中图分类号：TN915.01 文献标志码：A 文章编号：1000-582X（2025）08-086-13

Design and evaluation of an offload feedback strategy framework based on a service characteristic model in the internet of vehicles

WANG Shi， CAO Dayanab， ZHU Xiaoyingab， WANG Mingyua， WANG Haoyinga （a.SchoolofElectronic and Information Engineering; b.Institute of Graduate，Liaoning Technical University， Huludao，Liaoning，P.R. China）

Abstract： With the proliferation of diverse service characteristics in the internet of vehicles （IoV）under the mobile edge computing （MEC） paradigm，evaluating server-to-end transmission performance presents a significant challenge，particularly due to the complex modeling requirements that must account for service-specific traits in offload feedback strategies.To address this，acache scheduling evaluation framework is proposed，incorporating time-varying and multi-type services based on queuing theory and a Markov-modulated service procesThe proposed framework supports flexible adjustments to service characteristics，bidirectional procesing rates，and offload feedback strategies，enabling itto adapt to various communication environments.Within this framework， an offload feedback strategy based onstatistical prediction is proposed.Numerical simulationsshow that the the proposed strategy improves transmission performance by approximately 50% compared with traditional approaches. These findings indicate that the proposed framework provides a valuable reference for designing adaptive strategies under diverse network conditions and hardware configurations.

Keywords： internet of vehicles; mobile edge computing; sevice modeling; Markov-modulated services; resource allocation

近年来，随着无线通信和物联网技术的发展，车联网（internet ofvehicles，IoV）已成为5G的重要应用场景。在移动边缘计算技术背景（mobile edge computing，MEC）下，车联网中路边单元携带的MEC服务器和智能车辆配备的车载单元（onboard unite，OBU）都具备计算和存储能力l。基于MEC系统计算和缓存的功能，学者提出服务缓存和边缘缓存技术并衍生了任务卸载、资源分配等研究。在网络层上，任务卸载和资源分配问题被建模为最优化问题。针对最优化模型，张建军等4提出一种多MEC联合卸载的方案，李方伟等[5提出了V2X（vehicle-to-everything）协同缓存与资源分配机制。由于引人多样化业务模型会使最优化模型出现计算成本高的问题，上述研究在完成资源分配时未考虑到业务相关性和优先级等多样化特征。然而，不同类型业务的网络需求和处理方式并不相同。例如业务为时延敏感和上下文敏感的应用程序，则应卸载到MEC服务器，其他为安全性服务的重要业务应该在本地进行服务和保存。

目前车联网的典型业务包括：驾驶安全、交通效率、信息服务和管理综合4类业务。这些业务在网络端可定义为流量特性、可靠实时特性、忙时特性、移动性、触发特性和附着特性的量化组合]。但相关研究.]尚缺少针对车联网中车辆业务通信的系统建模，都是在网络层通过多要素预测单一业务特性的变化。因此，对设计任务卸载策略而言，建立业务模型呈现车联网中多种业务特性对边缘计算效率的影响有重要意义。……