Research On Task Offloading Strategy Based On UAV-Assisted Edge Computing

Under the background of the commercial use of the fifth generation(5G)mobile communication network,services such as augmented reality,virtual reality,and cloud gaming have developed rapidly,and edge computing technology has also been widely used due to its low latency and proximity.Therefore,task offloading strategy has great research significance as the key to edge computing technology.At the same time,considering the limitation that edge computing servers cannot provide efficient computing services for various scenarios such as natural disasters,remote areas,and temporary hot spots when they are deployed in fixed base stations.Therefore,,the UAV-assisted edge computing system formed by using unmanned aerial vehicles(UAVs)to carry edge computing servers and communication base stations has been extensively studied.On the other hand,research on the network architecture and key technologies of the Sixth Generation(6G)mobile communication network has also been carried out in academia.Among them,the Space-Aerial-Ground Integrated Network(SAGIN)has the characteristics of wide-area coverage.Through satellite/UAV-assisted edge computing,it can provide reliable communication and computing services for user equipment in remote areas such as deserts and oceans.However,the mobility and battery capacity constraints of the UAV,the difference of computing nodes and the task scheduling under the partial offloading model have all become the challenges that this paper needs to face.This paper first studies how to reduce the task execution delay by optimizing the task offloading strategy for the scenario of UAV-assisted edge computing.Secondly,for the scenario of satellite/UAV-assisted edge computing,a task offloading strategy based on multi-agent reinforcement learning is proposed to minimize the weighted sum of delay and energy consumption.The main work of this paper is as follows:(1)For the UAV-assisted edge computing scenario,a partial offloading strategy based on deep reinforcement learning is proposed in the second chapter,aiming at the insufficiency of the full offloading strategy in the current research that cannot fully utilize the system computing resources.Firstly,the coordinates and service range of UAV are defined.Secondly,the delay and energy consumption related to communication and computing under different task offloading strategies are analyzed.Afterwards,a problem model for realizing delay minimization considering UAV maneuverability and its battery capacity constraints is established.For this optimization problem,a partial offloading decision-making scheme based on deep deterministic policy gradient algorithm is proposed.Finally,the simulation results show that,compared with the complete offloading strategy proposed by the benchmark scheme,the strategy proposed in this paper can effectively reduce the delay through task division and UAV trajectory optimization.(2)Different from the task offloading scheme for UAV-assisted edge computing system,a satellite/UAV-assisted edge computing scenario is designed in this paper.Firstly,the delay and user equipment energy consumption generated by satellites and UAVs in the process of providing communication and computing services are defined.Secondly,according to the binary offloading model,a task offloading strategy based on multi-agent deep deterministic policy gradient algorithm is proposed,which optimizes the selection of offloaded nodes by exploring the global state and sharing the actions of all agents.Finally,the simulation results show that the algorithm proposed in this paper can effectively reduce the computational cost of the system.