TL;DR
This paper presents the first comprehensive reliability analysis for multi-hop routing in multi-tier Low Earth Orbit (LEO) satellite networks using stochastic geometry. By modeling satellites as random points on concentric spheres, the authors derive analytical expressions for "interruption probability" and propose priority strategies that optimize how data hops between different satellite altitudes and ground gateways.
What Problem Does This Paper Address?
As mega-constellations like Starlink and OneWeb populate the skies, we are moving toward a "network of networks" where data must hop across multiple satellites at different altitudes (tiers) and ground stations to reach its destination.
Traditional routing analysis often relies on deterministic models (assuming we know exactly where every satellite is at all times), which is computationally exhausting for thousands of moving parts. Furthermore, satellite routing faces unique physical hurdles: the Earth itself can block line-of-sight, signals attenuate over long distances, and different constellations operate at different heights. There was a lack of a rigorous mathematical framework to calculate the reliability of a multi-hop path in such a dynamic, multi-tiered environment.
Methods
The authors employ Stochastic Geometry to treat satellite locations as a Binomial Point Process (BPP) on spherical surfaces. This allows them to calculate the probability of finding a relay within a specific "constraint region" defined by:
- Direction Angle: Ensuring the next hop moves toward the destination.
- Minimum Dome Angle: Preventing tiny, inefficient hops.
- Maximum Dome Angle: Accounting for signal decay and Earth blockage.
To analyze multi-hop paths, they use a Markov Chain-based framework. They define a Transition Probability Matrix (TPM) where each state represents a satellite tier. An "absorbing state" is added to represent a routing interruption (failure).
They propose three priority strategies for selecting relays:
- Stationary Optimal Priority Strategy (SOPS): Uses the "steady-state" distribution of the network to minimize the overall interruption probability.
- Single-hop Interruption Probability (SHIP) Inspired Strategy: A lower-complexity alternative that prioritizes tiers based on their individual hop reliability.
- Density Inspired Strategy: A "naive" approach that simply prioritizes tiers with the most satellites.
Datasets
The paper does not use a traditional training dataset. Instead, it validates its mathematical models using Monte Carlo simulations ($10^6$ iterations) based on parameters from real-world LEO constellations:
- Kepler-like tier: 140 satellites at 575 km.
- OneWeb-like tier: 720 satellites at 1200 km.
- Terrestrial tier: 300 ground gateways.
Key Insights and Ideas
- The "Inverted Pyramid" Advantage: Networks are more reliable when more satellites are deployed in higher altitude tiers rather than lower ones. Higher satellites have a wider "view," making them more likely to find a valid next hop.
- Satellites > Gateways: Increasing the number of satellites is significantly more effective at boosting reliability than increasing the number of ground gateways.
- Convergence to Stationary Distribution: Regardless of where a transmission starts (e.g., a ground station), the probability of the data being at a certain tier quickly converges to a "stationary distribution" after a few hops. This simplifies long-distance reliability calculations.
- Multi-flow is King: The authors found that "multi-flow" routing (sending data along multiple paths simultaneously) provides a much larger reliability boost than simply using "smarter" tier-selection logic (dynamic priority).
Key Results
- Model Accuracy: The analytical formulas for interruption probability matched the Monte Carlo simulation results almost perfectly across various scenarios.
- Hop Count: In a three-tier case study (ground + two satellite layers), the average number of hops required for a successful cross-globe transmission was approximately 6.08.
- Strategy Performance: The proposed SOPS strategy approached the ideal lower bound for interruption probability. While the SHIP strategy was slightly less reliable, it offered a significant reduction in computational complexity ($O(K^3)$ vs $O(K^3 \cdot K!)$ for $K$ tiers).
Limitations and Future Work
- Traffic and Congestion: The current model assumes satellites are always available if they are in range. In reality, network congestion could block a path. The authors suggest using Marked Poisson Point Processes (MPPPs) in future work to account for traffic loads.
- Orbital Mechanics: While BPP is a good proxy for massive constellations, it doesn't capture the specific "street-of-coverage" patterns of actual orbital planes. Integrating more precise orbit geometry models is a noted next step.
- Dynamic Environments: The paper suggests further research into "intelligent" routing that adjusts directional angles dynamically to avoid interruptions in real-time.
Summary updated