Non-Terrestrial Networks Overview

Updated 28 January 2026

Non-Terrestrial Networks are integrated space and airborne communication infrastructures combining GEO, MEO, LEO, HAPs, and UAVs to enhance global wireless coverage.
They leverage both transparent and regenerative payload models with 3GPP NR NTN standards to support multi-connectivity, dynamic resource allocation, and low-latency links.
NTNs boost 5G/6G architectures by improving throughput, reducing coverage gaps, and enabling broadband, IoT, and mission-critical applications in remote areas.

Non-Terrestrial Networks (NTNs) refer to integrated space- and airborne communication infrastructures—primarily satellite constellations (LEO, MEO, GEO), high-altitude platforms (HAPs), and unmanned aerial vehicles (UAVs)—offering radio access and backhaul beyond the traditional terrestrial network, extending wireless coverage, capacity, and resilience worldwide. NTNs are a cornerstone of 5G/6G architecture, providing direct-to-device, relay, and multi-layer connectivity for broadband, IoT/mMTC, mission-critical, and disaster scenarios, especially in rural, remote, and underserved regions (Hernandez et al., 2023, Wang et al., 2024, Majamaa, 2023, Rossato et al., 21 Jan 2026, Araniti et al., 2021).

1. Taxonomy and System Architectures

Platform Types and Orbits

Non-terrestrial platforms span:

Geostationary Orbit (GEO): ≈35,786 km altitude, fixed with respect to Earth, round-trip delay ≈500 ms, very large footprint (radius ≈2,500–3,500 km), negligible Doppler (Hernandez et al., 2023, Araniti et al., 2021).
Medium Earth Orbit (MEO): 2,000–35,000 km, intermediate delay (RTT ≈ 70–150 ms), moderate Doppler, regional coverage (Hernandez et al., 2023, Araniti et al., 2021).
Low Earth Orbit (LEO): 300–2,000 km, RTT ≈20–50 ms, high mobility (orbital speed ≈7.5 km/s), significant Doppler (tens of kHz at GHz bands), small and rapidly moving beams (Hernandez et al., 2023, Wang et al., 2024).
High-Altitude Platforms (HAPs): ≈20 km altitude, RTT ≈10 ms, stationary/controlled mobility, footprint ≈100–200 km (Hernandez et al., 2023).
UAVs: 0.1–5 km, low latency, limited coverage, extreme flexibility (Hernandez et al., 2023).

Architecture integrates these platforms with terrestrial base stations (TNs) via unified core networks (3GPP NG-Core/5GC/6GC), supporting hybrid access modes (direct, relay, multi-connectivity) and backhaul via feeder/gateway stations (Rossato et al., 21 Jan 2026, Hernandez et al., 2023).

Payload Models and Integration

Transparent ("bent-pipe") payload: Simple frequency translation/amplification; all routing at ground stations. Standard in earlier GEO/MEO/LEO deployments (Hernandez et al., 2023, Araniti et al., 2021).
Regenerative payload: On-board demodulation, routing/switching, ISL support; allows flexible routing, traffic isolation, and lower end-to-end latency (Hernandez et al., 2023, Rossato et al., 21 Jan 2026).
3GPP NR NTN-compliant: Both direct access and relay are implemented; hybrid functional splits (e.g., DU on satellite, CU on ground) extend terrestrial RAN with satellite/air segment (Rossato et al., 21 Jan 2026, Araniti et al., 2021).

NTNs form the upper layers of the "Space–Air–Ground Integrated Network" (SAGIN), where all segments share resources and control (Nguyen et al., 2024).

2. Physical Layer and Propagation: Key Characteristics

Channel Impairments

Free-space path loss (FSPL): For distance $d$ (m) and carrier frequency $f$ (Hz),

$L_{\mathrm{fs}} = 20\log_{10}(4\pi d f/c)$

Large-scale, elevation- and frequency-dependent (Hernandez et al., 2023, Rossato et al., 21 Jan 2026).

