Coexistence-Aware Antenna Tilt and Resource Allocation for Cellular-Connected UAV Corridors

This work has been accepted for publication in the IEEE ANTS 2026 Conference WS-08.

Coexistence-Aware Antenna Tilt and Resource Allocation for Cellular-Connected UAV Corridors Tarun Singh Rawat

Pradyumna Kumar Bishoyi

Department of Electrical Engineering Indian Institute of Technology Jodhpur Rajasthan, India m25cps010@iitj.ac.in

Department of Electronics Engineering Indian Institute of Technology Jodhpur Rajasthan, India pradyumna@iitj.ac.in

arXiv:2611.06832v1 [cs.NI] 4 Nov 2026

Abstract—Cellular-connected UAV corridors reuse terrestrial infrastructure to provide predictable aerial connectivity, but coexistence with ground users (GUEs) creates competing demands on antenna tilt and physical resource blocks (PRBs). Prior work optimizes tilt for signal quality alone and does not address joint tilt-PRB planning under corridor rate and GUE delay constraints, nor the allocation policy needed when combined resource claims exceed the PRB budget. We address both gaps by formulating a joint antenna-tilt and PRB-allocation problem that maximizes GUE sum-log-rate subject to a minimum UAV corridor rate and an M/G/1-based GUE mean-delay bound. We decompose the problem into two subproblems: antenna tilts are adapted via projected gradient ascent, while PRB allocation follows a techno-economic approach. Under resource scarcity, we cast the PRB scarcity as a bankruptcy problem and apply the weighted Talmud rule to balance GUE and UAV claims. Simulation results demonstrate that relaxing the GUE delay bound consistently reduces UAV rate dissatisfaction under the proposed allocation. Index Terms—UAV Corridor, Coexistence, PRB, FR1, NTN

I. I NTRODUCTION

The growing demand for continuous and reliable uncrewed aerial vehicle (UAV) connectivity in logistics, inspection, surveillance, and urban mobility motivates research on extending cellular services to aerial users in upcoming 6G standard [1]. As UAV traffic increases, providing reliable connectivity throughout unrestricted three-dimensional airspace becomes difficult. In this context, cellular-connected UAV corridors have gained significant attention [2]–[4]. Such corridors confine UAV operations to predefined routes, thereby simplifying air traffic management and concentrating cellular coverage requirements within known aerial regions. Unlike a general cellular-connected UAV, whose trajectory is mission-dependent, or a UAV swarm, whose members coordinate their motion and tasks, UAVs sharing a corridor follow a spatially constrained route without requiring such coordination. Reusing terrestrial cellular infrastructure for UAV corridors is cost-effective but creates coexistence challenges with existing cellular ground users (GUEs) [5], [6]. UAV corridors require a minimum downlink rate along predefined routes spanning multiple base stations (BSs). Two coupled design challenges then arise. First, down-tilted BS antennas provide weak aerial coverage, whereas uptilting antennas to improve corridor coverage strengthens interference over LoS links from neighboring

BSs and can degrade GUE service [7]. Second, since UAVs and GUEs share the same limited pool of physical resource blocks (PRBs), congestion arises whenever their combined resource demand exceeds the available PRB budget. As antenna-tilt configuration and PRB allocation jointly determine the achievable rates for both user classes, cellular-connected UAV corridor systems require joint planning of the two to meet the service requirements of both. In [2], [3], the authors employ quantization theory to optimize the vertical antenna-tilt of existing cellular base stations, with the objective of maximizing the average received signal strength (RSS) for both GUEs and UAVs. While [4] extends this approach via Bayesian Optimization to maximize the mean signal-to-interference-noise ratio (SINR) of both users. In [8], the authors compare single- and dual-antenna designs for UAV corridor systems. Although these works improve UAV corridor connectivity, signal-quality optimization alone does not address joint antenna tilt and PRB planning under corridor rate and GUE delay requirements, nor does it provide an explicit allocation policy for when combined resource demands exceed the PRB budget, a gap that makes congestion management essential to network design. In this work, we address two coupled network-design questions for cellular-connected UAV corridor systems. First, how should the BS antenna tilts be configured to guarantee the required rate along the corridor without significantly degrading GUE throughput? Second, how should the limited PRBs be allocated to satisfy GUE delay requirements and corridor rate demands, particularly when their aggregate resource claims exceed the available budget? To address these questions, we develop an economics-inspired resource-allocation framework that jointly optimizes antenna tilts and PRB allocations to maximize the sum of logarithmic GUE rates while imposing a minimum downlink rate at every sampled corridor location. We model each BS’s aggregated GUE traffic as an M/G/1 queue and derive a closed-form service-rate threshold ensuring queue stability and the prescribed cell-level mean-delay bound. The service-rate threshold determines each BS’s minimum GUE PRB reservation. For each UAV corridor location, we derive the minimum PRB requirement from its downlink rate constraint. These quantities define the user’s resource claims. When the available budget covers all claims, we satisfy them

