PaperlayerView exact revision on arXiv

Source-linked AI summary

Multi-Timescale Context-Aware Radio Resource Optimization for Harmonious NTN-TN Coexistence: A Digital Twin and Open RAN Enabled Framework

Ecem Izde Kaya, Md. Zoheb Hassan, Georges Kaddoum

arXiv:2610.32827v1cs.NIeess.SP

TL;DR

Terrestrial downlinks sharing spectrum with satellite uplinks can cause harmful aggregate interference at LEO receivers. DT-STABLE uses site-specific digital-twin maps and O-RAN control for multi-timescale beam and user scheduling. In 28 GHz simulations, it balanced rate and satellite protection while reducing measurement overhead and optimization runtime.

  • Problem

    Overlapping terrestrial FR2 and satellite spectrum creates a coexistence challenge because terrestrial downlink transmissions may cause harmful interference to satellite uplinks, including through sidelobes and environmental upscattering.

    Source: §I · p. 1

  • Method

    DT-STABLE uses site-specific channel and satellite-interference maps to coordinate beam activation and user association through a multi-timescale O-RAN control loop.

    Source: §I · p. 2, A. System Overview · p. 3

  • Results

    DT-STABLE achieved 89.63% INR satisfaction and a mean sum rate of 210.96 bps/Hz, while reducing measurement overhead by 99.71% and reaching up to 33.1x faster execution than exact optimization.

    Source: §VII · p. 15

  • Takeaways & Limitations

    In the evaluated 28 GHz scenarios, site-specific DT knowledge and O-RAN control supported resource optimization that balances terrestrial throughput with satellite protection.

    Source: §VII · p. 15

Abstract

from arXiv · show

Coexistence between non-terrestrial networks (NTNs) and terrestrial networks (TNs) is essential for future wireless systems, but terrestrial downlink transmissions can generate harmful aggregate interference at low-Earth orbit (LEO) satellite receivers. We present DT-STABLE, an Open Radio Access Network (O-RAN)-enabled, digital twin (DT)-assisted framework for context-aware NTN-TN coexistence. Using site-specific information, the DT estimates interference at LEO satellites, enabling proactive management with reduced real-time measurement overhead. We formulate joint beam activation and user association under RF-chain limits, load balancing, and aggregate satellite interference constraints; prove NP-hardness; and develop an efficient DT-guided scheduling algorithm. DT-STABLE couples short-timescale radio resource adaptation with periodic mid- and long-timescale DT updates through standardized O-RAN interfaces. Site-specific 28 GHz simulations under worst-case satellite geometry and low elevation angles show that DT-STABLE balances rate and interference utility, achieving 89.63% interference-to-noise ratio (INR) satisfaction and a mean sum rate of 210.96 bps/Hz, with nearly 100% INR satisfaction when satellite protection is prioritized. The framework reduces beam measurement overhead by 99.71% relative to exhaustive sweeping and achieves up to 33.1 times lower runtime than the Gurobi-based exact optimization benchmark.

I. INTRODUCTION

Terrestrial downlinks can create harmful aggregate interference for satellite uplinks in overlapping spectrum, motivating coordinated NTN–TN resource control. DT-STABLE combines site-specific digital-twin maps with O-RAN control to coordinate beam activation and user association across timescales.

I. INTRODUCTION

  • Spectrum overlap between terrestrial FR2 downlinks and satellite bands creates a coexistence challenge because terrestrial transmissions may harm satellite uplinks, with sidelobes and environmental upscattering contributing to aggregate interference.The concern is network-wide interference, not only the emissions of an individual beam or base station.Source: §I · p. 1
  • DT-STABLE uses site-specific channel and interference maps for proactive beam activation and user scheduling under satellite-visibility, aggregate-interference, RF-chain, and beam-load constraints.Its O-RAN control loop detects prediction mismatches and selectively triggers low-overhead beam sweeping for DT refinement.Source: §I · p. 2
  • The coexistence problem is formulated as constrained joint beam activation and user association that balances terrestrial service utility with satellite protection, and the resulting optimization problem is NP-hard.DT-STABLE combines long-term CKM construction, mid-term refinement, and short-term beam–user scheduling.Source: §I · p. 2

II. RELATED WORKS

  • Prior work addresses NTN resource management, interference mitigation, and digital twins, while the proposed framework integrates site-specific context into resource control and coordinates it through standardized O-RAN interfaces.The authors distinguish their approach from work that uses environmental information mainly for channel modeling or evaluation.Source: §II · pp. 2–3

