Long-Term Energy Consumption Optimization-Based Task Offloading Algorithm for Satellite-IoT Systems

Deploying edge computing servers on satellites to provide computing services for Internet of thing (IoT) devices will be an indispensable example for satellite IoT (SIoT) systems. In this paper, we study task offloading problem for SIoT systems. Considering the IoT devices and satellites task queue state and satellite-ground channel state, under the conditions of meeting the limited computing resources and transmit power, and the queue stability, the task offloading problem for SIoT systems is formulated as a minimizing the long-term average system energy consumption problem. In order to solve this problem, We first use Lyapunov optimization method to decouple the joint optimization problem into three subproblems, i.e., computing resource allocation subproblem of IoT devices, computing resource allocation subproblem of satellites, joint task offloading and power allocation subproblem. For computing resource allocation subproblems, we apply Lagrange dual algorithm to solve them. To solve joint task offloading and power allocation subproblem, we formulate it as a Markov decision process (MDP), and propose a parameterized deep Q-network (PDQN)-based task offloading and power allocation algorithm. Finally, the simulation results show that the proposed algorithm has good performance in optimizing the long-term average system energy consumption.

I. Introduction

In recent years, the explosive growth of Internet of things (IoT) devices and the generated massive data have brought great challenges to the computing power and energy consumption of IoT devices. Applying edge computing technology [1] in satellite IoT (SIoT) systems allows IoT devices to offload their tasks to satellites, as a result, task execution performance can be enhanced.