Doppler shift: LEO satellites at 2 GHz can induce $f_d ≈ 50$ kHz, requiring Doppler estimation and compensation, particularly for OFDM (Majamaa, 2023, Rossato et al., 21 Jan 2026).
Propagation delay: $τ = d/c$ . GEO: $τ ≈ 120$ ms one-way; LEO: $τ ≈ 2–4$ ms (Majamaa, 2023).
Beam dynamics: Fast-moving spot beams in LEO require frequent handovers (every few minutes) and fine-grained timing/synchronization.

Channel Modeling

Large-scale fading: Dominant LoS in HAPs/LEO/MEO, Ricean or Shadowed-Rician fading (Huang et al., 2023).
Coherence time: Rapid satellite movement (LEO) yields short coherence time ( $T_c \sim 0.1–10$ μs at mmWave); LOS fraction crucial (Zheng et al., 2024).
Atmospheric impairments: Rain fade, gaseous absorption, and scintillation are significant at mmWave/FSO; ITU models standard (Hernandez et al., 2023, Rossato et al., 21 Jan 2026).
Beamforming: Active phased arrays on satellites/HAPs enable flexible spot-beam formation and tracking (Wang et al., 2024).

3. Radio Resource, Access, and Network Slicing

Multi-Connectivity and Scheduling

Multi-connectivity (MC): PDCP-layer MC (MR-DC) allows UE to connect to multiple nodes (satellite & terrestrial BSs, or multiple satellites). MC in NTN faces new challenges: delay asymmetry, Doppler, out-of-order delivery, frequent beam/sat handovers, and buffer sizing (Majamaa, 2023).
Resource allocation: Critical parameters include dynamic bandwidth fission ( $\varepsilon$ ), power adaptation, BS activation, and association, with optimization frameworks (BLASTER, log-sum utility, proportional-fair) balancing TN/NTN capacity, energy, and fairness (Alam et al., 2024, Alam et al., 2023, Trankatwar et al., 27 Jan 2026).
Network slicing: SDN/NFV-enabled NTN virtualizes resources into eMBB, URLLC, mMTC slices. Slices are isolated at radio, transport, and core. End-to-end constraints include latency, reliability, rate, and power (Nguyen et al., 2024, Wang et al., 2024).

MAC and HARQ Adaptations

Random Access (RA): Window and timer extensions (add 2× propagation delay). Pre-compensation needed due to timing uncertainty (Rossato et al., 21 Jan 2026).
HARQ: Large RTT requires increasing number of HARQ processes (e.g., 32+ for LEO), or switching to RLC-ARQ as fallback, with throughput/latency trade-offs (Rossato et al., 21 Jan 2026, Traspadini et al., 2024).

Duplexing and TDD

TDD in NTN: Guard period set by maximum cell one-way delay (e.g., LEO at 800 km: >5 ms GP). Enhanced slot allocation (ESSA) methods fill idle guard periods for higher channel utilization (Traspadini et al., 2024).

4. Mobility Management, Handover, and Orchestration

Mobility: LEO/MEO satellites produce frequent (minutes-scale) handovers. Optimized algorithms use predictive handover via GNSS/ephemeris, RA-less SN addition, elevation/location-based candidate selection, connected-mode pre-establishment (Majamaa, 2023, Wang et al., 2024).
Handover decision metrics: Composite utility functions ( $U = α \log(1+\mathrm{SINR}) − β \mathrm{latency} + γ T_\mathrm{stay}$ ) account for channel, latency, and dwell time in satellite cell (Wang et al., 2024).
Inter-satellite links (ISLs): Regenerative payloads and ISLs enable mesh networking, optimized routing (e.g., time-expanded graphs) for latency reduction (Wang et al., 2024).

Control and Artificial Intelligence

AI/ML for NTN: RL/DNN-based traffic steering, beam scheduling, dynamic slicing, and federated learning for control plane optimization: essential for complex, dynamic topologies and shadowing (Nguyen et al., 2024, Aygul et al., 2022, Al-Hraishawi et al., 2023).