A. System Overview

  • The system uses precomputed coverage and interference maps to estimate candidate beam rates and satellite exposure, with localized corrections when predictions mismatch; it assumes UE positions are available and neglects inter-cell and intra-cell inter-beam interference.A purely measurement-driven approach would require O(UJK) and O(JKM) measurements per scheduling epoch.Source: A. System Overview · pp. 3–4

B. DT-STABLE Framework

The framework links long-term digital-twin mapping, mid-term feedback-driven refinement, and short-term resource control through O-RAN interfaces. Its context-aware digital twin combines site data and propagation modeling into maps used by the near-real-time controller.

B. DT-STABLE Framework

  • The radio-twin rApp builds long-term DT maps, near-RT RIC measurements identify prediction mismatches for refinement, and the DT-STABLE xApp uses updated maps and RAN state to choose beams and associate users.The process uses O1, R1, A1, and E2 interfaces to collect context, deliver maps, receive feedback, and apply control decisions.Source: B. DT-STABLE Framework · p. 4

A. Modeling of the Context-Aware Digital Twin (CDT)

  • The context-aware digital twin combines real-world site and network data with TN and NTN propagation models, storing terrestrial received-power information in coverage map C and satellite interference coupling in map G.When measured rates deviate from map-based predictions, relevant entries are refined using feedback rather than regenerating the full map at every scheduling interval.Source: A. Modeling of the Context-Aware… · pp. 4–5

B. TN Antenna Modelling for Coexistence Analysis

  • Terrestrial beams use a finite DFT codebook on a half-wavelength-spaced URA, with element gain and array response combined to compute directional beam gain.The element pattern uses maximum gain, horizontal and vertical half-power beamwidths, and maximum attenuation.Source: B. TN Antenna Modelling for… · p. 5

C. NTN Modelling for Coexistence Analysis

The NTN model combines satellite receive-antenna directionality with time-varying BS–satellite geometry and propagation-channel modeling. These inputs characterize the interference coupling of terrestrial beams to visible satellite receivers.

1) NTN Antenna Modelling:

  • The satellite receive antenna is modeled with a circular-aperture reflector pattern normalized to unit power on boresight, and its directional gain-to-noise-temperature ratio depends on the off-axis angle.The model specifies the aperture radius and wave number and assumes receiver noise figure and antenna noise temperature do not depend on arrival direction.Source: 1) NTN Antenna Modelling · p. 5

2) NTN Channel Model:

  • OpenNTN generates offline BS–satellite channels conditioned on link geometry and deployment scenario, including LOS/NLOS state and large- and small-scale channel parameters.The model accounts for path loss, shadow fading, clutter, atmospheric attenuation, and stochastic multipath realizations; building-entry loss is omitted for outdoor terrestrial base stations.Source: 2) NTN Channel Model · pp. 5–6
  • The channel model uses time-varying satellite positions and BS–satellite geometry, including slant distance, elevation, and panel-relative angles, to determine directional transmit gain and link conditions.Satellite mobility is represented through changing link geometry and resulting Doppler characteristics.Source: 2) NTN Channel Model · p. 6, Figure 2 · p. 6
  • The NTN coupling retained for interference evaluation is derived from effective power coupling between each terrestrial beam and satellite across channel realizations.The passage defines this retained coupling but does not provide a numerical result.Source: 2) NTN Channel Model · p. 6

D. Construction of Multi-Antenna CKM

DT-STABLE uses a site-specific channel knowledge map (CKM) to estimate terrestrial link quality by location and beam, alongside an interference map that predicts exposure at visible satellites. Together, these maps give near-real-time resource allocation information without requiring exhaustive measurements.

D. Construction of Multi-Antenna CKM

  • Each O-RU/gNB uses a finite DFT beamforming codebook, with the full codebook containing one beam per antenna element.The precoding vectors are formed from horizontal and vertical DFT steering vectors.Source: D. Construction of Multi-Antenna CKM · p. 6
  • Site-specific ray tracing generates offline coverage maps for candidate beams, whose average beamformed path gains support estimates of terrestrial service utility.The CKM can be used to evaluate received SNR and achievable spectral efficiency for candidate O-RU–beam–UE associations.Source: D. Construction of Multi-Antenna CKM · pp. 6–7
  • The CKM stores beam-dependent received signal strength by coverage cell and uses a UE’s mapped cell to approximate received power, with accuracy depending on cell size and within-cell variation.The stored values support estimates of received SNR and achievable spectral efficiency for candidate associations.Source: D. Construction of Multi-Antenna CKM · p. 7
  • The environment-aware multi-beam CKM incorporates site-specific blockage, terrain, and multipath, unlike location-only or idealized uniform-coverage models.It provides spatially consistent channel gain for each beam within the O-RU coverage.Source: D. Construction of Multi-Antenna CKM · p. 7

