Achieving Predictable and Scalable Load Balancing Performance in LEO Mega-Constellations

With the increasing deployment scale of Low Earth Orbit (LEO) mega-constellations, more satellites are expected to become visible to users simultaneously, bringing a new opportunity to optimize the network performance by properly assigning users to satellites. In this paper, we consider LEO mega-constellations without Inter-Satellite Links (ISLs) while assuming there are enough ground relays for inter-satellite communications. To establish a path from a user terminal to the nearest ground station (which serves as a gateway to the Internet), shortest path routing is usually adopted. To focus on the problem of user-satellite assignment, as well as to make routing more scalable, we divide the routing process into two parts: assigning the user terminal to a visible satellite, and finding a path from the satellite to the nearest ground station. For simplicity, shortest path routing is assumed in the second part. Aiming at minimizing the Maximum Satellite Utilization (MSU), a Mixed Integer Linear Programming (MILP), called Optimal User-Satellite Assignment (OUSA), is formulated. Performance evaluations are conducted based on Starlink Phase I mega-constellation and AWS ground station locations. As compared with the existing solutions, we show that the average load balancing performance can be improved by up to 33.29%.