Traditional cellular networks are fundamentally constrained by two-dimensional geography. Building physical cell towers across oceans, rugged mountain ranges, or vast deserts is economically prohibitive and often physically impossible. As a result, large swaths of the planet remain connectivity "dead zones," isolating maritime vessels, remote industrial infrastructure, and underserved rural populations.
To bridge this global connectivity gap and achieve the ubiquitous coverage promised by 6G, researchers are expanding the network architecture into the third dimension. By weaving together traditional ground infrastructure with airborne platforms and orbital satellites, the telecommunications field is shifting toward Space-Air-Ground Integrated Networks (SAGIN). This multi-tiered framework transforms non-terrestrial assets from isolated relay pipes into an intelligent, collaborative extension of the global internet fabric.
The Multi-Tier Spatial Hierarchy
The core intuition behind SAGIN is that no single tier can solve the global coverage problem alone. Instead, the architecture balances a fundamental physical trade-off: altitude versus latency and footprint.
▲ Space Tier (GEO, MEO, LEO Satellites) ─── High Latency, Massive Footprint
│ Air Tier (HAPs, Balloons, UAVs) ─── Medium Latency, Flexible Deployment
▼ Ground/Sea Tier (BSs, Vessels, Nodes) ─── Low Latency, Localized Capacity
As outlined by Nasir Saeed and Heba Almorad in Point-to-Point Communication in Integrated Satellite-Aerial 6G Networks: State-of-the-Art and Future Challenges, a robust 3D network topology distributes responsibilities across three primary layers:
- The Space Tier: Composed of Geostationary (GEO), Medium Earth Orbit (MEO), and Low Earth Orbit (LEO) satellites. As Hazer İnaltekin et al. explain in Future satellite communications: Satellite constellations and connectivity from space, multi-layer satellite diversity allows ground terminals to maintain continuous paths through orbital handovers, providing wide-area backhaul and global coverage.
- The Air Tier: Operating in the stratosphere and troposphere, this layer utilizes High Altitude Platforms (HAPs), solar-powered balloons, and Unmanned Aerial Vehicles (UAVs). These nodes can be rapidly deployed to provide localized, high-bandwidth coverage over disaster zones or high-density events, acting as a flexible intermediary layer.
- The Ground and Sea Tier: Comprises traditional terrestrial base stations, mobile vehicles, marine vessels, and specialized remote terminals.
By combining these tiers, the network can dynamically route data based on environmental demands. For example, Dengke Wang et al. demonstrate in The Potential of Multi-Layered Hierarchical Non-Terrestrial Networks for 6G that while full three-hop systems (GEO-LEO-HAP) offer maximum coverage, a hybrid GEO-HAP architecture often provides the optimal balance of performance and complexity for regional deployment. Similarly, Houcem Ben Salem et al. show in Exploiting Hybrid Terrestrial/LEO Satellite Systems for Rural Connectivity how integrating sparse terrestrial infrastructure with LEO constellations significantly elevates data rates in underserved rural areas.
Dynamic Coordination and Resource Optimization
Integrating moving aerial and orbital nodes with static ground infrastructure introduces severe mathematical and operational challenges. Unlike fixed cell towers, LEO satellites travel at several kilometers per second, and UAVs are subject to wind currents and strict battery limitations. Maintaining a stable link requires continuous, real-time optimization of power, beamforming, and user association.
To evaluate these highly dynamic environments, researchers rely on advanced spatial modeling. In Outage Probability Analysis of Uplink Heterogeneous Non-terrestrial Networks: A Novel Stochastic Geometry Model, Wen-Yu Dong et al. introduce a novel point process model to capture the clustered and exclusive spatial distribution of heterogeneous aerial vehicles, providing a mathematical foundation for predicting uplink outages under realistic fading conditions.
With a mathematical model established, the network must actively manage its resources. Ahmad Alsharoa addresses this in Improvement of the Global Connectivity Using Integrated Satellite-Airborne-Terrestrial Networks With Resource Optimization by proposing a two-stage optimization framework that jointly manages transmit power, user associations, and physical HAP placement, boosting throughput by up to 88%.
