Space-Terrestrial Integrated Network (STIN)

Updated 3 July 2026

Space-Terrestrial Integrated Network (STIN) integrates terrestrial infrastructures with LEO, MEO, and GEO satellites, and optional aerial platforms to provide seamless global connectivity.
It coordinates control, user-plane functions, and resource allocation across diverse network segments to address challenges such as rapid mobility and Doppler shifts.
Advanced STINs employ digital twins, multi-agent reinforcement learning, and dynamic spectrum-sharing protocols to optimize performance and support varied applications.

Space-Terrestrial Integrated Network (STIN) is a unified, multi-layer network paradigm that integrates terrestrial communication infrastructure (cellular base stations, edge servers, and gateways) with non-terrestrial assets such as Low Earth Orbit (LEO), Medium Earth Orbit (MEO), Geostationary Orbit (GEO) satellites and, in some configurations, aerial platforms (HAPs/UAVs), to provide seamless, resilient, and globally ubiquitous connectivity. By tightly coupling these heterogeneous segments, STINs coordinate control and user-plane functionality, joint resource allocation, and interference management across both terrestrial and space-based domains. Advanced STIN architectures also leverage digital twins, quantum-assisted design, and spectrum-sharing protocols enabling 3D coverage, dynamic service provisioning, and robust adaptation to environmental and infrastructural diversity.

1. Fundamental Architecture and Elements

The canonical system structure of STINs delineates three main segments (Voicu et al., 2021, Hernandez et al., 2023, Tao et al., 2024):

Space segment: Encompasses NGSO (LEO/MEO) and GEO constellations, supporting inter-satellite links (ISLs), feeder links to terrestrial gateways, and, in advanced systems, on-board processing, beamforming, and satellite as-a-service (e.g. satellite gNB for 5G NR-Uu). Private constellations such as OneWeb, Amazon Kuiper, and Starlink embody Walker-type orbital designs with hundreds to thousands of satellites, specifying orbital parameters (altitude, inclination), ISL architecture (laser or Ka-band), and dynamic spot beam coverage.
Aerial segment (optional): HAPs/UAVs may supplement coverage, enable edge computing, or facilitate relay for shadowed terrestrial regions (Wang et al., 2024).
Terrestrial segment: Comprises macro/micro base stations, edge/cloud-edge servers, terrestrial access points, and gateways interfaced with satellites via high-capacity RF or optical feeder links. Network Operations Centers (NOCs), virtualized RAN cores, and digital twins can provide further orchestration.

The STIN control plane permits end-to-end session establishment, mobility and handover management, and cross-domain resource control, with the user plane carrying IP-encapsulated traffic over a unified protocol stack. Mapping the 5G NR stack to satellite links necessitates adaptation of numerology (to address Doppler shifts, large RTT), and protocol timers (RLC/HARQ), along with cross-layer APIs for beam handover prediction (Voicu et al., 2021).

2. Key Technical Challenges in STIN Design

2.1 Mobility and Handover

Rapid satellite mobility, especially in large LEO constellations, introduces challenges:

Beam footprints scan the Earth surface at velocities (up to 7 km/s), resulting in short visibility windows and frequent handovers (e.g., 70–110 handovers/hour for mega-LEO constellations, compared to <10 for MEO/HEO) (Voicu et al., 2021).
Two key handover strategies are recognized:
- “Best-signal” (switch to maximum elevation)
- “Longest-visibility” (hold beam as long as possible)
An explicit trade-off exists between handover rate, latency, and control-plane overhead, with LEO systems favoring “longest-visibility” to suppress excess signaling and MEO/HEO preferring “best-signal” for minimal path-loss (Voicu et al., 2021).

2.2 Doppler Shift and Propagation Delay

LEO satellites at f_c=30 GHz can incur Doppler shifts >100 kHz, requiring adaptive OFDM subcarrier spacings and on-board GNSS pre-compensation (Voicu et al., 2021, Hernandez et al., 2023).
End-to-end Round Trip Time (RTT) and protocol timer adaptation is paramount: GEO RTT can reach ~550 ms, MEO 100–250 ms, and LEO 20–50 ms (Hernandez et al., 2023, Voicu et al., 2021).

