【摘要】 随着低轨（LEO）卫星通信技术的进步，LEO卫星星座通信系统进入高速发展阶段。这些系统可以有效地弥补地面蜂窝网的覆盖盲区，作为地面蜂窝网的补充对实现全球互联具有重要意义。针对LEO卫星通信领域中重要的卫星广播调度问题（SBSP），深入分析并构建一个整数规划模型，并设计用于求解该模型的改进变邻域搜索算法（MVNS-SBS）。MVNS-SBS算法通过随机生成的方式获得初始可行的星地连接方案，然后进入VND(Variable Neighborhood Descent)阶段，并利用add和switch两种邻域结构交替进行局部搜索。当VND阶段陷入局部最优时，算法跳转进入shaking阶段，通过remove操作为方案添加扰动，并返回VND阶段。如果在多次扰动后仍未找到更优方案，则认为算法已收敛并终止运行。实验中，为了模拟实际工程场景，在不同规模的经典算例上，把所提算法与粒子群优化（PSO）、竞争Hopfield神经网络（CHNN）和SDBDE(Stochastic Diffusion Binary Differential Evolution)3种有效算法进行对比。实验结果表明，所提算法可以有效求解SBSP，并展现出良好的鲁棒性和实时性。更多还原

【Abstract】 With progress in Low-Earth-Orbit(LEO) satellite communication technology, LEO satellite constellation systems are entering a rapid development phase. These systems can address coverage blind areas in terrestrial cellular networks effectively and play an important role as a supplement to terrestrial cellular networks in achieving global connectivity. Focusing on the important Satellite Broadcast Scheduling Problem(SBSP) in the field of LEO satellite communications, in-depth analysis was carried out, an integer programming model was constructed, and a Modified Variable Neighborhood Search algorithm for Satellite Broadcast Scheduling(MVNS-SBS) was designed for solving this model. In MVNS-SBS algorithm, an initially feasible satellite-ground connection scheme was generated randomly, then the Variable Neighborhood Descent(VND) phase was entered, and local searches were conducted using two neighborhood structures: add and switch alternately. When a local optimum was encountered in the VND phase, a shaking phase was jumped to and entered, perturbations were introduced through remove operations, and the VND phase was returned to. If no better solution was found after several perturbations, the algorithm was considered to be converged and the operation was terminated. In experiments, to simulate real engineering scenarios, on classic examples of different scales, the proposed algorithm was compared with three effective algorithms: Particle Swarm Optimization(PSO), Competitive Hopfield Neural Network(CHNN), and SD-BDE(Stochastic Diffusion Binary Differential Evolution). The results indicate that the proposed algorithm solves SBSP effectively, demonstrating good robustness and real-time performance.更多还原