Unlicensed Full Allocation (UFA)

• Unlicensed Full Allocation (UFA) is defined as aggressive and persistent use of unlicensed spectrum by LAA, keeping the channel ON until the FIFO queue is empty.
• Analytical models based on finite Markov chains and queueing theory assess UFA’s impact on packet acceptance, service rates, and coexistence with Wi‑Fi.
• UFA also serves as a baseline full-use assumption in HetNet optimization, highlighting trade-offs between maximizing LAA performance and ensuring Wi‑Fi fairness.

Unlicensed Full Allocation (UFA) is a paper-dependent term for aggressive use of unlicensed spectrum rather than a single standardized allocation mechanism. In the most explicit usage, UFA is the baseline unlicensed-band allocation policy for a License-Assisted Access (LAA) small cell coexisting with Wi‑Fi: once the LAA small cell gains access to an unlicensed channel, it keeps that channel in the ON state and continues serving queued LAA packets until its FIFO buffer becomes empty (Chou, 27 Sep 2025). In adjacent licensed–unlicensed resource-allocation literature, the same expression maps to baseline-like or special-case “full-use” assumptions rather than to a distinct proposed optimization scheme, notably in heterogeneous-network spectrum allocation (Zhou et al., 2015) and in assistive shared-spectrum extensions of Configured Grant for semi-deterministic traffic (Singh et al., 2020).

1. Terminological scope

The term is not used uniformly across the cited literature. In one line of work it is an explicit coexistence policy; in others it is an interpretive label for a full-use assumption or for assistive use of unlicensed/shared spectrum. This makes terminological precision necessary when comparing results across papers.

Source	Meaning of UFA	Status in the paper
"Unlicensed Band Allocation for Heterogeneous Networks" (Chou, 27 Sep 2025)	Unlicensed Full Allocation; the channel stays ON until the LAA queue empties	Explicit baseline policy
"Licensed and Unlicensed Spectrum Allocation in Heterogeneous Networks" (Zhou et al., 2015)	Extreme/full-use case of the unlicensed band	Baseline / special case, not a named optimization scheme
"Configured Grant for Semi-Deterministic Traffic for Ultra-Reliable and Low Latency Communications" (Singh et al., 2020)	Assistive shared-spectrum usage interpreted as Unlicensed Full Allocation	Concept clearly present, acronym not explicit

Two recurring points follow from this usage. First, UFA is not the paper’s main proposed method in every source. Second, “full allocation” does not denote the same control primitive in all settings: it may refer to persistent channel occupation by LAA, to a degenerate active-set configuration in a HetNet optimizer, or to completion of missing repetitions over unlicensed/shared resources.

2. UFA as persistent channel occupation in LAA/Wi‑Fi coexistence

In the LAA/Wi‑Fi coexistence framework, UFA is defined operationally. The system is modeled as a queueing system with $D$ unlicensed channels total, $D_f$ free unlicensed channels, an LAA FIFO buffer of size $Q$, LAA packet arrivals with Poisson rate $\lambda_\ell$, Wi‑Fi packet arrivals with Poisson rate $\lambda_w$, LAA service time on an unlicensed channel with exponential mean $1/\mu_{\ell,u}$, and Wi‑Fi service time on an unlicensed channel with exponential mean $1/\mu_w$ (Chou, 27 Sep 2025). The LAA small cell operates in SDL mode, where downlink traffic is sent over unlicensed spectrum while control/uplink remain on licensed bands. It senses the unlicensed band via LBT, performs RRC and random-access signaling, and then transmits user data during the channel occupancy phase.

UFA is the most aggressive use of the unlicensed channel by LAA in that model. If an LAA packet arrives and there is a free unlicensed channel, it is assigned immediately. If no unlicensed channel is free, the packet is buffered in the FIFO queue, if space exists. The controller keeps the unlicensed channel ON until the FIFO queue becomes empty. Only then is the unlicensed channel switched OFF/released. The decisive distinction from Unlicensed Time-Division Allocation (UTA) is that UFA does not enforce a periodic ON/OFF duty cycle; the occupation interval is traffic-driven and persists as long as there is queued LAA traffic.

The same paper studies UFA together with UTA, UFAB, and UTAB. UTA alternates between sensing/occupancy and OFF states using a timer. UFAB keeps the UFA logic but adds a buffering threshold $Q_\theta$. UTAB combines periodic ON/OFF behavior with the buffering threshold mechanism. Within that family, UFA is the most LAA-aggressive basic scheme, UTA is the most explicit fairness-oriented scheme, UFAB adds buffering to UFA, and UTAB adds buffering to UTA. The paper’s main conclusion is that UFA tends to maximize LAA acceptance, while UTA/UTAB tend to protect Wi‑Fi more effectively; buffering helps balance the two.