dB is given by Ab,q = Amax + AVb,q + AH b,q , where Amax denotes the maximum antenna gain at the boresight, AVb,q and AH b,q represent the vertical and horizontal antenna gains in dB, respectively, given by AVb,q = − θ212 [θb,q − θb ]2 and 3dB 12 AH [ψb,q − ψb ]2 . The vertical antenna gain AVb,q 2 b,q = − ψ3dB depends on the vertical tilt θb of BS b and on the elevation angle θb,q between BS b and user location q ∈ E, calculated as   hq − hb,B θb,q = tan−1 , (1) ||q − pb ||

Figure 1. Illustration of Cellular-connected UAV corridor system.

via a water-filling algorithm; otherwise, we cast the problem as a bankruptcy problem [9], and apply the weighted Talmud allocation (WTA) rule [10]. The main contributions of our work are as follows, • We formulate a joint antenna-tilt and PRB-allocation problem to maximize the GUE throughput utility while ensuring UAV corridor minimum-rate requirements, GUE meandelay constraints, and per-BS PRB budgets. • We decompose the joint problem into antenna-tilt and PRB-allocation sub-problems, and solve them iteratively. The antenna-tilt subproblem is solved using projected gradient ascent with a logarithmic barrier to ensure UAVrate feasibility. For fixed tilts, each BS uses water-filling when all minimum claims are feasible and WTA otherwise. • Numerical results show that the fraction of UAVs failing to meet the minimum-rate requirement decreases under WTA as the GUE mean-delay constraint is relaxed. II. S YSTEM M ODEL

where hq , hb.B , pb represents height of user q , height of BS b and 2D coordinates of BS b respectively. The horizontal antenna gain AH b,q depends on the difference [ψb,q − ψb ] ∈ [−180◦ , +180◦ ], where ψb is azimuth orientation of BS b and ψb,q is the azimuth angle between BS b and user location q, given by

ψb,q =

   qy − pb,y −1 ◦   tan + 180 × 2c,   qx − pb,x 

if qx − pb,x > 0,

    qy − pb,y  −1 ◦  + 180 × (2c + 1),  tan qx − pb,x

if qx − pb,x < 0.

(2)

where subscripts x and y denote the horizontal and vertical coordinates of a point, respectively, and the integer c is chosen in a way that −180◦ ≤ ψb,q − ψb ≤ +180◦ . Channel model: We consider BSs are equipped with a directive antenna with a maximum gain of 14 dBi and a vertical (resp. horizontal) half-power beamwidth of 10◦ (resp. 65◦ ). All UEs are equipped with a single omnidirectional antenna. For notational compactness, q ∈ E denotes either a GUE g or a UAV corridor location a. Let Gb,q (θb ) denote the large-scale power gain from BS b to user q, incorporating path loss, lognormal shadowing, and the antenna gain. Let hb,e denote the normalized small-scale fading coefficient. We assume Rayleigh fading for GUE links and pure LoS propagation for UAV links, modeled as hb,g ∼ CN (0, 1), and |hb,a |2 = 1, respectively.

A. Network Topology We consider a cellular network that consists of a set of B = {1, 2, · · · B} base stations (BSs) operating in frequency range 1 (FR1). The nominal inter-site distance (ISD) between neighboring BSs is denoted as dISD . We consider two types of users are being served by the BSs. First, the ground UEs (GUEs), indexed by g ∈ Eg , are distributed uniformly around BS. Second, UAVs are inside the corridor at an altitude ha from the ground. The UAV locations in the corridor are indexed by a ∈ A. The complete evaluation set is denoted as E = Eg ∪ A. Each BS b ∈ B provides service to both GUEs and UAVs using a finite number of physical resource blocks (PRBs). We consider a total of Nb PRBs for BS b, and the per-PRB bandwidth is denoted as BPRB (Hz). Further, the azimuth boresight direction ψb , and electrical tilt angle θb ∈ [θmin , θmax ]. For a given tilt vector θ, let Ubg (θ), and Uba (θ) denote the set of GUEs and UAVs associated with BS b, respectively. Further, Ub (θ) = Ubg (θ) ∪ Uba (θ), denotes all users associated with BS b. Antenna gain: The vertical and horizontal half-power beamwidths of the directional antennas are denoted by θ3dB and ψ3dB , respectively. The total antenna gain of BS b in

