【作者】 马欣；

【导师】 王晓轩；

【作者基本信息】 北京交通大学 ， 信息与通信工程， 2024， 硕士

【摘要】 在车联网移动边缘计算中,车辆通过计算卸载的方式,将任务卸载到边缘节点进行处理,从而缓解其计算压力。车联网中的一些任务与道路安全息息相关,在进行计算卸载时,必须确保这些任务准确高效的完成,以避免引发安全风险。为了确保计算卸载的顺利进行,系统可靠性成为了不可忽视的重要因素。车联网系统的可靠性直接关系到任务完成的准确性和稳定性,进而影响道路安全和用户体验。因此,实现车联网移动边缘场景的可靠卸载对于保障道路安全、提升服务质量具有重要意义。本研究主要通过计算卸载策略的设计和边缘节点动态部署策略的设计来实现车联网移动边缘场景的可靠卸载,具体研究内容如下:(1)针对可靠卸载中任务的多样性、系统的动态性等问题,本研究提出了基于软演员评论家(Soft Actor Critic,SAC)的两阶段优先级卸载策略,以提高系统可靠性。首先,本研究考虑到任务的重要性和紧迫性,提出了一种优先级评估方法用以评估任务的优先级。基于优先级,本研究设计了初始卸载策略,用于周期性初始的一次性高效任务卸载。然后,本研究设计了动态队列插入卸载策略,用于执行初始卸载策略期间陆续到达任务的卸载。接着,对系统可靠性进行建模,并将其作为目标函数,将卸载问题建模为最优化问题,采用SAC算法解决。最后,针对初始卸载策略和动态队列插入卸载策略分别进行了仿真。结果表明,初始卸载策略在可靠性方面优于其它卸载策略,采取动态队列插入的卸载策略相较于无动态队列插入的卸载策略对系统可靠性有明显提升。(2)针对可靠卸载中路边单元(Road Side Unit,RSU)覆盖不足,RSU部署难以满足时变的服务要求等问题,本研究提出了基于异步优势演员评论家(Asynchronous Advantage Actor Critic,A3C)的边缘节点动态部署策略。首先,本研究采用公共汽车作为移动路边单元(Mobile-Road Side Unit,m-RSU)填补覆盖空白,并提出了基于路径相似度的快速m-RSU选择算法,以用于RSU覆盖空白地区的m-RSU接入选择。然后,设计了基于SAC的边缘节点动态部署策略,通过调整f-RSU(Fixed-Road Side Unit,f-RSU)和m-RSU的状态,优化系统中RSU的部署以最大化系统可靠性并最小化能耗。接着,将部署问题建模为一个最大化系统可靠性,最小化能耗的双目标优化问题,并采用A3C算法进行解决。最后,对所提出的动态部署策略进行仿真评估。结果显示,该策略部署效果优于有m-RSU协助的边缘节点非动态部署策略。除此之外,有m-RSU协助的策略在部署效果方面优于无m-RSU协助的策略。更多还原

【Abstract】 In vehicular edge computing,vehicles alleviate their computational burden by offloading tasks to edge nodes,easing their computational load.Some tasks in vehicular networks are closely related to road safety,and when performing task offloading,it is essential to ensure the accurate and efficient completion of these tasks to avoid safety risks.Therefore,system reliability becomes a crucial factor to ensure smooth task offloading.The reliability of vehicular network systems directly affects the accuracy and stability of task completion,thereby impacting road safety and user experience.Hence,achieving reliable offloading in vehicular edge scenarios is vital for safeguarding road safety and improving service quality.This study primarily focuses on designing task offloading strategies and dynamic deployment strategies for edge nodes to achieve reliable offloading in vehicular edge scenarios.The specific research contents are as follows:(1)To address task diversity and system dynamism,this study proposes a two-stage priority offloading strategy based on Soft Actor Critic(SAC)to enhance system reliability.First,a priority evaluation method is developed to assess task importance and urgency.Based on these priorities,an initial offloading strategy is designed for efficient one-time task offloading at the start of each cycle.Then,a dynamic queue insertion strategy is created for offloading tasks that arrive during the execution of the initial strategy.System reliability is modeled as the objective function,and the offloading problem is formulated as an optimization problem,solved using the SAC algorithm.Simulations for both the initial and dynamic queue insertion strategies show that the initial strategy is more reliable than others,and the dynamic queue insertion strategy significantly improves system reliability compared to not using dynamic queue insertion.(2)To address RSU coverage gaps and the challenges of meeting time-varying service demands,this study proposes a dynamic edge node deployment strategy based on Asynchronous Advantage Actor Critic(A3C).Buses are used as mobile RSUs(m-RSUs)to fill coverage gaps,and a fast m-RSU selection algorithm based on path similarity is proposed for m-RSU access in areas with RSU gaps.The A3C-based dynamic deployment strategy adjusts the states of fixed RSUs(f-RSUs)and m-RSUs to optimize their deployment,maximizing system reliability and minimizing energy consumption.The deployment problem is formulated as a bi-objective optimization problem,solved using the A3 C algorithm.Simulations show that this dynamic deployment strategy outperforms the static deployment strategy with m-RSU assistance.Moreover,the strategy with m-RSU assistance is more effective than the one without m-RSU assistance.更多还原