基于任务抢占的VEC计算卸载与缓存优化方法

打开文本图片集

关键词：车载边缘计算；计算卸载；内容缓存；任务抢占；改进异步深度强化学习中图分类号：TP929.5 文献标志码：A 文章编号：1001-3695（2025）10-029-3122-07doi：10.19734/j. issn.1001-3695.2025.04.0083

Tsk preemption-sed computtionl offloding nd cche optimiztion method in VEC

Shi Jie a，b† ，Tang Honga，ʰ，Liu Zixingaʰ，Yan Xingruia，b （a.SchoolofComuications&IforationEnginering，.ChongqingKeyLaboratoryfobileCommunictionsTecologyChonging University of Posts& Telecommunications，Chongqing 40o065，China）

Abstract：Withtherapidadvancementof intelligenttransportationsystems，challengessuchasunevencomputingresourceallocationand task latencysensivity in vehicularedge computing have become increasinglyprominentHowever，staticresource allcationmethods inexisting studiesareinsuffcientfordynamicvehicularenvironments.Toaddress thisisse，this paper proposedataskpreemption-based jointoptimization frameworkforcomputationofloading andcaching，enabling elasticcache expansionandeficienttask scheduling.Firstly，thispaper establishedacollborativecomputing modelformultiplevehicles androadsideunit，considering task priorityandutilizingidleresourcesfromparkedvehicles.ItintroducedaZipfdistributionbasedcaching strategy toimprovecontent placement.Thenthis paper integratedatask preemptionmechanism toprioritizeurgenttasksandallviateRSUoverload.Subsequently，this paper formulateda jointoptimizationproblemto minimizeaverage systemdelaybyoptimizingofloadingandcaching strategies.Astheproblem was high-dimensionalandnonlinear，this paper proposedadeepreinforcementlearning based improvedasynchronousadvantageactor-criticalgorithm.Simulationresultsdemonstratethattheproposedmethod significantlyreduceslatencyand improves task completionrates，enhancinguserexperience in practical vehicular environments.

Key words：vehicular edge computing（VEC）；computation ofloading；content cache；task premption；improved asynchronous deep reinforce learning

0 引言

随着物联网技术的不断发展，移动边缘计算（mobileedgecomputing，MEC）已广泛应用于智能家居、智慧医疗、新能源车辆等对实时性敏感的领域[1]。（剩余16736字）