Hybrid MOEA with problem-specific operators for beam-hopping based resource allocation in multi-beam LEO satellites

Abstract

The efficient allocation of satellite communication resources has become increasingly vital due to the dynamic and growing nature of traffic demand. The beam-hopping resource allocation technique addresses this challenge by enabling sequential and adaptive beam illumination, along with a dynamic distribution of power and bandwidth based on existing user demand. This work formulates the large-dimensional and highly constrained beam-hopping problem for low-power, low earth orbit satellites as a multi-objective optimization problem. It considers three key objectives: unserved capacity (UC), which measures the portion of traffic demand that remains unmet; extra served capacity (EC), which reflects excess traffic delivered beyond the requested demand, indicating possible inefficiencies; and time to serve (TTS), which represents the average waiting time for users in non-illuminated cells. Aiming at innovating in optimization, specialized initialization, crossover, mutation, and local search operators for multi-objective evolutionary algorithms (MOEAs) have been proposed. Performance is assessed through Hypervolume (HV) metrics and statistical confidence analysis. An extensive experimental analysis is presented first for a canonical NSGA-II algorithm, characterizing the impact of the new operators and hybrid components on performance. Then, beyond Pareto-based approaches such as NSGA-II, a study of both decomposition- and indicator-based MOEAs, namely MOEA/D and SMS-EMOA is assessed, demonstrating the generalizability of the presented approach. Compared to previous results in the literature, our hybrid approaches achieve up to 5% improvement in UC, up to 100% gains in EC, and up to 60% improvement in TTS for the best configuration.