B. Downlink SINR Assuming full frequency reuse, the downlink SINR of user e ∈ Ub (θ) on a PRB is given by γe (θ) =

PbPRB Gb,e (θb )|hb,e |2 X

2 PjPRB Gj,e (θj )|hj,e |2 + σPRB

.

(3)

j∈B\{b}

Here, PbPRB denotes the transmit power per PRB at BS b, and 2 σPRB is the noise power over bandwidth BPRB . Setting e = g or e = a yields the GUE SINR γg (θ) or the UAV SINR γa (θ), respectively. C. Downlink Communication Throughput We consider downlink communication throughput as a performance metric for both GUEs and UAVs. For GUEs, we consider the downlink log-sum rate, which is commonly used for maintaining fairness [11]. Given antenna tilt and PRB allocation, the log-rate is X Ug (θ, n) = log(ng BPRB log2 (1 + γg (θ))) . (4) g∈Ubg

Similarly, the downlink communication rate along the UAV corridor points indexed by a ∈ A is Uu (θ, n) = na BPRB log2 (1 + γa (θ)) .

(5)

D. GUE delay modelling We model the aggregate downlink GUE traffic at BS b as an M/G/1 queue. Packets for GUE g ∈ Ubg arrive according to an independent Poisson process with rate λg packets/s and mean size Lg , yielding an offered load Λg = λg L̄g . By the superposition property of independent Poisson processes,Pthe aggregate arrival rate and traffic load at BS b are λtot b = g∈Ubg λg and P Λb = g∈U g Λg , respectively. For fixed antenna tilts (θ) and b PRB allocations (ng ), the aggregate GUE downlink rate is X rb = BPRB ng log2 (1 + γg (θ)) , (6) g∈Ubg

We approximate the aggregate queue as a constant-rate, firstcome-first-served server with packet service time Sb = Lb /rb . Under the stability condition rb > Λb , the mean sojourn time including queueing and transmission can be expressed using the Pollaczek–Khinchine formula as
Λb L̄b 1 + CL2 b L̄b (7) + . Wb = rb 2rb (rb − Λb ) P P where L̄b = ( g∈U g λg L̄g )/( g∈U g λg ). CL2 b denotes the b b squared coefficient of variation. Each BS must satisfy Wb ≤ Dcell , where Dcell > 0 denotes the prescribed cell-level meandelay bound. The following proposition converts this requirement into a minimum service-rate condition. Proposition 1. For a BS b with Λb > 0, the stability condition rb > Λb and mean-delay requirements Wb ≤ Dcell are jointly satisfied if and only if rb ≥ rb∗ , (8) where

rb∗ =

v" # u u Dcell Λ − L̄
2 b b cell t
D Λb + L̄b + + 2Dcell Λb L̄b 1 + CL2 b

in proportion to the offered traffic loads. We define each GUE’s requirement as Λg GUE,min cg = N , (10) Λb b A GUE with heavier traffic (i.e., a larger Λg /Λb ) receives a proportionally larger reservation. Thus, the individual GUE P requirements sum to the total reservation at BS b, i.e., g∈U g cg = b

NbGUE,min . Under this allocation, we obtain the minimum reservation required to provide rb∗ as   ∗ r GUE,min b . (11) Nb = P   Λ  BPRB g∈Ubg Λgb log2 (1 + γg )  E. Problem Formulation Our goal is to determine optimal antenna tilt and PRB allocation that maximizes the aggregated sum-rate of GUEs while ensuring minimum downlink rate at each UAV corridor point and adhering to GUE mean-delay constraints. The optimization problem is formulated as P0 : maximize

