车载云计算系统中资源分配的优化方法  

Optimization Method of Resource Allocation in Vehiclular Cloud Computing System

作　　者：董晓丹[1,2] 吴琼 DONG Xiao-dan;WU Qiong(Jiangsu vocational college of information technology,Wuxi 214153,China;Jiangnan University,Wuxi 214122,China)

机构地区：[1]江苏信息职业技术学院,江苏无锡214153 [2]江南大学,江苏无锡214122

出　　处：《中国电子科学研究院学报》2020年第1期92-98,共7页Journal of China Academy of Electronics and Information Technology

基　　金：国家自然科学基金(61701197);江苏省高职院校教师专业带头人高端研修项目(2019GRGDYX049);江苏信息职业技术学院重点科研课题(JSITKY201901)。

摘　　要：随着车联网(IoV)应用服务的发展,提升网络的任务卸载能力成为满足用户服务需求的关键。文中针对动态场景中车辆计算资源共享问题,提出了车载云计算(VCC)系统的最优计算资源分配方案,以实现任务卸载能力的提升。该方案将VCC系统中任务卸载的收入(节省的功耗、处理时间和转移成本的量化值)和开销(预期的任务处理开销)的差值作为系统奖励值,将最优分配问题转化为求解长期预期奖励的最大值问题。进一步将问题表述为无限时域半马尔可夫决策过程(SMDP),定义和分析了状态集、动作集、奖励模型以及转移概率分布,并考虑忙碌(Busy)车辆离开的情况,称为B-SMDP的解决方案。最后,仿真结果表明,与模拟退火算法(SA)和贪婪算法(GA)相比,B-SMDP方案有较明显的性能提升。With the development of Internet of Vehicle( IoV) application services,improving the offloading ability of network tasks has become the key to satisfying user service needs. Aiming at solving the problem of vehiclular computing resource sharing in dynamic scenarios,this paper proposes an optimal computing resource allocation scheme for vehiclular cloud computing( VCC) system to improve the task offloading capability. This solution uses the difference between the revenue( Quantified value of power savings,processing time,and transfer costs) and the overhead( expected task processing overhead) of the task offload in the VCC system as the system reward value,and converts the optimal allocation problem into solving the problem of maximizing the long-term expected rewards. The problem is further expressed as an infinite time domain semi-Markov decision process( SMDP). The state set,action set,reward model,and transition probability distribution are defined and analyzed,and the case of a busy vehicle leaving is considered,we name the proposed solution as B-SMDP solution. Finally,simulation results show that compared with the simulated annealing algorithm( SA) and greedy algorithm( GA),the B-SMDP solution has a significant performance improvement.

关 键 词：车载云计算 半马尔可夫决策过程 忙碌车辆 资源分配 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] TN915[自动化与计算机技术—计算机科学与技术]