Non-Terrestrial Network Operation

• Non-Terrestrial Network (NTN) operation is a layered architecture combining satellites, HAPS/UAVs, and terrestrial systems to provide seamless, ubiquitous connectivity.
• It addresses unique challenges such as high Doppler shifts, significant propagation delays, and dynamic beam and resource management across space-air-ground links.
• Integration with terrestrial networks and use of AI/ML-driven orchestration enable real-time scheduling, adaptive resource optimization, and enhanced network slicing.

A Non-Terrestrial Network (NTN) is a network architecture that integrates spaceborne (satellites in LEO/MEO/GEO), airborne (HAPS, UAVs), and terrestrial segments to provide ubiquitous wireless connectivity. NTN operation spans a diverse set of networking, signal processing, and resource management challenges unique to high-mobility, large-delay, and multi-domain environments. This article presents a comprehensive synthesis of NTN operation from architectural principles through physical layer specifics, resource management, integration with terrestrial systems, and emergent orchestration and optimization methodologies.

1. NTN System Architecture and Operational Modes

NTNs are multi-layered systems comprising satellites (LEO, MEO, GEO), HAPS, UAVs, and ground segments including gateways and terrestrial infrastructure. Operational modes are categorized as "bent-pipe" (transparent) or "regenerative":

• Bent-pipe (transparent) mode: The satellite or HAPS payload performs only RF frequency translation and amplification. All protocol processing, including NR stack and user or control-plane processing, is ground-based. This is the dominant mode in commercial LEO mega-constellations and is the basis for testbeds such as SpaceNET (Rahman et al., 23 Dec 2025) and 6GStarLab (Ruiz-de-Azua et al., 19 Mar 2025).
• Regenerative mode: The payload includes onboard SDRs and baseband/PHY/MAC computation (e.g., gNB/UE stack in SDR CPU/FPGA), performing demodulation, decoding, scheduling, and protocol-leading functions onboard (Ruiz-de-Azua et al., 19 Mar 2025). Onboard resources are dynamically allocated, supporting experiment-driven deployments under 3GPP Release 17/18 NTN requirements.

Typical NTN operation involves periodic satellite passes (LEO/MEO), with dynamic beam footprints, adaptive tracking and handover, and integration with inter-satellite links (ISL) for mesh routing and data aggregation (Wang et al., 2024, Baena et al., 21 Feb 2025).

Space-Ground Segmentation

• Space segment: Satellites host SDRs, multiband RF chains, and, increasingly, optical laser terminals (e.g., 6GStarLab: UHF, S, X, Ka bands; bidirectional Gb/s optical terminal).
• Ground segment: RF and optical ground stations, user terminals (UEs, IoT sensors, VSATs), control and operations centers (Ruiz-de-Azua et al., 19 Mar 2025).
• Airborne segment: HAPS or UAVs supplement satellite coverage and can be equipped with advanced beamforming (including RIS (Tekbıyık et al., 2020)).

The network can be dynamically scheduled (experiment time slots, power allocation), with on-the-fly reconfiguration of frequency, beam pattern, waveform, and processing mode.

2. Physical Layer: Channel Modeling, Link Budget, and Synchronization

Channel and Propagation

NTN links are dominated by free-space path loss, atmospheric attenuation (especially in Ka-band and above), and Doppler effects due to satellite velocity (Sandri et al., 2023, Araniti et al., 2021). The canonical link budget model accounts for:

$\text{FSPL}(f, R) = 20\log_{10}(4\pi R f / c)$

$$P_r = P_t + G_t + G_r - \text{FSPL}(f, R) - L_{atm} - L_{scint} - L_{misc}$$

Where additional loss terms include oxygen/rain attenuation, environmental scintillation, shadowing/clutter (per 3GPP TR 38.811/821), and possible pointing or polarization errors (Rahman et al., 23 Dec 2025, Sandri et al., 2023).

High-precision satellite mobility and geolocation modeling is enabled by Earth-Centered Earth-Fixed (ECEF) coordinates with accurate orbital propagators (Sandri et al., 2023). Signal propagation is simulated per scenario, with elevation angle and frequency dictating loss tables.

