Beam-to-Satellite Scheduling for High Throughput Satellite Constellations Using Particle Swarm Optimization

The next generation of satellite communications will be characterized by a paradigm shift that will transform a traditionally static market into a frequently shifting environment. Fluctuating demand, highly flexible payloads, and the usage of non-geostationary orbits will boost the constellations' capacity to unthinkable limits, at the cost of additional complexity. This novel operational context carries unique problems that were non-existent in earlier stages of this industry. In this paper, we formulate and solve one of these novel problems, the Beam-to-Satellite scheduling problem, which focuses on deciding when to activate and deactivate a specific beam on a particular satellite. First, we describe the problem in terms of its scheduling variables, time-related constraints, and objective function, based on a combination of load balancing between the satellites and interference minimization. Second, we derive a linear-integer programming formulation of the problem, which can be optimally solved using common mathematical solvers. Given that those are computationally infeasible for high-dimensional scenarios (i.e.,
$> \mathbf{200}$
beams), we then propose a single-objective PSO implementation. Finally, we test the algorithm over different high-dimensional scenarios taken from a realistic dataset, with tens of thousands of beams, provided by a satellite operator. Our PSO approach proves to be an effective technique to scan the search space and reach a satisfactory solution in a reasonable time. Using the same dataset, we benchmark the PSO implementation against heuristic solutions and show that it improves by between 39% and 73% the resource consumption overhead and around 30% the demand balancing between the satellites. In addition, we also demonstrate that it outperforms other metaheuristics, such as genetic algorithms and probabilistic algorithms, over all scenarios considered.