θ,{ng },{na }

s. t

Ug (θ, n) Uu (θ, n) ≥ Rumin , X X ng + na ≤ Nb , ∀ b ∈ B, g∈Ubg

a∈Uba

ng ≥ cg ,

g ∈ Ubg , ∀b ∈ B,

θmin ≤ θb ≤ θmax ,

∀b ∈ B,

(12) (13) (14) (15)

Constraints (12)-(15)enforce the corridor rate, the per-BS PRB budget, the delay-driven GUE PRB floor, and the tilt range, respectively. Problem P0 is nonconvex since both the utility functions (Ug (·) and Uu (·)) are not jointly concave in the antenna tilts (θ) and PRB allocations (ng ). In Section III, we develop a solution framework for this coupled design and address resource allocation when the available PRBs cannot satisfy all minimum service requirements. III. P ROPOSED S OLUTION

. (9) 2Dcell Proof: Under the queue-stability condition rb > Λb , the term 2rb (rb −Λb ) is strictly positive. Substituting (7) into Wb ≤ Dcell and multiplying 2rb (rb − Λb ) both sides we get
2L̄b (rb − Λb ) + Λb L̄b 1 + CL2 b ≤ 2Dcell rb (rb − Λb ).

We adopt a two-step approach to solve the above optimization problem. We decouple the tilt-dependent SINR and PRB allocation by alternating between two subproblems until convergence. For fixed PRB allocations ({ng }, {na }), we optimize the BS tilts; for fixed tilts, we optimize the PRB allocation independently at each BS.

Rearranging the above inequality with respect to rb yields 2
Λb L̄b (1−CL ) b Dcell rb2 − Dcell Λb + L̄b rb + ≥ 0. Let fb (rb ) 2 denote the left-hand side of the inequality. Since Dcell > 0, fb is an upward-opening quadratic, and substituting rb = Λb gives fb (Λb ) = −Λb L̄b (1 + CL2 b )/2 < 0, so Λb lies strictly between its two roots. Hence, within the stable region rb > Λb , fb (rb ) ≥ 0 holds iff rb is no smaller than the larger root, which by the quadratic formula is rb∗ in Eq. (9). Since rb∗ > Λb , the condition rb ≥ rb∗ also guarantees queue stability. To translate the service-rate threshold in Proposition 1 into PRB requirements, we distribute the GUE reservations at BS b

A. Optimizing Antenna Tilt To decouple the antenna tilt (θ) from PRB allocation, we formulate a modified optimization problem that retains the tilt-dependent component of the objective in the problem P0. Further, the UAV rate constraint in Eq. (12) is incorporated through an equivalent spectral-efficiency feasibility threshold. Let nmax be the maximum PRB that a BS can allocate to a a UAV. We define the required spectral-efficiency threshold as τa = (Rumin )/(nmax BPRB ). Accordingly, UAV-rate feasibility a during tilt optimization requires log2 (1 + γa (θ)) ≥ τa . Clearly, τa ensures UAV link feasibility throughout the corridor. The

modified objective function for optimizing antenna tilt in our considered scenario is X log(log2 (1 + γg (θ))) Φ(θ) = g∈Eg

+κ

X

log(log2 (1 + γa (θ)) − τa ) .

(16)

a∈A

The first term preserves the GUE utility in Problem P0, while the second is a logarithmic barrier enforcing UAV spectralefficiency feasibility, with κ > 0 controls the strength of the barrier. To this end, the antenna tilt optimization problem is formulated as, P1 :

max Φ(θ) θ∈Θ

|B|

where Θ = [θmin , θmax ] is the feasible set. We solve P1 using projected gradient ascent [12]. At iteration t, the antenna-tilt vector is updated as h
i θ (t+1) = ΠΘ θ (t) + ηt ∇θ Φ θ (t) (17) where ηt > 0 is the step-size, ΠΘ (·) denotes projection onto the feasible tilt set Θ. The iteration continues till the difference between the θ (t) in two consecutive iterations is sufficiently small. B. Optimized PRB Allocation Now, in the second stage, we aim to find the optimal PRB allocation for GUEs and UAVs associated with each BS. For a fixed antenna tilt θ ∗ , PRBs are allocated independently at each BS. Therefore, the PRB allocation subproblem can be formulated as X P2 : max log(ng ) s. t (12), (13), (14) {ng },{na }

g∈Ubg

denote the minimum PRB requirement for UAV. Let nmin a Then, from the Eq. (12), the expression for nmin can be a m l min R min u expressed as na = BPRB log (1+γ ∗ ) . Using Eq. (11) and 2 a nmin , the total PRB demand at BS b can be expressed as a X Fb = NbGUE,min + nmin (18) a a∈Uba