Synchronization and Doppler

• LEO velocity (~7.5 km/s) generates significant Doppler shifts: $f_D = (v_{rel} / c) \cdot f_c$, necessitating Doppler pre-compensation at both the satellite and UE level, frequent numerology selection ($\mu$ in NR), and/or subcarrier spacing adaptation (Araniti et al., 2021).
• Timing Advance (TA) and guard periods: Large slant ranges (up to 36 000 km in GEO) yield propagation delays of $\sim$2–600 ms, with extended TA fields and differentiated PRACH formats; TDD operation requires enhanced synchronized slot allocation (ESSA) to exploit guard intervals (Traspadini et al., 2024).
• Random Access and Handover: Extended RACH/PRACH preamble, increased contention windows, HARQ timer adaptation, and conditional handover procedures are implemented (Araniti et al., 2021).

3. Resource Management, Scheduling, and Optimization

NTN resource management must address spectral efficiency, fairness, and real-time adaptation under high-latency, heterogeneous connectivity, and cross-tier constraints:

Power and Bandwidth Allocation

For GEO/LEO spot beams, power allocation among UEs is formalized as a non-convex fractional programming problem, with coupled intra-beam interference (Rahman et al., 2023). The alternate fractional programming (Alt-FP) algorithm achieves 10–15% spectral efficiency gains over classical FP, supporting per-beam NP-hard sum-log-of-fractions problems via Lagrangian and quadratic transforms with per-iteration complexity $O(I\cdot J)$.

In integrated TN–NTN, joint allocation of spectrum split $\varepsilon$, UE association, and per-BS (ground or satellite) transmit powers is solved as a mixed-integer, log-utility maximization (Alam et al., 2024, Alam et al., 2023). Dynamic algorithms (e.g., BLASTER, BCOMD) enable real-time adjustment, yielding up to 250% throughput and 45% energy savings relative to static 3GPP benchmarks (Alam et al., 2024, Alam et al., 10 Jun 2025).

Scheduling and Network Slicing

• Mini-slot scheduling and flexible numerology: Used in LEO/MEO to address short dwell times, fast handover, and URLLC latency (Araniti et al., 2021).
• Network slicing and closed-loop orchestration: Multi-layer per-slice SLAs are enforced by dynamic resource allocation (see Section 4); AI/ML-based RICs schedule beam/UE/spectrum assignments (Nguyen et al., 2024, Baena et al., 21 Feb 2025).
• Fairness-throughput trade-off: Multi-objective optimization (FTA-NTN) with Bayesian parameter tuning yields globally optimal LEO/MEO constellation designs supporting both high throughput (e.g., $>9$ Gb/s) and fairness (Jain index $\sim$0.4 for 500 users), with adaptive user clustering and beam allocation (Trankatwar et al., 27 Jan 2026).

Slot Filling and Time Division Duplexing in NTN

• The ESSA method enables the fill of large TDD guard intervals with downlink (DL) transmissions, leveraging differential delay spread and tailored scheduling (e.g., SNR- or delay-spread-based UE selection). Simulation results show up to $4\times$ improvement in capacity (Traspadini et al., 2024).

4. Integration with Terrestrial Networks, Orchestration, and Standardization

TN–NTN Interworking

NTNs are integrated with terrestrial (5G/6G) networks via standard interfaces and co-optimized BBU placement, spectrum assignment, and bandwidth partitioning (Alam et al., 2024, Alam et al., 2023, Alam et al., 10 Jun 2025). The NR stack is extended with minimal protocol changes to support both bent-pipe and regenerative payloads, enabling direct-to-satellite NB-IoT, 5G NR (Rel-17: n256/n511), and LEO/GEO access types (Lin et al., 2021, Ruiz-de-Azua et al., 19 Mar 2025).

• Operational scenarios: Real deployments dynamically adjust satellite beams for rural/underserved area coverage, offload terrestrial BSs in low-traffic periods (shutdown for energy savings), and shift bandwidth $$P_r = P_t + G_t + G_r - \text{FSPL}(f, R) - L_{atm} - L_{scint} - L_{misc}$$0 adaptively according to UE association distributions (Alam et al., 2024, Alam et al., 2024).
• User association & energy: Joint association schemes balance terrestrial/NTN loads, incorporating pricing-based and probabilistic MAB (multi-armed bandit) control to achieve low outage and power consumption across dynamic traffic profiles (Alam et al., 10 Jun 2025, Alam et al., 2024).