E. Interference Map Construction

  • The interference map combines satellite geometry, directional antenna responses, and NTN propagation to predict time-varying exposure from candidate BS–beam pairs along satellite trajectories.It lets near-real-time resource allocation account for potential violations of the satellite protection threshold.Source: E. Interference Map Construction · p. 7

V. RADIO RESOURCE MANAGEMENT FOR HARMONIOUS NTN–TN COEXISTENCE

The resource-allocation formulation jointly chooses active beams and UE associations to balance terrestrial utility, satellite protection, and efficient use of beam resources. It is NP-hard, motivating an efficient suboptimal solution.

A. Optimization Problem Formulation

  • The binary beam-activation and user-association decisions require each UE to use one active BS–beam pair while respecting each BS’s RF-chain limit.Beam load is used in the load-aware utility.Source: A. Optimization Problem Formulation · p. 8
  • The formulation penalizes aggregate INR above each satellite’s threshold, with the interference price controlling the trade-off against terrestrial network utility.No penalty applies below the threshold; violations are increasingly penalized as the price increases.Source: A. Optimization Problem Formulation · p. 8
  • Problem P0 is NP-hard, established by reducing 0–1 knapsack to a restricted instance of the formulation.The construction makes any capacity violation suboptimal, leaving the knapsack objective over selected item beams.Source: A. Optimization Problem Formulation · pp. 8–9

B. DT-STABLE Algorithm Development

DT-STABLE uses a price-based iterative heuristic to coordinate beam activation and UE association under interference and load constraints. It is computationally efficient, but global optimality is not guaranteed.

B. DT-STABLE Algorithm Development

  • The algorithm iteratively adjusts interference and load prices to guide beam selection and UE association, then stops when price changes meet a tolerance or the iteration limit is reached.It first forms provisional associations, activates up to the RF-chain limit, and reassigns UEs from deactivated beams.Source: B. DT-STABLE Algorithm Development · p. 9
  • The algorithm’s complexity over at most Tmax iterations is O(Tmax(UJK + JKM + JK log K)).The per-iteration terms account for association and load evaluation, interference-aware scoring, and beam selection.Source: B. DT-STABLE Algorithm Development · p. 9
  • Because P0 is a nonconvex mixed-integer problem, DT-STABLE is a heuristic and does not guarantee global optimality.It adaptively balances terrestrial capacity and satellite protection, especially when protecting additional satellite links would require substantial utility sacrifice.Source: B. DT-STABLE Algorithm Development · p. 9

C. O-RAN Compliant Multi-Time Scale Optimization

DT-STABLE separates long-term digital-twin map maintenance from near-real-time feedback and scheduling. It selectively probes when measurements diverge from predictions, then uses the refined inputs for beam and user decisions.

C. O-RAN Compliant Multi-Time Scale Optimization

  • The non-RT RIC updates terrestrial coverage and satellite-interference maps from site context and satellite visibility, then sends maps and feasible beam sets to the near-RT RIC via A1.The near-RT RIC combines DT predictions with UE locations, serving-cell information, and measured rate KPMs received through E2.Source: C. O-RAN Compliant Multi-Time Scale… · pp. 9–10
  • When a UE’s measured serving rate exceeds the DT-prediction mismatch threshold, the xApp selects only the top-Kp predicted BS–beam pairs for selective probing.Measured received powers replace the corresponding DT estimates for that scheduling instance, and achievable rates are recomputed.Source: C. O-RAN Compliant Multi-Time Scale… · p. 10
  • DT-guided ranking avoids exhaustive beam sweeping, reusing the map when measured rates agree with predictions and probing only UEs with detected mismatch.For mismatched UEs, only selected candidate BS–beam pairs are measured.Source: C. O-RAN Compliant Multi-Time Scale… · pp. 9–10
  • The near-RT xApp runs Algorithm 1 using refined rates, interference couplings, feasible beam sets, INR thresholds, and the RF-chain limit, then sends beam and association decisions to base stations via E2.DT-STABLE performs short-term scheduling after DT-map updates and selective probing.Source: C. O-RAN Compliant Multi-Time Scale… · pp. 10–11

A. Simulation Setting

The evaluation uses a site-specific digital twin and physical-network simulation for terrestrial downlink coexistence with a LEO satellite system. Satellite visibility and interference are assessed under a conservative low-elevation geometry.

