卫星物联网中面向多类型任务的计算卸载策略.

Integrating edge computing with satellite Internet of Things (SIoT) allows tasks generated by local IoT devices to be processed closer to the edge, significantly relieving the computational burden on local IoT devices. However, given the diversity of computational tasks in satellite IoT, deploying a specific computing offloading strategy for each type of task leads to wastage of computational resources due to varying requirements based on task characteristics. Moreover, the randomness in the generation of computational tasks in satellite IoT requires a long-term optimization approach; otherwise, it may result in unde-rutilization of computational resources of the devices, further increasing the task processing delay. To address these issues, this paper proposed a multi-type task computing offloading strategy based on DQN, incorporating queue theory to minimize the long-term average delay in system task processing. Additionally, to enhance the training outcomes, it redesigned the exploration method during the training phase. Finally, extensive simulation results indicate that the proposed strategy effectively reduces the average task processing delay in the system compared with other offloading strategies. [ABSTRACT FROM AUTHOR]