Clearly, two regimes arise according to whether Fb ≤ Nb , as explained in subsequent subsections. 1) Feasible regime: When Fb ≤ Nb , the PRB budget of BS b is sufficient to meet the minimum resource demand. We assign each UAV a minimum requirement nmin and allocate the a remaining PRBs among GUEs using the traditional water-filling algorithm. The GUE allocation is n∗g = max (cg , ν ∗ ), where the optimal water level ν ∗ satisfies X X max (cg , ν ∗ ) = Nb − nmin (19) a g∈Ubg

a∈Uba

2) Resource-deficit regime: When Fb > Nb , Problem P2 is infeasible because all minimum claims cannot be simultaneously satisfied. We therefore model this using bankruptcy

theory [10], which allocates a limited resource among users whose aggregate claims exceed its availability. We employ the weighted Talmud allocation (WTA) rule to account for both heterogeneous resource claims and service priorities [10]. The WTA rule extends the classical Talmud allocation by incorporating user-specific priority weights alongside their resource claims. More specifically, the WTA is characterized by a triplet (Eb , cb , ω b ), ∀b ∈ B, where Eb = Nb is the estate, c is the claim vector consisting of cg , ∀g ∈ Ubg and nmin , ∀a ∈ Uba . a b b Finally, ω b = (ωg , ωa ) is the weight vector that represents the user’s priority, where ωgb and ωab are the weights assigned by BS b for GUEs and UAVs, respectively. Algorithm 1: Per-Cell PRB Allocation Inputs : Ubg ,Uba , {cg }g∈U g , {nmin a }a∈Uba , and Nb P b Outputs: {ne }e∈Ub satisfying e∈Ub ne ≤ Nb P Set ca ← nmin a , for each UAV; Compute Fb ← e∈Ub ce if Fb ≤ Nb then // Water-filling allocation Assign xa ← nmin a , for each UAV Determine ν ∗ from Eq. (19) Assign xg ← max(cg , ν ∗ ), for each GUE else // Weighted Talmud allocation Eb ← Nb if Eb ≤ Fb /2 then Find λ by bisection using Eq. (20) ; xe ← min(ce /2, λωe ), ∀e ∈ Ub else Find µ by bisection using Eq. (21); xe ← ce − min(ce /2, µ/ωe ), ∀e ∈ Ub end end ne ← ⌊xe ⌋, ∀e ∈ Ub .

The WTA rule is governed by the half-claim principle and operates in two regions depending on the available estate relative to one-half of the aggregated claim. • Scarcity region: When Nb ≤ (Fb /2), the budget cannot cover more than half the aggregate claims. We allocate PRBs in proportion to the priority weights until individual allocations reach half their respective claims   ce xe = min , λωe (20) 2 where λ is a tuning parameter that is chosen such that P e∈Ub xe = Nb . Before reaching the cap, a larger weight gives a user a larger allocation. The cap limits each user’s share relative to its own requirement. • Relative-abundance region: When Fb /2 < Nb < Fb , the budget covers more than half the aggregate claims but remains insufficient to satisfy them fully. We therefore allocate the deficit Fb −Eb as losses from the users’ claims:   ce µ , , (21) xe = ce − min 2 ωe tuning parameter µ is  ce µ e∈Ub min 2 , ωe = Fb − Eb .

The P

chosen

such

that

6

8

1

5

−500

18

0.6

13 10

14

0.4

19

0.2 7

3

0.8

CDF

Corridor 1

y (m)

0

16

Corridor 2

2

500

1.0

BS sites

11

1000

12

4

17

0.0

Corridor 1 GUEs Corridor 1 UAVs

0

15

−1000

−500

0 x (m)

500

20 30 Throughput (Mbps)

40

50

(a) GUE and UAV throughput under water-filling.

9

−1000

10

Corridor 2 GUEs Corridor 2 UAVs

1.0

1000

0.8

Figure 2. A 19-site deployment with 500 m ISD and UAV corridors at 150 m and 120 m. Colors indicate the sites serving each corridor. Corridor 1 contains 5 UAVs and Corridor 2 contains 10 UAVs; sites are populated with 50 GUEs per BS.

