CECTO: Cloud-Edge Collaborated Task Offloading Based on Dual Time-Scale A2C

Lina Ni, Jiehan Zhou, Shiping Chen, Chen Yu

Abstract: Optimizing task offloading in the context of Power IoT (PIoT) through cloud-edge collaboration is crucial for smart grid development. This paper combines the most advanced Space-Air-Ground Internet (SAGIN) with PIoT and proposes a cloud-edge collaborated computing scheme namely CECTO based on the Double Time-Scale Advantage Actor-Critic (DTS-A2C) algorithm. It decouples the collaborated computing model into subproblems using the Lyapunov method. For the task offloading, it applies the DTS-A2C algorithm to optimize the data transmission power on a small time-scale and task offloading decisions on a large time-scale. Experimental results demonstrate that the CECTO method can significantly reduce task delays and device energy consumption.