.png){kind=icon}
.png){kind=icon}
Return
A hybrid learning-assisted multi-parallel algorithm for a large-scale satellite-ground networking optimization problem
.png){kind=icon}
PRE
AI
DOI：10.1007/s42524-025-4098-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The rapid expansion of satellite Internet deployments, driven by the rise of Space-Ground Integration Network (SGIN) construction, has led to a significant increase in satellite numbers. To address the challenge of efficient networking between large-scale satellites and limited ground station resources, this paper presents a hybrid learning-assisted multi-parallel algorithm (HLMP). The HLMP features a multi-parallel solving and deconflicting framework, a learning-assisted metaheuristic (LM) algorithm combining reinforcement learning (RL) and Tabu simulated annealing (TSA), and a linear programming (LP) exact-solving algorithm. The framework first divides the problem into parallel sub-problems based on the time domain, then applies LM and LP to solve each sub-problem in parallel. LM uses LP-generated scheduling results to improve its own accuracy. The deconflicting strategy integrates and refines the planning results from all sub-problems, ensuring an optimized outcome. HLMP advances beyond traditional task-driven satellite scheduling methods by offering a novel approach for optimizing large-scale satellite-ground networks under the new macro paradigm of maximizing linkage to the greatest extent feasible. Experimental cases involving up to 1,000 satellites and 100 ground stations highlight HLMP's efficiency. Comparative experiments with other metaheuristic algorithms and the CPLEX solver further demonstrate HLMP's ability to generate high-quality solutions more quickly.
Keywords:
satellite-ground networking
multi-parallel framework
metaheuristics
reinforcement learning
linear programming
Journal
IF:
7.7
Papers:
437
Citations:
1.5K