Modeling and Analysis of Massive Low Earth Orbit Communication Networks

Abstract

Non-terrestrial networks are foreseen as a crucial component for developing 6th generation (6G) of wireless cellular networks by many telecommunication industries. Among non-terrestrial networks, low Earth orbit (LEO) communication satellites have shown a great potential in providing global seamless coverage for remote and under-served regions where conventional terrestrial networks are either not available or not economically justifiable to deploy. In addition, to the date of writing this summary, LEO communication networks have became highly commercialized with many prominent examples, compared to other non-terrestrial networks, e.g., high altitude platforms (HAPs) which are still in prototyping stage.

Despite the rapid promotion of LEO constellations, theoretical methodologies to study the performance of such massive networks at large are still missing from the scientific literature. While commercial plans must obviously have been simulated before deployment of these constellations, the deterministic and network-specific simulations rely on instantaneous positions of satellites and, consequently, are unable to characterize the performance of massive satellite networks in a generic scientific form, given the constellation parameters.

In order to address this problem, in this thesis, a generic tractable approach is proposed to analyze the LEO communication networks using stochastic geometry as a central tool. Firstly, satellites are modeled as a point process which enables using the mathematics of stochastic geometry to formulate two performance metrics of the network, namely, coverage probability and data rate, in terms of constellation parameters. The derivations are applicable to any given LEO constellation regardless of satellites’ actual locations on orbits. Due to specific geometry of satellites, there is an inherent mismatch between the actual distribution of satellites and the point processes that are used to model their locality. Secondly, different approaches have thus been investigated to eliminate this modeling error and improve the accuracy of the analytical derivations.

The results of this research are published in seven original publications which are attached to this summary. In these publications, coverage probability and average achievable data rate of LEO satellite networks are derived for several communication scenarios in both uplink and downlink directions under different propagation models and user association techniques. Moreover, the analysis was generalized to cover the performance analysis of a multi-altitude constellation which imitates the geometry of some well-known commercial constellations with satellites orbiting on multiple altitude levels. While direct communication between the satellites and ground terminals is the main studied communication scenario in this thesis, the performance of a LEO network as a backhaul for aerial platforms is also addressed and compared with terrestrial backhaul networks.

Finally, all analytical derivations, obtained from stochastic modeling of the LEO constellations, are verified through Monte Carlo simulations and compared with actual simulated constellations to ensure their accuracy. Through the numerical results, the performance metrics are evaluated in terms of different constellation parameters, e.g., altitude, inclination angle, and total number of satellites, which reveals their optimal values that maximize the capacity and/or coverage probability. Therefore, other than performance analysis, several insightful guidelines can be also extracted regarding the design of LEO satellite networks based on the numerical results.

Stochastic modeling of a LEO satellite network, which is proposed for the first time ever in this thesis, extends the application of stochastic geometry in wireless communication field from planar two-dimensional (2D) networks to highly heterogeneous three-dimensional (3D) spherical networks. In fact, the results show that stochastic modeling can also be utilized to precisely model the networks with deterministic nodes’ locations and specific distribution of nodes over the Euclidean space. Thus, the proposed methodology reported herein paves the way for comprehensive analytical understanding and generic performance study of heterogeneous massive networks in the future.