ISTNs-Based Computing and Communication Resource Allocation and Optimization for IoT

Abstract

With the growing requirements of computation tasks in the Internet of Things (IoT), integrated satelliteterrestrial networks (ISTNs) are being applied in the IoT increasingly, among which low earth orbit (LEO) satellites as edge computing nodes attract extensive attention. Nevertheless, the limited computing resources on LEO satellites as well as the constrained spectrum resources of satellite-terrestrial transmission links pose significant challenges for computation task offloading to LEO satellites. Therefore, suitable selection of computing nodes and transmission links is necessary to fully leverage existing computing resources and enhance transmission efficiency. In this paper, an edge computing network infrastructure utilizing LEO satellites is introduced. Within this framework, a comprehensive scheme for computation task offloading that encompasses the selection of both computing nodes and transmission links is proposed. Then, an optimization objective is designed for the purpose of minimizing computational latency and system energy consumption jointly, and the proposed problem is formulated as a Markov decision process (MDP). Meanwhile, the proximal policy optimization (PPO) is employed to make optimal decisions and a utility function is defined to quantify the system overhead, thus computational latency as well as system energy consumption can be reduced significantly. Compared with conventional DRL-based offloading methods, the proposed approach can reduce the average task execution delay by about 16% and the average system energy consumption by about 20%.