5. System-Level Performance and Optimization

Analytical and Simulation Results

Throughput–coverage–fairness trade-offs: Integration of NTN boosts mean data rates by >200% in rural scenarios, reduces coverage holes by >90%, and enhances energy efficiency by up to 45% (Alam et al., 2023, Alam et al., 2024, Trankatwar et al., 27 Jan 2026).
Stochastic geometry models: System-level performance (coverage, capacity, association probabilities) is rigorously analyzed using spherical PPP, BPP, Cox process models, with closed-form metrics for coverage probability, SINR, and beam clustering (Wang et al., 13 Jan 2025, Huang et al., 2023).
Slot usage: ESSA in TDD can increase channel utilization by 3–10× over naïve schemes, especially with delay-based user grouping to minimize differential delays (Traspadini et al., 2024).
IoT over NTN: Hybrid offloading (probabilistic association to gateway vs. satellite) significantly improves success probability in mMTC (Wang et al., 2022).

Orbit	Band	Max Throughput	One-way Delay	Peak SNR	Area Capacity Density
LEO 600 km	Ka	302.7 Mbps	4–8 ms	8.5 dB	170 kbps/km²
GEO 35,786	Ka	469.9 Mbps	120 ms	11.6 dB	1.4 kbps/km²
LEO 600 km	S	76.4 Mbps	4–8 ms	6.6 dB	—
GEO 35,786	S	21.7 Mbps	120 ms	0 dB	—

Key observations: LEO offers order-of-magnitude latency advantage and higher area capacity density; GEO saturates at higher absolute throughput due to larger BW/EIRP but performance is challenged by path loss and delay (Figaro et al., 8 Jan 2026, Shang et al., 3 Nov 2025).

6. Open Challenges and Future Research Trajectories

HARQ/ARQ protocol design: Large RTT/propagation delays stretch conventional mechanisms. New schemes and extended timers (buffering, window, redundancy).
Reliance on GNSS: Satellite/UE synchronization and Doppler pre-compensation depend on GNSS, which is vulnerable to spoofing/jamming. GNSS-resilient operations and alternative timing protocols are required (Figaro et al., 8 Jan 2026).
Interference and coexistence: Multi-layer (GEO/LEO/HAP/UAV/TN) spectrum sharing, rate-splitting, and cognitive access, dynamic interference management across systems (Wang et al., 2024).
Security and privacy in network slicing: Lightweight isolation and authentication mechanisms per slice topologies, especially for dynamic SD-NTN (Wang et al., 2024).
Cross-layer and distributed AI: Federated/distributed ML for resource control, dynamic switching among centralized, decentralized, and federated learning according to topology, latency, privacy, and compute constraints (Aygul et al., 2022, Al-Hraishawi et al., 2023, Nguyen et al., 2024).
E2E orchestration: Digital-twin-based network control, on-board edge processing, O-RAN adoption for flexible NTN orchestration (Al-Hraishawi et al., 2023, Nguyen et al., 2024).
Next-gen physical-layer: Design of THz/optical ISLs, RIS/IRS for reconfigurable coverage, adaptive waveforms robust against ultra-high Doppler and delay (Araniti et al., 2021, Shang et al., 3 Nov 2025).

7. Standardization and Practical Deployment

3GPP NR NTN (Rel-17–Rel-20): Ongoing inclusion of satellite direct access, regenerative payloads, enhanced frequency bands (L/S/Ka), GNSS-resiliency, advanced mobility & handover, RedCap/IoT extensions (Rossato et al., 21 Jan 2026, Figaro et al., 8 Jan 2026, Majamaa, 2023).
Interoperability and scaling: Focus on practical issues (gateway placement, beam/user density, feeder/gateway link variability, dual connectivity form-factors, uplink power constraints) and the massive scaling expected in NTNs (e.g., >10,000 LEO satellites) (Hernandez et al., 2023, Rossato et al., 21 Jan 2026).
Industry/field validation: LEO and GEO constellations (Starlink, OneWeb, HughesNet) have demonstrated >100 Mbps/user, LEO round-trip delays <40 ms, with full 3GPP NR integration and open-source evaluation suites (e.g., ns3-NTN) available for E2E validation (Figaro et al., 8 Jan 2026, Azari et al., 2021).

NTNs, as a fundamental 6G pillar, are rapidly advancing toward seamless, software-defined, AI-orchestrated, energy-efficient, and globally available space–air–ground communication fabrics (Wang et al., 2024, Nguyen et al., 2024, Majamaa, 2023, Rossato et al., 21 Jan 2026, Wang et al., 13 Jan 2025).