基于算力路由的空间信息网络低时延在轨协同计算策略

针对当前空间信息网络（Spatial Information Network， SIN）地面数据处理模式中存在的高传输时延问题，提出了基于算力路由的低时延在轨协同计算策略。考虑到卫星网络的动态变化及卫星计算资源的异构性，提出了要素时空扩展图模型，以实现在屏蔽SIN动态性的同时对星上资源进行精确表征。在此基础上，构建基于算力路由的多星在轨协同计算数学模型，提出时延优化问题，并利用改进的异构最早完成时间（Heterogeneous Earliest Finish Time， HEFT）算法进行求解。仿真结果表明：卫星在轨协同计算可有效降低卫星数据的任务处理时延；同时，本文所提的改进HEFT算法以较小且可接受的收敛性能为代价，换取了业务处理时延的大幅度降低。

Computation dependent routing based on-orbit collaborative computing strategy for satellite information networks

In order to solve the problem of high transmission latency in the current data processing mode of Spatial Information Network (SIN), a low latency on-orbit collaborative computing strategy based on computation dependent routing is proposed. Considering the dynamic change of satellite network and the heterogeneity of satellite computing resources, an element time expanded graph model is proposed to realize the computing resources of satellite and shield the SIN dynamic characteristic. On this basis, a mathematical model of multi-satellite on-orbit collaborative computing based on computation dependent routing is constructed, and the latency optimization problem is proposed. To solve the above problem, an improved Hetergenetous Earliest Finish Time (HEFT) algorithm is designed. The simulation results show that the on-orbit collaborative computation can effectively reduce the task processing latency of satellite date, and the proposed improved HEFT algorithm reduce the task processing latency substantially with a little and acceptable convergence performance.