【摘要】 卫星在星地网络中计算能力和数量的快速提高，使天地协同任务卸载成为可能。天地协同卸载需要一个有效的任务调度算法，将任务分配到适合的服务器以提升任务卸载性能。然而，卫星高动态性带来的网络拓扑变化，以及终端任务子任务间的并行性导致服务器资源挤兑与依赖性引起的调度顺序要求，对任务协同卸载提出了新的挑战。修剪路径(PruningPath)算法实现了星地设备间任务的联合计算，并在满足子任务间依赖关系的前提下，将最小任务能耗问题转化为最短路径问题。实验结果表明，PruningPath在降低任务响应延迟方面具有显著优势，可降低卸载时延达31.9%。更多还原

【Abstract】 The rapid improvement of the computing power and the number of satellites in the network makes satellite-ground collaborative task offloading possible.It is necessary to have an effective task scheduling algorithm to allocate tasks to appropriate servers to improve task offloading performance.However, changes in network topology brought by high dynamics of satellite and the scheduling sequence requirement caused by server resource crowding, and dependency caused by the parallelism between terminal task subtasks, pose new challenges to collaborative task offloading.PruningPath implements joint calculation of tasks between satellite and ground devices, and on the premise that the dependency between subtasks is satisfied, transforms the problem of minimum task energy consumption into the problem of the shortest path.Experimental results show that the algorithm has significant advantages in reducing task response latency, and can reduce the offloading latency by up to 31.9%.更多还原