A. Simulation Setting

  • The simulated site covered 4×4 km2 near Québec City and included 19 macro O-RUs/MBSs, each with a 25 m height, a 1000 m service region, and 16 candidate beams.Each O-RU used a half-wavelength-spaced directional 4×4 URA and a finite DFT codebook.Source: A. Simulation Setting · p. 11
  • Satellite positions came from CelesTrak TLE data, and visibility was determined using a 25° minimum elevation threshold.The CDT and physical-network simulation were synchronized, with site-specific radio information informing decisions.Source: A. Simulation Setting · p. 11
  • Low-elevation satellites were the critical regime: they were more prevalent and less protected by downtilt, leaving p95 INR above the −6 dB threshold in the 25°–45° range.Downtilt reduced p95 INR by up to 5.19 dB above 70° and by 3.09 dB at 25°–45°; subsequent analysis used the lowest-elevation visible satellites at each observation.Source: A. Simulation Setting · p. 11

B. Optimization Strategies

The evaluation contrasts interference-optimal, rate-optimal, and balanced operating points, then compares DT-STABLE with terrestrial and interference-aware benchmarks. DT-STABLE provides a middle ground in satellite protection and terrestrial throughput across the reported settings.

B. Optimization Strategies

  • Adaptive interference-price parameters produce three operating points: interference-optimal prioritizes satellite protection, while rate-optimal prioritizes terrestrial throughput.The 5G baseline exhaustively sweeps candidate beams without considering satellite interference; PBS selects beams locally and does not coordinate aggregate interference across base stations.Source: B. Optimization Strategies · p. 11, C. Benchmark Schemes · p. 12
  • 99.99% INR satisfaction with 133.55 bps/Hz characterized the interference-optimal setting, while the rate-optimal setting achieved 257.24 bps/Hz and 66.99% INR satisfaction; DT-STABLE achieved 89.63% and 210.96 bps/Hz.DT-STABLE improved INR satisfaction by 22.64 percentage points over the rate-optimal setting.Source: B. Optimization Strategies · p. 11
  • DT-STABLE reached 210.96 bps/Hz, outperforming NL-Moderate/Strict and PBS-Medium/High, while achieving INR control comparable to NL schemes with 89.63% satisfaction.The 5G baseline had a mean INR of −6.53 dB and 55.32% satisfaction; DT-STABLE reduced mean INR to −13.07 dB.Source: C. Benchmark Schemes · p. 12
  • DT-STABLE attained the highest normalized ideal-point score, Sa = 0.813, among the evaluated benchmarks.The score measures proximity to the normalized ideal point, with larger values indicating closer proximity.Source: C. Benchmark Schemes · p. 12
  • At U = 100, DT-STABLE achieved 89.63% INR satisfaction with a 25.99% rate loss relative to the 5G baseline, and its INR satisfaction exceeded PBS-Low across evaluated user densities.As users increased from 50 to 300, DT-STABLE’s INR satisfaction decreased from 91.09% to 87.56%.Source: C. Benchmark Schemes · pp. 12–13
  • At U = 100, DT-STABLE achieved 68.67% QoS satisfaction; across U = 50 to U = 300, QoS satisfaction decreased from 73.91% to 49.73%.QoS satisfaction counted UEs meeting the minimum spectral efficiency of 0.5 bit/s/Hz.Source: C. Benchmark Schemes · p. 13

D. Advantages of Digital Twin Predicted Channel

Site-specific DT channel predictions outperform the 3GPP UMa baseline, while feedback-driven CKM refinement helps DT-STABLE adapt to channel mismatch with lower beam-measurement overhead. The results show gains under blockage and preserve much of exhaustive-probing rate with DT-guided candidate beams.