2.3 Interference and Integrated Resource Management

Dense satellite deployments cause intra-system (co-channel) and inter-system (e.g., LEO→GEO GS) interference, demanding coordinated spectrum allocation (beam management, subchannel matching) and explicit constraints on interference to guarantee QoS (Zhang et al., 2023).
Technically, matching-based and water-filling power allocation algorithms, as well as Lagrangian decomposition for joint handover/resource association, are essential for optimal throughput/backhaul trade-offs.

2.4 Dynamic Resource Allocation and Digital Twins

STIN must address user mobility and channel variation across both terrestrial and satellite domains. Digital twin (DT) frameworks provide predictive control, enabling real-time mirroring of physical topology and enabling proactive traffic steering, resource reservation, congestion minimization (Tao et al., 2024, Nguyen-Kha et al., 28 Jul 2025, Nguyen-Kha et al., 9 Feb 2026).
Multi-agent reinforcement learning (MARL) schemes and two-stage optimization (coarse-grained with DT, fine-grained in real-time) are prevalent for scalable, self-organizing deployment and management (Tao et al., 2024).

3. Protocol, Spectrum, and Physical Layer Aspects

Layer	Terrestrial	Satellite¹	Unified Adaptation
PHY, MAC	5G-NR, OFDM	OFDM (adaptive numerology)	Doppler compensation,
			extended subcarrier spacing
RLC/HARQ/RRC	Standard	Timer extension, beam event	HARQ/feedback adaptation
User interface	NR-Uu, Wi-Fi, etc.	Satellite relay/gNB, ISL	Cross-layer API
Spectrum	C-band, mmWave	Ka-band, V/Ku, laser ISL	Dynamic spectrum sharing
Control-plane	IP/MPLS, SDN/NFV	SDN, ISL routing	Unified CN-RAN mapping

^{1(Voicu et al., 2021,} Hernandez et al., 2023)

Spectrum sharing (SS) is increasingly central: four principal STIN SS frameworks—NTN-DL↔TN-DL, NTN-DL↔TN-UL, NTN-UL↔TN-DL, NTN-UL↔TN-UL—entail diverse interference mitigation mechanisms: protection zones, energy-detection sensing, dynamic spectrum access, and advanced game-theoretic allocation (Shang et al., 6 Jan 2025). For IoT-centric and LPWAN use, NB-IoT and LoRa/LoRaWAN technologies are recommended, each with tailored MAC, ALOHA/scheduled access, and extended protocol stacks (Fraire et al., 2021).

4. Performance Metrics and Analytical Models

The performance and capacity evaluation in STINs centers on:

Network capacity: Defined by infrastructure link bottlenecks (ISLs, ground-satellite links), separately from throughput (under specific routing and traffic matrices). Cap-uISL models account for ISL reliability (failures, repair time), revealing that practical throughput saturates well below infrastructure maximum, especially under fixed-path routing (Yang et al., 2024).
Coverage and SINR: Analytical stochastic geometry and Markov models characterize the coverage probability, rate coverage, and cross-tier interference (with explicit dependence on satellite/BS density, pathloss, beamforming gains, and LoS probabilities) (Park et al., 2023, Park et al., 13 May 2026). Integration of terrestrial and satellite links consistently raises the overall coverage and reliability, with up to 20% absolute coverage gain in mid-SINR regimes for blockage-aware deployments (Park et al., 13 May 2026).
QoS and Age of Information (AoI): 6G-motivated metrics extend to finite blocklength coding, with statistical exponents for delay, AoI, and reliability, enabling precise allocation for mURLLC services (Wang et al., 2024).
Resource slicing and slicing delay/cost: Distributed resource slicing (DRS) algorithms minimize aggregate cost and delay violation probability, leveraging hybrid optimization and reinforcement learning across satellite coverage regions (He et al., 2024).