IV. P ERFORMANCE E VALUATION Setup and Scenario: We consider the cellular deployment in Figure 2, with three sectors per BS oriented at 0◦ , 120◦ , and 240◦ . We consider 5G NR FR1 carrier at 3.5 GHz with a channel bandwidth of 100 MHz and subcarrier spacing of 30 KHz. Each BS has a PRB budget of Nb = 275 PRBs. The BS’s transmitting power is set to 43 dBm, uniformly distributed across its PRBs. Unless otherwise stated, we set Dcell = 50 ms and Rumin = 10 Mbps, η = 0.02, and barrier coefficient κ = 10. Simulation Results and Discussion: Figure 3 presents the downlink throughput cumulative distribution functions (CDFs) of GUEs (solid line) and UAVs (dashed line) over 104 Montecarlo realizations under the two PRB-allocation regimes. Figure 3(a) shows the distribution under a feasible regime, i.e., the minimum resource demand is less than the BS PRB budget. We observe that the UAV throughput generally exceeds GUE throughput, with Corridor 2 providing higher median UAV throughput than Corridor 1. This is because the antenna-tilt favoring corridor UAVs reduces antenna gain towards GUEs. UAV reservations also limit the PRBs available to GUEs, whose service adequacy depends on the mean-delay requirement. Figure 3(b) shows the distributions under WTA when aggregate claims exceed the PRB budget. UAV rates concentrate near 10 and 18 Mbps in Corridors 1 and 2, respectively, while

CDF 0.4 0.2 0.0

Corridor 1 GUEs Corridor 1 UAVs

0

10

20 30 Throughput (Mbps)

Corridor 2 GUEs Corridor 2 UAVs

40

50

(b) GUE and UAV throughput under WTA rule. Figure 3. Throughput CDFs for GUEs and UAVs in the two corridor scenarios: (a) water-filling; (b) WTA under PRB scarcity.

25 20

Claimed PRBs Allocated PRBs Equal split

15

PRBs

Here we can observe that the higher the value of ωe , i.e., the higher the priority of users, the smaller the resource loss in the relative-abundance region. The WTA relaxes (12) and (14) in problem P0 while guaranteeing each user at least half its claim in the relative-abundance region. The detailed allocation rule and corresponding per-BS PRB allocation are summarized in Algorithm 1. Complexity analysis: Let Mb = |Ub | and Gb = |Ubg |. Waterfilling algorithm requires O(Mb + Gb log Gb ), where as WTA rule requires O(Mb Ib ), where Ib is the number of bisection iterations.

0.6

10 5 0

G1 G2 G3 G4 G5 G6 G7 G8 G9 G10 G11 G12 U1 U2 U3 U4 Users associated with Site 2 (12 GUEs + 4UAVs)

Figure 4. PRB claims and allocations under the proposed method and equal sharing for 12 GUEs and four UAVs at Site 2.

GUE rates remain low. We observe that compared with Figure 3(a), Figure 3(b) exhibits a narrower UAV throughput range and stronger concentrations at specific rates. This is because the WTA distributes the insufficient budget according to claims and priorities. It therefore manages resource shortfalls without guaranteeing all service requirements are met. Figure 4 compares individual PRB claims with the proposed allocation and equal sharing. We randomly select 12 GUEs and four UAVs associated with Site 2, as shown in Figure 2. Equal sharing assigns the same number of PRBs to each user, regardless of their requirements. It exceeds the claims of G1 and U1 but fails to satisfy G10 and U4. However, our proposed method meets all displayed claims before distributing surplus PRBs among GUEs. It achieves this by accounting for traffic-dependent GUE requirements and rate-dependent UAV

requirements. Figure 5 illustrates how GUE delay tolerance

Unsatisfied UAVs (%)

100 80 60 40 20 0

Del25 50 ay d ead 75 line (ms 100 )

2

16 13 te i 12 BS S g n i v Ser

Figure 5. Unsatisfied UAV percentage versus GUE mean-delay bound in the resource-deficit regime.

affects UAV rate satisfaction in the resource-deficit regime. A larger delay deadline value indicates applications that tolerate longer delivery times, rather than arbitrarily relaxing a fixed application requirement. Increasing it from 25 to 100 ms reduces UAV dissatisfaction at all four serving sites. This relaxation lowers the GUE service-rate threshold (Eq. (8)) and corresponding PRB claims, allowing WTA to allocate more resources to UAVs. Thus, accommodating GUE applicationspecific delay requirements can improve UAV corridor service without increasing the PRB budget.