3. Analytical model, state space, and performance measures

The analytical treatment of UFA in the LAA/Wi‑Fi setting uses a finite Markov chain / queueing model with a global system state. For UFA, the state is

$\mathbf{n}=(w,x,y,z)$

where $w$ is channel status, $D_f$0 is the number of LAA packets being served, $D_f$1 is the number of Wi‑Fi packets being served, and $D_f$2 is the number of LAA packets waiting in the FIFO queue. For UFA specifically, the channel is always ON, so

$D_f$3

and the state space is

$D_f$4

The model assumes exponential distributions for analytical tractability: sensing duration mean $D_f$5, occupancy duration mean $D_f$6, LAA packet arrival process Poisson rate $D_f$7, Wi‑Fi packet arrival process Poisson rate $D_f$8, LAA service time mean $D_f$9, and Wi‑Fi service time mean $Q$0. The paper explicitly notes that real traffic need not be exponential; the exponential assumption is used to obtain a tractable mean-value/steady-state analysis and to validate simulation.

Steady-state probabilities $Q$1 are obtained from balance equations written in the generic form

$Q$2

where $Q$3 aggregates incoming transition flows and $Q$4 aggregates outgoing transition rates from a state. Representative transitions include

$Q$5

$Q$6

$Q$7

$Q$8

and

$Q$9

Performance is expressed mainly through packet dropping probability. For UFA,

$\lambda_\ell$0

and

$\lambda_\ell$1

The corresponding acceptance rates are

$\lambda_\ell$2

Within this model, UFA gives the best LAA acceptance performance across the compared schemes, because it outperforms the other allocations in terms of LAA dropping probability $\lambda_\ell$3. Its coexistence cost is explicit: UFA is worst for Wi‑Fi fairness, and under UFA the performance between LAA and Wi‑Fi “has significantly differed and cannot be compensated for by any mechanism.” Buffering improves coexistence, with UFAB improving over UFA for Wi‑Fi and UTAB generally better balanced overall. The analytical and simulation values for UFA match closely for both $\lambda_\ell$4 and $\lambda_\ell$5, with errors described as very small and Wi‑Fi errors as especially tiny (Chou, 27 Sep 2025).

4. UFA as a special full-use case in licensed–unlicensed HetNet optimization

In the heterogeneous-network formulation of licensed and unlicensed spectrum allocation, UFA is not a named optimization scheme proposed by the authors. Instead, it is best understood as a baseline / special-case allocation behavior in a broader framework where the unlicensed band is treated as a distinct RAT with extra contention and vacation delay (Zhou et al., 2015). The system has $\lambda_\ell$6 access points, $\lambda_\ell$7 user groups, and $\lambda_\ell$8 RATs. Each RAT $\lambda_\ell$9 has bandwidth $\lambda_w$0, and the spectrum is split into patterns, where a pattern $\lambda_w$1 is a subset of APs allowed to use a time-frequency segment. The fraction of RAT-$\lambda_w$2 spectrum assigned to pattern $\lambda_w$3 is $\lambda_w$4, and user-level bandwidth allocation is represented by $\lambda_w$5.

The consistency constraints are

$\lambda_w$6

and

$\lambda_w$7

User association is implicit: group $\lambda_w$8 is associated with AP $\lambda_w$9 over RAT $1/\mu_{\ell,u}$0 if $1/\mu_{\ell,u}$1 for some $1/\mu_{\ell,u}$2.

The unlicensed RAT is modeled with lower effective spectral efficiency, reduced reliability, and additional delays due to contention and/or listen-before-talk requirements. Under the conservative model, the service rate for user group $1/\mu_{\ell,u}$3 on RAT $1/\mu_{\ell,u}$4 is

$1/\mu_{\ell,u}$5

For LTE-U on unlicensed spectrum, the paper uses a queue with single vacation and nonexhaustive service. The average delay is given by

$1/\mu_{\ell,u}$6

where $1/\mu_{\ell,u}$7, and for user group $1/\mu_{\ell,u}$8,

$1/\mu_{\ell,u}$9

Two optimization-based allocation schemes are proposed: Conservative allocation (P1) and Utilization-dependent allocation (P2). The UFA-like full-use case appears only as a degenerate subcase of P2, obtained by setting

$1/\mu_w$0

together with $1/\mu_w$1 for all $1/\mu_w$2. The paper describes this as a feasible suboptimal solution and uses it only to show that P2 reduces to P1 under that special case. In that sense, UFA is a maximal-access assumption for the unlicensed band rather than one of the paper’s main contributions.

The main simulation findings further delimit that role. The proposed schemes substantially outperform both orthogonal and full-reuse allocations. The gain is largest in heavy traffic. The utilization-dependent scheme performs best overall because it captures dynamic interference more accurately. In light traffic, the utilization-dependent scheme tends to behave more like full reuse; as traffic increases, it shifts toward more orthogonalization to mitigate interference. When unlicensed interference is low, more traffic is pushed onto unlicensed spectrum; when interference/vacation delay is high, traffic is shifted toward licensed spectrum. The utilization-dependent allocation can reduce average delay by about 40% in low traffic relative to the conservative allocation.

