上海航天（中英文）2025，Vol.42Issue(5)：121-130,10.DOI:10.19328/j.cnki.2096-8655.2025.05.014
基于GPU加速改进粒子群算法的多波束卫星通信资源优化
GPU-Accelerated Particle Swarm Optimization for Resource Allocation in Multibeam Satellite Communications
摘要
Abstract
With the widespread application of low-Earth-orbit(LEO)constellations and multi-beam satellite systems in broadband access,Internet of Things(IoT),and other fields,the scenario requirements of constellation dynamic Earth-oriented communication and dynamic service node selection have become increasingly prominent;thus,the efficiency and optimization quality of intelligent resource scheduling have become key factors influencing system performance.Traditional optimization algorithms encounter challenges such as limited encoding capability,cumbersome constraint handling,and slow convergence when addressing joint optimization problems involving multi-dimensional decision variables(e.g.,beams,power,and bandwidth).To address these issues,this paper proposes an improved particle swarm optimization algorithm(PPSO)based on a hybrid Stick-breaking encoding mechanism and GPU parallel acceleration,designed to efficiently solve intelligent resource allocation optimization problems in LEO constellation-based multi-beam satellite systems.The proposed method reconstructs the solution space of particles through hybrid Stick-breaking encoding,ensuring that normalized variables inherently satisfy global constraints,thereby eliminating complex constraint correction operations required in traditional methods.Furthermore,GPU-accelerated parallel computation is employed for particle swarm evolution and fitness evaluation,significantly improving algorithmic efficiency while preserving solution quality.Experimental results demonstrate that the proposed method outperforms existing approaches in optimizing key performance metrics,including total system delay,packet loss rate,and energy consumption,especially in large-scale constellation scenarios requiring dynamic Earth-oriented communication and dynamic node selection,exhibiting superior scalability and computational advantages.关键词
低轨(LEO)星座/动态调度/图形处理器(GPU)加速/优化设计/粒子群算法(PSO)Key words
low-Earth-orbit(LEO)constellation/dynamic scheduling/graphics processing unit(GPU)acceleration/optimization design/particle swarm optimization分类
航空航天引用本文复制引用
宋自阳,张廷尧,赵家庆,慕忠成,黄益新,付哲楷..基于GPU加速改进粒子群算法的多波束卫星通信资源优化[J].上海航天（中英文）,2025,42(5):121-130,10.基金项目
上海市自然科学基金面上资助项目(23ZR1432400) （23ZR1432400）