D. Advantages of Digital Twin Predicted Channel

  • DT-predicted channels outperformed the 3GPP UMa baseline, with average sum rate rising from 146.85 to 315.64 bps/Hz as users increased from 50 to 300, versus 25.49 to 43.45 bps/Hz.The DT model used site-specific ray tracing in a reconstructed 3D environment.Source: D. Advantages of Digital Twin… · p. 13

E. Advantages of Multi-time Scale Digital Twin Assisted Resource Management

  • When prediction error exceeded a predefined mismatch threshold, DT-STABLE selectively probed DT-ranked candidate beams and updated CKM information before resource allocation.The controlled mismatch came from temporary mixed blockage affecting 30% of UEs during slots 60–90.Source: E. Advantages of Multi-time Scale… · p. 13
  • 14.8% higher average sum rate under blockage: DT-STABLE reached 197.91 bits/s/Hz versus 172.40 bits/s/Hz with a static DT.The experiment used 50 UEs, 19 BSs, 16 candidate beams per BS, and feedback every 10 slots.Source: E. Advantages of Multi-time Scale… · p. 13
  • 99.71% fewer beam measurements than full sweeping: DT-STABLE used 535 measurements for 107 mismatch events with Kp = 5, versus 182,400 for full sweeping.The DT-guided candidate set at Kp = 8 achieved 91.9% of the full-probing rate on average; Table IV reports recovered rate versus beam-sweeping budget.Source: E. Advantages of Multi-time Scale… · pp. 13–14, Table IV · p. 13, Figure 12 · p. 13

F. Robustness Evaluation

DT-STABLE's robustness tests show sensitivity to positioning and interference-map errors, while satellite protection varies with elevation. Its conclusion reports a rate–protection balance and faster execution than exact optimization.

F. Robustness Evaluation

  • The robustness evaluation examined UE positioning errors, interference-map uncertainty, and satellite elevation angle as sources of mismatch or geometry variation.The scheduler's decisions under mismatched DT information were validated over the real-world layer.Source: F. Robustness Evaluation · p. 14

1) Impact of Imperfect UE Positioning Estimation:

  • At 10% UE positioning error, mean total sum rate fell to 64.06 bps/Hz and QoS satisfaction to 68.6%, declines of 56.5% and 19.8 percentage points from accurate positioning.The tested values with accurate positioning were 147.33 bps/Hz and 88.4%.Source: 1) Impact of Imperfect UE Positioning… · p. 14

2) Impact of Interference Prediction Uncertainty:

  • As interference-map error increased from 0 to 5 dB, INR satisfaction fell from 89.66% to 79.15%, while mean INR rose from −13.01 dB to −10.88 dB.The terrestrial channel rates were held unchanged to isolate interference-map uncertainty.Source: 2) Impact of Interference Prediction… · p. 14

3) Impact of Elevation Angle:

  • After scheduling, mean worst-case INR exceedance fell by 75.0%, 70.6%, and 75.0% for low-, mid-, and high-elevation profiles, respectively.Post-scheduling INR satisfaction was 89.63%, 92.47%, and 98.50% across those profiles.Source: 3) Impact of Elevation Angle · pp. 14–15

G. Computational Complexity and Gurobi Benchmarking

  • 0.207–0.757 s for DT-STABLE versus 3.26–21.21 s for Gurobi on reduced instances, with INR satisfaction up to 30 percentage points lower for DT-STABLE.DT-STABLE runtime remained below 0.45 s across tested UE populations, including 150 UEs; Table V compares runtime, sum rate per BS, and INR satisfaction.Source: Table V · p. 15, G. Computational Complexity and Gurobi… · p. 15

VII. CONCLUSION

  • For 100 users, DT-STABLE achieved 89.63% INR satisfaction and 210.96 bps/Hz mean sum rate, with nearly complete INR satisfaction when satellite protection was prioritized; execution was 15.5x to 33.1x faster than exact optimization.Source: §VII · p. 15

APPENDIX A DERIVATION OF DT-STABLE

The lower-level allocation uses load prices to guide association and beam selection under resource constraints. The upper level adapts satellite-interference prices from residuals, yielding a tractable but not globally certified solution.

A. Lower-Level Resource Allocation

  • Load prices shape association and beam scores: associations are selected at simplex vertices, then beams are filtered to the RF-chain limit and users on discarded beams are reassigned.The relaxed lower-level program is convex, but the sequential restoration procedure does not guarantee global optimality.Source: A. Lower-Level Resource Allocation · p. 15

B. Upper-Level Adaptive Interference Pricing

  • The upper level adjusts each satellite's interference price upward for positive residuals and downward for negative residuals, projecting prices into [0, λmax,m] until convergence or Tmax.The resulting solution is computationally tractable but not certifiably globally optimal.Source: B. Upper-Level Adaptive Interference… · p. 16

Related papers

Loading 2610.32827v1…