5. Advanced Algorithms and Orchestration Frameworks

Digital Twin–Aided STIN: Real-time DTs mirror UE location, traffic, satellite state, and channel conditions, enabling joint optimization of bandwidth, traffic steering, and power allocation. Convex relaxation and compressed sensing techniques approximate binary allocation constraints, with SCA methods addressing non-convexity in rates/SINRs (Nguyen-Kha et al., 28 Jul 2025, Nguyen-Kha et al., 9 Feb 2026).
Interference Management: Multi-tier coordinated handover and matching algorithms mitigate intra- and inter-system interference, exploit satellite trajectory knowledge, and optimize backhaul allocation while maintaining GEO GS protection (Zhang et al., 2023).
Beamforming and Multi-Connectivity: Multi-connectivity (MC) frameworks support multi-satellite/multi-BS association, spatial macro-diversity, and traffic offloading (Li et al., 2024). Joint MC beamforming maximizes sum rate under cross-link backhaul constraints. Rate-splitting multiple access (RSMA) implements robust joint beamforming for fairness and resilience to CSI and phase uncertainty (Yin et al., 2021, Kim et al., 2023).
Vehicular Edge Computing: Hybrid THz–RF resource allocation for vehicular distributed computing leverages alternating optimization and many-to-one stable matching for efficient offloading across satellite and terrestrial paths (Zhang et al., 26 Jan 2025).

6. Application Scenarios and Benchmark Results

Emergency, Rural, and Mobility Scenarios: STINs underpin emergency comms, cross-continental backbones, and vehicular edge clouds, with digital twins and MC substantially lifting service continuity and QoS under adverse conditions (Wang et al., 11 Mar 2025, Zhang et al., 26 Jan 2025, Wang et al., 2024).
Capacity and Throughput Evaluation: For four major constellations (Kuiper, OneWeb, Telesat, Starlink), cap-uISL modeling shows average ISL unreliability of λ=40 drops Starlink throughput by 28% (to 9.65 Tbps), with utilization at peak traffic staying at ~46% due to spatial bottlenecks and fixed routing (Yang et al., 2024).
Cross-cell Coordination and Slicing: Hybrid data-model DRS using distributed RL and optimization achieves up to 30% cost/reliability improvement over baselines, with fast convergence and robustness to asynchronous/heterogeneous traffic (He et al., 2024).
Urban Blockage and Beam Coverage: Integrated coverage models reveal blockages can complementarily suppress interference, highlighting design leverage between satellite density, beam footprint, and urban penetration (Park et al., 13 May 2026).

7. Future Directions and Research Frontiers

Several major directions are identified:

Multi-orbit, Multi-tier Spectrum Sharing spanning GEO/LEO/UAV, integrating joint STIN/NOMA, advanced dynamic access, and co-primary sharing with interference/fairness guarantees (Shang et al., 6 Jan 2025, Li et al., 2024).
Quantum-assisted and AI-driven Design for combinatorial topology problems (satellite/gateway selection, spectrum assignment), with quantum adiabatic algorithms yielding near-optimal results in fixed practical timeframes (Vercellino et al., 4 Feb 2026).
Standardization and Protocol Evolution: 3GPP/ITU processes are at the forefront, advocating Satellite Access Functions in 5GC, extended NR numerology for high-Doppler/large-RTT, and cross-layer resource slicing (Voicu et al., 2021, Hernandez et al., 2023).
Integrated Edge Computing and Digital Twin Realization: Embedding digital twins and multi-tier orchestration (cloud, edge, on-board processing) for proactive, low-latency service adaptation under strong mobility, delay, and freshness constraints (Tao et al., 2024, Nguyen-Kha et al., 9 Feb 2026).
Blockage-aware, Topology-Optimized Deployment: Stochastic models guide satellite–terrestrial density balancing, elevation angle selection, and backhaul provisioning in heterogeneous terrain, facilitating robust 6G global coverage (Park et al., 2023, Park et al., 13 May 2026).