Research On Task Offloading Strategy Optimization Method In Low Earth Orbit Satellite Internet Of Things

As an important communication technology,IoT technology undertakes a series of important functions such as detection,monitoring,transmission,and processing.However,for IoT devices located in remote areas,their computing resources and device power are not sufficient to support complex computing tasks,and they cannot offload the tasks to the cloud for processing via 5G cellular networks.Low earth orbit satellite networks become an important complement to 5G cellular networks due to their wide coverage,wide bandwidth,low impact by geographical factors,and high robustness.By deploying mobile edge computing servers on satellites,low-latency and highefficiency edge computing services can be provided to IoT devices in remote areas.However,how to efficiently utilize the overall computational and bandwidth resources of the network and improve the communication efficiency and task processing efficiency of the network is a pressing issue in the low earth orbit satellite IoT system.In this paper,we address the offloading decision and power allocation problems of IoT devices and low earth orbit satellites,model the complex environment of satellite networks,and propose a computational offloading and power allocation algorithm based on multiagent reinforcement learning to improve the overall system efficiency.Firstly,this paper designs a three-layer computational offloading architecture of local device-low earth orbit satellite-cloud computing center around the computational task offloading problem of IoT devices,and models the communication and computational aspects of the system model.For each task,this paper considers the delay and energy consumption of all computing tasks in the system as optimization objectives in a shorter time scale,and models the optimization problem as a partially observable Markovian decision process,proposes an offloading decision and power allocation algorithm based on a multi-agent deep deterministic policy gradient,and verifies the convergence and effectiveness of the algorithm proposed in this paper through simulation experiments.On the basis of the above research,because the computing resources of a single satellite are not enough to load the continuously dense arrival of computing tasks,and considering the unbalanced load in the satellite network,this paper designs a collaborative processing architecture for low earth orbit satellite computing tasks around the task offloading problem in the low earth orbit satellite network,in order to achieve balanced load,reduce latency,and optimize user experience.In this paper,the satellites in the LEO satellite network are considered as fully collaborative relationships,and a collaborative LEO satellite task processing algorithm based on multi-agent proximal policy optimization is proposed to optimize the secondary assignment decision of computational tasks to satellites.The subsequent simulation results of this paper verify the optimization effect of the proposed algorithm on the overall latency and energy consumption.The research results in this paper provide a new perspective on the optimization of computational task offloading decision in LEO satellite IoT and are of reference significance for the optimization of edge computing decision in future LEO satellite IoT.