This paper explores the architectural "sweet spot" for 6G non-terrestrial networks (NTNs), comparing standalone satellite systems against multi-layered hierarchies that include Low Earth Orbit (LEO) satellites and High Altitude Platforms (HAPs). The authors demonstrate that integrating HAPs as relays (the GEO-HAP-Earth configuration) provides the best balance of capacity and reliability, outperforming more complex three-hop architectures.
TL;DR
The paper evaluates how to best combine satellites and aerial platforms to achieve 6G's promise of global, high-capacity connectivity. It finds that using High Altitude Platforms (HAPs) as intermediate relays for geostationary signals offers up to a 6x capacity boost and significantly better reliability compared to standalone satellites or LEO-based relaying.
What Problem Does This Paper Address?
While 5G is rolling out, 6G aims for "ubiquitous connectivity"—meaning high-speed internet even in the middle of the ocean, deep deserts, or remote rural areas where cell towers are impractical.
To do this, we need Non-Terrestrial Networks (NTNs). Traditionally, this meant Geostationary (GEO) satellites, which cover huge areas but suffer from massive signal loss and high latency. Low Earth Orbit (LEO) satellites and High Altitude Platforms (HAPs, like solar-powered gliders or balloons at 20km altitude) offer lower latency but smaller coverage. The challenge is determining the optimal way to "stack" these layers. Should a signal go straight from a GEO satellite to your phone, or should it be relayed through a LEO satellite or a HAP first?
Methods
The authors developed a mathematical framework to compare four distinct networking architectures:
- GE (GEO-to-Earth): The baseline standalone satellite approach.
- GLE (GEO-LEO-Earth): Using LEO satellites as relays.
- GHE (GEO-HAP-Earth): Using HAPs as relays.
- GLHE (GEO-LEO-HAP-Earth): A complex three-hop integrated architecture.
They modeled these as 3D multi-hop networks using an Amplify-and-Forward (AF) relay protocol. The study incorporates sophisticated channel models that account for atmospheric gas attenuation, ionospheric scintillation, rain fading, and clutter loss. They tested these across two frequency bands: the S-band (2 GHz) for reliability and the Ka-band (20/30 GHz) for high capacity. The theoretical models were validated using 10,000 Monte Carlo simulation runs.
Datasets
Not discussed in the paper. The study relies on standardized system parameters provided by 3GPP (Rel-17/18) and ITU-R technical reports to ensure the simulations reflect realistic hardware constraints and environmental conditions.
Key Insights and Ideas
- HAPs are the "Atmospheric Bridge": The most significant signal degradation happens in the lower atmosphere. HAPs sit at 20km, just above most of this "weather." By relaying signals through a HAP, the "bottleneck" Earth-link is shortened from thousands of kilometers to just 20km, drastically improving signal quality.
- The LEO Hardware Bottleneck: Surprisingly, LEO satellites (GLE) can sometimes perform worse than direct GEO links in the Ka-band. This is because LEO satellites are often power-constrained and have smaller antennas compared to massive GEO satellites, resulting in a lower Effective Isotropic Radiated Power (EIRP).
- Complexity vs. Performance: The most complex architecture (GLHE) actually showed a 42% decrease in capacity compared to the simpler GHE (GEO-HAP) model. The extra hop through a LEO satellite adds more noise and hardware limitations without providing a proportional benefit in signal strength.
- Elevation Matters: Performance drops sharply at low elevation angles (below 20°) because the signal has to travel through a much thicker slice of the atmosphere, increasing "clutter" and fading.
Key Results
- Capacity Gains: In the 2 GHz band, the GHE architecture offers 6x higher capacity than the standalone GE baseline.
- Frequency Trade-offs: Transitioning to the Ka-band (20/30 GHz) provides 10x higher capacity than the S-band, but it is much more sensitive to alignment and atmospheric conditions.
- Reliability: For a target outage probability of 1%, the GHE architecture can operate at a Signal-to-Noise Ratio (SNR) threshold of 24 dB, whereas the standalone GEO baseline fails at anything above -6 dB.
- LEO Altitude: For LEO-based relaying, a 600km orbit is significantly better than 1200km due to lower path loss in the satellite-to-ground link.
Limitations and Future Work
The authors acknowledge that their current model is a "snapshot" in time. LEO satellites and HAPs are mobile, and their movement introduces dynamic topology changes and Doppler shifts that weren't the primary focus here.
Future research directions include:
- End-to-End Latency: Quantifying the delay introduced by multi-hop processing.
- Mobility Management: How to handle seamless handovers as LEO satellites fly overhead or HAPs drift.
- Packet Delivery Rates: Moving from physical layer analysis (SNR/Capacity) to network layer performance (throughput/packet loss).
Summary updated