┌──────────────────────────┐
│ Joint Optimization │
│ ┌────────────────────┐ │
│ │ User Association │ │ ──► Maximized Throughput
│ ├────────────────────┤ │ ──► Controlled Interference
│ │ Beamforming Matrix │ │
│ └────────────────────┘ │
└──────────────────────────┘
This joint optimization philosophy is echoed by Shasha Liu et al. in Joint User Association and Beamforming in Integrated Satellite-HAPS-Ground Networks, where directional beamforming is coordinated across space and air tiers to mitigate inter-cell interference. Furthermore, hardware-level adjustments are critical: Seong-Jun Kim et al. demonstrate in Non-Terrestrial Networks for UAVs: Base Station Service Provisioning Schemes With Antenna Tilt that dynamically tilting base station antennas is essential for minimizing network outages when serving ground users and aerial drones simultaneously.
In scenarios where real-time channel measurements are too costly or delayed—such as deep-sea maritime networks—predictability becomes the primary asset. In Hybrid Satellite-UAV-Terrestrial Networks for 6G Ubiquitous Coverage: A Maritime Communications Perspective, Yanmin Wang et al. exploit the highly predictable trajectories of ocean vessels to schedule links proactively, achieving up to 83% energy savings while maintaining continuous broadband coverage for the "Internet of Ships."
Moving Computation to the Sky
The vision of SAGIN extends beyond merely routing communication packets; it also encompasses the distribution of computational power. Autonomous systems, such as the Internet of Vehicles (IoV) or offshore environmental sensors, generate massive amounts of data that require low-latency processing. When these systems operate outside the range of terrestrial edge servers, the non-terrestrial network must step in as a distributed cloud.
This paradigm, termed Aerial Computing, is detailed by Quoc-Viet Pham et al. in Aerial Computing: A New Computing Paradigm, Applications, and Challenges. By mounting edge computing servers directly onto UAVs and low-orbit satellites, the network creates an "in-the-sky" processing infrastructure.
[Remote Vehicle / Node]
│
▼ (Computation Offloading)
┌────────────────────────────────────────────────────────┐
│ Sky-Based Edge Infrastructure │
│ ├── [UAV Swarm] ─── Localized, Low-Latency Compute │
│ └── [LEO Satellite] ── Wide-Area Backhaul / Fallback │
└────────────────────────────────────────────────────────┘
To make this work, the network must intelligently decide when and where to offload tasks. Haofei Li et al. tackle this in Deep Deterministic Policy Gradient-Based Algorithm for Computation Offloading in IoV, utilizing a Deep Reinforcement Learning (DRL) approach to dynamically split complex tasks between a vehicle's local processor and orbital LEO servers.
Because individual drones face strict energy and payload constraints, multi-agent collaboration is required. In Collaboration in the Sky: A Distributed Framework for Task Offloading and Resource Allocation in Multi-Access Edge Computing, Yan Kyaw Tun et al. leverage the Alternating Direction Method of Multipliers (ADMM) to let a swarm of UAVs cooperatively balance computational loads without relying on a centralized ground controller. To prevent system failures when a drone's battery is exhausted, modern frameworks increasingly integrate space assets as a robust computational fallback, as explored by Long He et al. in QoE Maximization for Multiple-UAV-Assisted Multi-Access Edge Computing via an Online Joint Optimization Approach and Yulan Gao et al. in Cost-Efficient Computation Offloading in SAGIN: A Deep Reinforcement Learning and Perception-Aided Approach.
Real-World Implementations and Future Horizons
The transition to space-air-ground integration has moved past pure mathematical modeling into physical validation. A major milestone in this transition is documented by Ruolin Xing et al. in From Earth to Space: A First Deployment of 5G Core Network on Satellite, which details the first successful deployment of a unmodified 5G Core (5GC) network on a LEO satellite. This deployment proved that standard terrestrial cellular protocols can operate natively in orbit, paving the way for seamless consumer-device-to-satellite connectivity.
Looking toward the extreme performance metrics of 6G, Xiaohu You et al. demonstrate in Toward 6G TKμ Extreme Connectivity: Architecture, Key Technologies and Experiments how integrating native AI with "Super Edge Nodes" across a three-layer architecture can shift networks from session-centric to task-centric operations, achieving Terabit-per-second data rates in real-world testbeds.
On the horizon, the boundaries of SAGIN are expanding even further. In the long term, the architectural principles used to manage latency and handovers between Earth and LEO satellites may serve as the foundation for interplanetary communication. As Lik‐Hang Lee et al. outline in Beyond the Blue Sky of Multimodal Interaction: A Centennial Vision of Interplanetary Virtual Spaces in Turn-based Metaverse, overcoming the extreme multi-minute propagation delays between Earth and Mars will require decentralized orbital server nodes and predictive "digital twins"—extending the concept of space-air-ground integration to a truly interplanetary scale.