flexible design parameter ($\alpha$) that allows for partial overlap between uplink and downlink channels. This captures everything from pure HD ($\alpha=0$) to pure FD ($\alpha=1$).
  • Realistic Pulse Shaping: Unlike many theoretical papers that assume "perfect" square signals, this model accounts for time-limited pulse shapes (rectangular and triangular) and matched filtering, which more accurately reflects how signals bleed into adjacent frequencies.
  • 3-Node Topology (3NT) vs. 2-Node Topology (2NT): They rigorously compared a system where the base station talks to two HD users (3NT) against a system where the base station talks to one FD user (2NT).

Datasets

As a theoretical communication theory paper, it does not use external datasets. Instead, the authors used independent system-level Monte Carlo simulations to validate their mathematical derivations. They simulated base stations across a 400 $km^2$ area to ensure their "tractable" equations matched reality.

Key Insights and Ideas

  • The "Good Enough" Base Station: The most significant finding is that upgrading the user equipment (UE) provides diminishing returns. If the base station is FD-capable, the network gains are largely realized regardless of whether the phone is HD or FD.
  • The Turning Point: There is a specific threshold for self-interference cancellation. If a phone's ability to cancel its own signal isn't nearly perfect, the 3NT approach (using HD phones) actually outperforms the 2NT approach (using FD phones) because HD phones don't suffer from self-interference at all.
  • The 5% Gap: Even under ideal conditions (perfect SI cancellation in phones), the 3NT setup only loses about 5% of the rate compared to a full FD 2NT setup.
  • Pulse Shaping as a Tool: By carefully choosing the pulse shape and the amount of channel overlap ($\alpha$), operators can maximize the uplink rate without destroying the downlink rate.

Key Results

  • Rate Gains: The study confirms that FD-capable base stations can provide substantial rate improvements (often cited around 30% in total network throughput) over traditional HD systems.
  • Validation: The analytical results (the math) perfectly matched the simulation results, proving the framework is a reliable tool for network planning.
  • UE Duplexing is Secondary: The results show that for the purpose of harvesting rate gains, the duplexing scheme of the user terminal is secondary to the duplexing capability of the base station.

Limitations and Future Work

  • Single-Tier Focus: The paper models a single-tier network. Future work could extend this to "Heterogeneous Networks" (HetNets) where macrocells and small cells coexist, creating more complex interference patterns.
  • Scheduling Complexity: The 3NT model assumes the base station can perfectly pair an uplink user and a downlink user. In practice, finding two such users with compatible channel conditions in every cell adds a layer of scheduling complexity not fully explored here.
  • User Mobility: The model assumes a snapshot in time; the impact of high-speed mobility on these FD gains remains an open question.

Reference: AlAmmouri, A., ElSawy, H., & Alouini, M. S. (2016). Harvesting Full-Duplex Rate Gains in Cellular Networks with Half-Duplex User Terminals. arXiv:1511.00207v2 [cs.IT].

Summary updated

Abstract

Full-Duplex (FD) transceivers may be expensive in terms of complexity, power consumption, and price to be implemented in all user terminals. Therefore, techniques to exploit in-band full-duplex communication with FD base stations (BSs) and half-duplex (HD) users' equipment (UEs) are required. In this context, 3-node topology (3NT) has been recently proposed for FD BSs to reuse the uplink (UL) and downlink (DL) channels with HD terminals within the same cell. In this paper, we present a tractable mathematical framework, based on stochastic geometry, for 3NT in cellular networks. To this end, we propose a design paradigm via pulse-shaping and partial overlap between UL and DL channels to maximize the harvested rate gains in 3NT. The results show that 3NT achieves a close performance to networks with FD BSs and FD UEs, denoted by 2-node topology (2NT) networks. A maximum of 5% rate loss is reported when 3NT is compared to 2NT with efficient self-interference cancellation (SIC). If the SIC in 2NT is not efficient, 3NT highly outperforms 2NT. Consequently, we conclude that, irrespective to the UE duplexing scheme, it is sufficient to have FD BSs to harvest FD rate gains.
BibTeX
@inproceedings{alammouri2016harvesting,
  author = {Ahmad AlAmmouri and Hesham ElSawy and Mohamed-Slim Alouini},
  title = {{Harvesting full-duplex rate gains in cellular networks with half-duplex user terminals}},
  booktitle = {ICC},
  year = {2016},
  url = {https://openreview.net/forum?id=K6EyzOtcUUR},
}

Loading related content...

Figure