【摘要】 由低轨卫星高速运动引发的星地网络拓扑高动态变化，导致星地多维资源映射关系难以稳定保持，资源利用率降低。提出一种星地资源异质图表征及协同调度方法。首先构建星地多维资源表征异质图，基于异质节点属性表征节点多维资源，实现资源与拓扑的映射关联，并建立异质边多维约束，适应星地资源状态的高动态变化。其次，构建星地传算模型，提出高维空间组合优化问题，在目标空间实现关联决策。最后，设计多目标双重深度Q网络算法，协同优化传输与计算任务调度策略。仿真实验表明，所提算法相比于粒子群优化等算法，资源利用率可提升90%以上，且性能更加稳定。更多还原

【Abstract】 In the satellite-terrestrial networks,the highly dynamic topology variations,caused by the rapid motion of low-orbiting satellites,pose significant challenges in maintaining stable multi-dimensional resource mapping relationships and thus reduce the resource utilization efficiency.To address this issue,a heterogeneous graph-based representation and cooperative scheduling method is proposed for satellite-terrestrial resources.First,a heterogeneous graph is constructed to characterize satellite-terrestrial multi-dimensional resources,where heterogeneous node attributes represent multi-dimensional resource,realizing the mapping between resources and topology,and heterogeneous edges are established with multi-dimensional constraints to adapt to the highly dynamic changes in satellite-terrestrial resource states.Second,a satellite-terrestrial transmission-computing model is developed,and a high-dimensional combinatorial optimization problem is formulated to realize the correlation decision-making in the target space.Finally,an algorithm based on a multi-objective double-deep Q-network is designed to cooperatively optimize the scheduling strategies for transmission and computing tasks.Simulation experiments show that compared with the other algorithms such as particle swarm optimization,the proposed algorithm significantly improves the resource utilization by more than 90% while exhibiting more stable performance.更多还原