Orchestration and Policy-Driven Control

• Hierarchical architectures: Multi-layered orchestration frameworks such as Space-O-RAN and the weak-control policy-driven multi-operator orchestration allow per-operator autonomy while guaranteeing end-to-end policy compliance, route negotiation, and dynamic re-orchestration under link failures or topology change (Abe et al., 5 Feb 2026, Baena et al., 21 Feb 2025).
• AI/ML-driven RIC and digital twins: Satellite-side dApps (onboard applications), near-RT Space-RICs (clusters using ISL <10 ms), and strategic SMO (ground-based) relax global and operator-specific constraints, drive digital-twin simulation, and periodically update low-latency and non-RT policies as required (Baena et al., 21 Feb 2025).
• Industry-standard interfaces: O-RAN functional splits (e.g., split 2 or 7.2) mapped to satellite radio links; open interfaces (E2, A1, O1) dynamically mapped to SL, ISL, FL, or GSL per link property (Nguyen et al., 2024, Baena et al., 21 Feb 2025).

5. Practical Testbeds, Standardization Platforms, and Experimentation Workflows

• 6GStarLab: 6U LEO CubeSat with SDR and optical payload, supporting real-time reconfiguration, experiment upload, bent-pipe/regenerative mode switching, and PHY/MAC standard conformance. Direct measurement of RACH, Doppler, handover, and waveform performance; empirical 3GPP feedback loop for channel model and KPI tuning (Ruiz-de-Azua et al., 19 Mar 2025).
• SpaceNET: End-to-end testbed integrating commercial Starlink LEO with Mininet-based IP-layer emulation, supporting transparent payload operation analysis, UDP/TCP performance benchmarking, cross-layer optimization trials, and dynamic routing/MAC experimentation live on commercial hardware (Rahman et al., 23 Dec 2025).
• Simulation/emulation: 3GPP-compliant channel and antenna models implemented in ns-3, including full support for path loss, fading, ECEF-based mobility, and hardware-conformant antenna radiation patterns (Sandri et al., 2023). Model calibration validated to <0.5 dB for FSPL, atmospheric, and scintillation components.

6. Research Challenges and Future Directions

Open issues in NTN operation remain at all layers:

• Inter-system interference and coexistence: Harmonization between LEO, GEO, HAP/UAV, and terrestrial networks, plus spectrum sharing strategies (Wang et al., 2024).
• Mobility management: Efficient handover (conditional/predictive/CHO), beam management, and load balancing to minimize handoff rate, interruption time, and signaling (Wang et al., 2024, Araniti et al., 2021).
• Network slicing and resource isolation: Dynamic req-provisioning for URLLC, eMBB, and mMTC slices under time-varying, cross-layer load and mobility. AI-based dynamic slice instantiation and re-optimization in software-defined NTNs (Nguyen et al., 2024, Wang et al., 2024).
• RIS and advanced beamforming: Integration of reconfigurable intelligent surfaces (RIS) at HAPS, LEO/MEO/GEO, and DSN nodes to mitigate free-space loss, misalignment fading, and scintillation, as well as support for low-SWaP relay and flexible waveform design (Tekbıyık et al., 2020).
• AI/ML frameworks and digital-twin orchestration: Advanced RL, CNN/LSTM/Transformer prediction for resource management, handover, and routing under partial CSIT, with federated learning and privacy-preserving designs across hybrid terrestrial-satellite environments (Wang et al., 2024, Nguyen et al., 2024).
• Security and reliability: Cross-operator policy-driven orchestration, blockchain-based slice security, and fault-tolerant slicing with on-demand VNF migration (Wang et al., 2024, Abe et al., 5 Feb 2026).

Large-scale, multi-layer NTN operation requires further research in scalable optimization, end-to-end protocol design under uncertain CSI and mobility, channel modeling for mmWave/optical NTN links, and standardization and open testbed development for real-world NTN deployments (Ruiz-de-Azua et al., 19 Mar 2025, Rahman et al., 23 Dec 2025).