Unsatisfied GUEs (%)

100 80 60 42.5

40

31.8

20 0

11.2 2.4

25

50 100 Corridor 1 height (m)

150

Figure 6. Percentage of unsatisfied GUEs versus UAV corridor height.

Figure 6 shows the percentage of unsatisfied GUEs as the UAV corridor altitude is increased from 25 m to 150 m under the resource-deficit regime. Unsatisfied GUEs rise sharply from 2.4% to 42.5%, as higher altitudes improve UAV line-of-sight to more BSs, inflating UAV PRB claims and biasing antenna tilts toward UAV coverage. This forces WTA rule to divert more PRBs away from GUEs, degrading their service. V. C ONCLUSIONS In this work, we proposed a joint antenna-tilt and PRBallocation framework for cellular-connected UAV corridors,

considering GUE mean-delay guarantees with UAV minimumrate requirements under per-BS PRB budgets. Antenna tilt and PRB allocation are decoupled into two subproblems: tilt adaptation via projected gradient ascent and PRB allocation via water-filling when claims are feasible, or via a weighted Talmud rule under PRB scarcity. Numerical results show that WTA reduces UAV rate dissatisfaction as GUE delay tolerance is relaxed, demonstrating the efficacy of the proposed approach. R EFERENCES [1] Y. Wang et al., “Toward Realization of Low-Altitude Economy Networks: Core Architecture, Integrated Technologies, and Future Directions,” IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 5, pp. 2788–2820, 2025. [2] S. Karimi-Bidhendi et al., “Analysis of UAV Corridors in Cellular Networks,” in Proc. IEEE ICC, 2023, pp. 1–6. [3] ——, “Optimizing Cellular Networks for UAV Corridors via Quantization Theory,” IEEE Trans. Wirel. Comm., vol. 23, no. 10, pp. 14 924–14 939, 2024. [4] M. Benzaghta et al., “Designing Cellular Networks for UAV Corridors via Bayesian Optimization,” in Proc. IEEE GLOBECOM, Dec 2023, pp. 4552–4557. [5] M. Bernabè et al., “Gradient-Based Optimization of Terrestrial Cellular Networks for Aerial Highways,” IEEE Trans. Veh. Techn., vol. 75, no. 5, pp. 8396–8413, 2026. [6] P. K. Bishoyi, C. C. Lee, N. Keshtiarast, and M. Petrova, “Dynamic interference management for tn–ntn coexistence in the upper mid-band,” in Proc. IEEE ICC 2026, 2026, pp. 1–6. [7] H. K. Armeniakos et al., “Performance Analysis and Experimental Validation of UAV Corridor-Assisted Networks,” IEEE Transactions on Communications, vol. 74, pp. 3830–3844, 2026. [8] A. K. Gupta et al., “Optimal antenna tilt configurations in hybrid terrestrial/ non terrestrial network,” in Proc. IEEE COMSNETS, 2026, pp. 366–374. [9] W. Thomson, “On the axiomatics of resource allocation: interpreting the consistency principle,” Economics & Philosophy, vol. 28, no. 3, pp. 385– 421, 2012. [10] R. J. Aumann and M. Maschler, “Game theoretic analysis of a bankruptcy problem from the Talmud,” Journal of Economic Theory, vol. 36, no. 2, pp. 195–213, 1985. [11] P. K. Bishoyi and S. Misra, “Enabling Green Mobile-Edge Computing for 5G-Based Healthcare Applications,” IEEE Trans. Green Commun. and Netw., vol. 5, no. 3, pp. 1623–1631, 2021. [12] S. Boyd and L. Vandenberghe, Convex optimization. Cambridge university press, 2004.

Related documents

• LLM-Assisted Hierarchical Interference Control for Satellite-Terrestrial Integrated Networks#1297460
• Enhancing 3GPP Urban Channel Models For Terrestrial-to-Non-Terrestrial Communication#792344
• Enhancing 3GPP Urban Channel Models For Terrestrial-to-Non-Terrestrial Communication#788344
• AI-Enabled Wireless Propagation Modeling and Radio Environment Maps for 5G Aerial Wireless Networks#1063294
• A Comprehensive Survey of Wireless Foundation Models for AI-Native 6G Networks#609127