5. Related interpretations beyond LAA coexistence

A distinct but related use appears in Configured Grant for semi-deterministic traffic. That paper does not introduce the acronym UFA explicitly, but the concept is described under “Assistive shared spectrum usage” and is best interpreted as Unlicensed Full Allocation: a design where the UE is given a full configured-grant resource block of $1/\mu_w$3 repetitions in the licensed CG period, and any repetitions that cannot be carried there are completed over unlicensed shared spectrum such as NR-U (Singh et al., 2020). The UE transmits as many repetitions as possible within the CG transmission occasions; if the number actually sent in CG is $1/\mu_w$4 with $1/\mu_w$5, the remaining $1/\mu_w$6 repetitions are transmitted in the shared spectrum within a specified latency budget. This is proposed to address the late-arrival problem of semi-deterministic traffic in URLLC. The paper provides no actual simulations; it states that these strategies will be analyzed analytically and verified using simulations in future work.

Another adjacent formulation is continuity-aware spectrum carving. The DRL-based licensed/unlicensed coexistence paper does not implement a dedicated UFA protocol, but it addresses a closely related problem: the base station schedules resource blocks to licensed users while deliberately preserving contiguous unused RBs so that unlicensed entities can opportunistically use them with less coordination overhead and fewer interruptions (Boroujerdi et al., 2020). The state includes a continuity vector, the action can leave the current RB unallocated, and the objective includes both a spectral-efficiency term and a continuity term. The paper reports that the proposed method achieves higher average spectral efficiency than greedy baselines in the studied setting, and with larger buffer $1/\mu_w$7 reaches up to 99.6% acceptance ratio at high arrival rate.

A third related interpretation is the unlicensed-band framework based on right sharing and Value-of-Rights (VoR). In the network-slicing paper, UFA is not explicitly defined; the unlicensed band is modeled through channel access probability rather than fixed bandwidth ownership, and cooperation is represented as sharing and trading access rights (Xiao et al., 2020). The relevant abstraction is that all MNOs have equal rights to access the unlicensed band $1/\mu_w$8, with utility

$1/\mu_w$9

This places “full allocation” closer to full access opportunity than to exclusive possession of a frequency block.

The robust distributed power-allocation paper operates in shared unlicensed spectrum with multiple selfish transmitter-receiver pairs and uncertain normalized interference-plus-noise (fard et al., 2011). It is not a UFA paper in nomenclature, but it studies the same class of non-exclusive unlicensed access in which users compete over common bands. In the bounded symmetric uncertainty model, the robust Nash equilibrium simplifies to a water-filling form with an inflated interference term $Q_\theta$0. When the equilibrium is unique, the social utility at robust Nash equilibrium is less than at the nominal Nash equilibrium; when multiple equilibria exist, uncertainty may push the system toward more orthogonal utilization of bandwidth.

6. Trade-offs, misconceptions, and broader significance

Across the cited literature, UFA is consistently associated with aggressive or persistent use of unlicensed resources, but the precise trade-off depends on the model. In the LAA/Wi‑Fi coexistence setting, a longer effective occupancy phase improves LAA acceptance but harms Wi‑Fi; because UFA holds the channel until the queue empties, it is good for LAA acceptance and poor for Wi‑Fi coexistence (Chou, 27 Sep 2025). In the HetNet optimizer, the full-use interpretation is explicitly a feasible suboptimal special case, while the principal contributions are the queueing abstraction for unlicensed delay and the conservative and utilization-dependent optimization schemes (Zhou et al., 2015). In Configured Grant, assistive unlicensed/shared usage reduces licensed-spectrum over-allocation but introduces contention and possibly sensing overhead such as Listen-Before-Talk (Singh et al., 2020).

Several misconceptions can therefore be addressed directly. UFA is not a universally standardized acronym across licensed–unlicensed spectrum literature. It is not always the main proposed scheme of a paper. It does not imply fair coexistence with Wi‑Fi. It also does not necessarily mean fixed unlicensed bandwidth ownership, because some formulations allocate access probability or exploit structured vacancies rather than assign exclusive spectrum blocks (Xiao et al., 2020, Boroujerdi et al., 2020).

A plausible implication is that UFA is best treated as a family of “full-use” unlicensed access assumptions whose technical meaning must be read within the surrounding system model. In queueing-based coexistence models, it means persistent ON-state occupation until backlog is exhausted. In joint licensed–unlicensed optimization, it denotes a degenerate full-reuse state. In CG-based URLLC enhancements, it denotes completion of missing repetitions over shared or unlicensed resources. This suggests that comparisons involving UFA are methodologically valid only when the channel model, buffering model, fairness criterion, and access-control assumptions are aligned.