物联网学报2025，Vol.9Issue(2)：39-50,12.DOI:10.11959/j.issn.2096-3750.2025.00490
PD-TD3:高速公路场景下边路协同计算卸载策略
PD-TD3:edge-cloud collaborative computation offloading strategy in highway scenarios
摘要
Abstract
In the highway scenarios,existing offloading models often overlook the network dynamics caused by the high-speed movement of vehicles,leading to increased latency and energy consumption,and exhibit insufficient effectiveness in reducing latency and energy consumption.To address these challenges,an offloading strategy utilizing the prioritized double-buffer pool experience replay twin delayed deep deterministic policy gradient(PD-TD3)algorithm was proposed.Initially,a three-layer distributed offloading model tailored for highway environments was developed.Subsequently,the computation offloading problem was formulated as a Markov decision process(MDP),with the reward function designed to optimize the trade-off between latency and energy consumption,aiming to maximize the reward.To address the limita-tions of the traditional TD3 algorithm,including slow convergence,Q-value underestimation bias,and inefficient experi-ence sampling,the PD-TD3 algorithm was introduced to solve the optimization problem.Simulation results indicate that,compared with the TD3 algorithm,the PD-TD3 algorithm can effectively improve the efficiency of early algorithm explo-ration and effectively reduces computation offloading latency by approximately 50%and energy consumption by about 70%.关键词
移动边缘计算卸载/深度强化学习/智能车辆/边路协同/时延/能耗Key words
mobile edge computing offloading/deep reinforcement learning/intelligent vehicle/side-lane synergy/time delay/energy loss分类
交通工程引用本文复制引用
刘毅,杨琪,李国燕,何军,张明辉..PD-TD3:高速公路场景下边路协同计算卸载策略[J].物联网学报,2025,9(2):39-50,12.基金项目
国家自然科学基金资助项目(No.61876131) （No.61876131）
天津市研究生科研创新资助项目(No.2022SKYZ391)The National Natural Science Foundation of China(No.61876131),Tianjin Research and Innovation Project for Postgraduate Students(No.2022SKYZ391) （No.2022SKYZ391）
