多低軌衛(wèi)星協(xié)作的邊緣計(jì)算卸載與資源分配策略

打開(kāi)文本圖片集

中圖分類(lèi)號(hào)：TN929.5-34 文獻(xiàn)標(biāo)識(shí)碼：A 文章編號(hào)：1004-373X（2025）17-0007-09

引用格式：,等.多低軌衛(wèi)星協(xié)作的邊緣計(jì)算卸載與資源分配策略[J].現(xiàn)代電子技術(shù),2025,48(17)：7-15.

Edge computing offloading and resource allocation strategy for multi-LEO satellitecollaboration

YANGLiming,ZHOU Yuqian,WANGWenhao,ZHAOHongjun (Schoolfoaidoatigoigsitosdeletiso

Abstract:Low earth orbit (LEO)satelites with widecoveragecan beequipped with mobile edge computing (MEC)server with computing power toprovidemoreeficientcommunicationservices for ground terminalequipment (GTE).Inthispaper,a strategyis proposedonthebasisofconsideringthefactorssuchas mobilityandresource limitationofLEOsatelites,as wellas thecontinuousactionspaceofthecomputingoffoadingproblem.ThecomputingoffoadingproblemistransformedintoaMarkov decisionprocess(MDP)forthejointofloading of multipleLEOsatelltes.Thesystem bandwidth resourcesandcomputing resourcesarealocated.Andten,acomputationalofloadingandresourceallcationstrategybasedondeepdeterministicpolicy gradient (DDPG)isproposed.Inthestrategy，theoveralllatencyofthesystemisoptimizedwhenserving multipleGTEs.The simulationresultsshowthattheproposedstrategycancompletethetaskoffoadingandresourcealocationefectively,andeduce thesystemdelaysignificantly.Theaveragedelayoftheproposedstrategyunderthetotaltaskvolumeof1~1OMBisreducedby about 40.9% ， 35.59% and 30.7% ,respectively,in comparison with the three strategies of no inter-satellite link,using deep Qnetwork and full offloading.

Keywords:LEOsatelite;MEC;task offloading;resource allocation;offloading strategy;deepreinforcementlearning

0 引言

隨著物聯(lián)網(wǎng)新興技術(shù)的發(fā)展，在各種智能設(shè)備上產(chǎn)生了許多延遲敏感、計(jì)算密集的應(yīng)用，例如VR（虛擬現(xiàn)實(shí)）、AR(增強(qiáng)現(xiàn)實(shí)）、各式各樣的游戲，這些應(yīng)用要求很強(qiáng)的計(jì)算能力，如果僅依靠地面終端設(shè)備（GroundTerminalEquipment,GTE)本身的計(jì)算能力很難去滿(mǎn)足用戶(hù)低時(shí)延、高穩(wěn)定性的需求4。（剩余21910字）