Computation-Dependent Routing Based Low-Latency Decentralized Collaborative Computing Strategy for Satellite-Terrestrial Integrated Network

By jointly optimizing the computation offloading and routing, the Computation-Dependent Routing (CDR) can significantly reduce the task processing latency in Satellite-Terrestrial Integrated Network (STIN). However, existing CDR-related works mainly focus on the performance improvement in the central static environment while ignoring the dynamic decentralized environment of STIN. To support the latencysensitive tasks in STIN, in this paper, we propose a CDR based low-latency decentralized collaborative computing strategy (LOCS), in which a Directed Diffusion enhanced Task Scheduling algorithm (DETS) is proposed to achieve the computing while transmitting distributedly by designing a novel computing gradient that both considers the computing and communication resources. Moreover, to overcome the STIN dynamics, a cross time slot mechanism based on the Time Expanded Graph (TEG) model is introduced. The latency performance of LOCS is analyzed. Numerical results verify that the proposed LOCS performs stronger resilience compared with the central scheme.