In wireless network design, engineers have long been trapped in a binary choice when allocating frequency resources: force transmissions to be completely orthogonal, or allow them to overlap entirely.
In traditional Half-Duplex (HD) systems, uplink (user to base station) and downlink (base station to user) transmissions are strictly separated in time or frequency to prevent interference. This isolation protects signal quality but cuts potential spectral efficiency in half. Conversely, In-Band Full-Duplex (FD) communication promises to double capacity by allowing simultaneous transmission and reception on the exact same frequency. In practice, however, pure FD introduces severe "cross-mode" interference: high-power downlink transmissions from base stations easily drown out the weak, power-constrained uplink signals from mobile devices.
To break this deadlock, a powerful paradigm has emerged: partial spectrum overlap. Instead of treating interference as a binary hazard to be completely avoided, this approach treats spectral overlap as a continuous, tunable parameter. By allowing channels to partially overlap, networks can smoothly trade off spectral efficiency against manageable interference, maximizing overall throughput.
Challenging the Orthogonality Dogma
The intuition that partial overlap could be beneficial was first validated in the context of Wi-Fi networks. Historically, wireless local area networks (WLANs) strictly adhered to non-overlapping channel assignments to prevent adjacent-channel interference.
To challenge this convention, Suman Banerjee et al. introduced a novel framework in Weighted coloring based channel assignment for WLANs. They defined an "I-factor" (Interference factor) to mathematically quantify the exact amount of interference generated between two channels based on their physical frequency separation.
By modeling this relationship, the authors demonstrated in Partially overlapped channels not considered harmful that avoiding overlap entirely leaves significant network capacity on the table. Instead, strategically assigning partially overlapping channels can drastically improve spectrum reuse and overall network throughput in dense deployments, proving that a managed level of interference is a acceptable price to pay for increased bandwidth.
Formalizing the Middle Ground: The $\alpha$-Duplex
Translating this concept to cellular networks led to the formalization of the $\alpha$-duplex scheme. Rather than forcing an all-or-nothing choice between half-duplex and full-duplex modes, an $\alpha$-duplex system introduces a continuous design parameter, $\alpha \in [0, 1]$, which defines the fraction of shared bandwidth between the uplink and downlink channels.
<img src="https://images.unsplash.com/photo-1544197150-b99a580bb7a8?q=80&w=1000&auto=format&fit=crop" alt="Conceptual diagram of spectrum allocation showing a tunable overlap region between uplink and downlink bands" style="width:100%; max-width:600px; display:block; margin:20px auto; border-radius:8px;"/>
- When $\alpha = 0$, the bands are perfectly orthogonal, replicating traditional Half-Duplex operation.
- When $\alpha = 1$, the bands completely merge, replicating pure Full-Duplex operation.
- When $0 < \alpha < 1$, the system operates in a partial overlap state, creating a flexible middle ground.
As proposed by Ahmad AlAmmouri et al. in In-Band Full-Duplex Communications for Cellular Networks with Partial Uplink/Downlink Overlap, tuning this overlap ratio allows the network to harvest the capacity gains of full-duplex without subjecting weak uplink transmissions to catastrophic bit error rates.
Why Partial Overlap Works: The Intuition
To understand why an intermediate value of $\alpha$ is often optimal, consider the mathematical tension between two competing physical effects:
- The Bandwidth Gain (Linear): As $\alpha$ increases, the total effective bandwidth available to both the uplink and downlink increases linearly. More spectrum means a higher potential data rate.
- The Interference Penalty (Non-linear): As $\alpha$ increases, the frequency separation between the opposing links shrinks. This causes the cross-mode interference to scale non-linearly, rapidly degrading the Signal-to-Interference-plus-Noise Ratio (SINR).
Because the interference penalty hits power-constrained uplink devices much harder than base stations, a pure full-duplex system ($\alpha = 1$) frequently collapses the uplink rate.
By using stochastic geometry to model multi-tier cellular networks, Ahmad AlAmmouri et al. proved in In-Band α-Duplex Scheme for Cellular Networks: A Stochastic Geometry Approach that the optimal network performance lies between the two extremes. A carefully managed partial overlap can achieve simultaneous rate improvements for both links—such as a 28% gain in downlink and a 56% gain in uplink relative to half-duplex.
The authors expanded on this in Flexible Design for α-Duplex Communications in Multi-Tier Cellular Networks, demonstrating that complete spectral overlap is rarely the most efficient configuration for protecting uplink performance. Similarly, work by Itsikiantsoa Randrianantenaina et al. in Interference Management with Partial Uplink/Downlink Spectrum Overlap confirmed that adjusting the shared bandwidth parameter acts as a reliable protection mechanism for power-constrained transmissions.
Dynamic Tuning and Asymmetric Hardware
A major advantage of the partial overlap framework is its adaptability to real-world network asymmetries. Rather than enforcing a single global $\alpha$ value across an entire network, the overlap can be optimized locally.
In Interference Management in Full-Duplex Cellular Networks With Partial Spectrum Overlap, Itsikiantsoa Randrianantenaina et al. developed an optimized scheme that dynamically adjusts both the spectrum overlap fraction and transmit powers for each cell individually. This localized tuning ensures that high-power base stations in neighboring cells do not overwhelm vulnerable, edge-of-cell user signals, yielding up to a 43% sum-rate gain over traditional half-duplex systems.
Furthermore, this framework solves a massive deployment hurdle: consumer hardware costs. Implementing pure full-duplex requires expensive self-interference cancellation circuitry inside every single smartphone.
To circumvent this, Ahmad AlAmmouri et al. proposed a "3-node topology" in Harvesting Full-Duplex Rate Gains in Cellular Networks with Half-Duplex User Terminals. In this setup, only the base station is upgraded to full-duplex hardware, while the consumer user equipments (UEs) remain inexpensive, traditional half-duplex devices. The full-duplex base station communicates with an uplink user and a downlink user simultaneously over the partially overlapped spectrum.
In the accompanying mathematical proofs published in the Appendices, the authors demonstrated that this asymmetric configuration captures nearly all the capacity gains of a fully full-duplex network. This insight provides telecom operators with a highly practical evolutionary path: they can harvest next-generation full-duplex speeds by upgrading centralized infrastructure, without waiting for consumers to replace their existing smartphones.
Key Foundational Work
Partially overlapped channels not considered harmful (2006) — contributed 1 concept elements to this direction.
Weighted coloring based channel assignment for WLANs (2005) — contributed 1 concept elements to this direction.