Scholarlink

<p>The rapid growth of Vehicular Big Data (VBD) generated by Autonomous Vehicles (AVs) has massive challenges for meeting hard latency requirements, efficient data transmission, and adaptive workload management. Conventional offloading strategies that rely on edge or cloud computing tiers often struggle to adapt in dynamic vehicular environments, where fluctuating workload and network congestion are frequently observed. Moreover, existing approaches focus on single-objective optimization and fail to consider diverse performance metrics and intermediate computing layers. To address these limitations, this paper presents a multi-tier, multi-objective offloading framework using Proximal Policy Optimization (PPO). The proposed approach dynamically offloads VBD tasks across vehicular edge, regional, and cloud layers. It utilizes real-time system feedback, including delay, congestion, cost, CPU utilization, and memory availability, to make adaptive offloading decisions. We introduce a Regional Computing tier, strategically positioned between the edge and cloud, which serves as a latency-aware buffer capable of absorbing excess load and mitigating cloud dependency. To validate the model, we developed <i>RegionalEdgeSimPy</i> simulator. This modular Python simulator enables experimentation with customizable task generation, hierarchical tier modeling, and Artificial Intelligence (AI) scheduling integration. Simulation results demonstrate that the proposed PPO model significantly performed in reducing delay, cost, congestion, CPU utilization, and storage usage while improving load distribution across tiers. The proposed framework demonstrates strong potential for real-time, VBD processing in future Intelligent Transportation Systems (ITS) and vehicular networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Proximal Policy Optimization for Vehicular Big Data Offloading Across Edge, Regional, and Cloud Layers

  • 期刊论文

  • 发表日期：

    2025年11月4日

查看全文
  • Tariq Alsahfi,
  • Afzal Badshah,
  • Raed Alsini,
  • Fouad Shoie Alallah,
  • Wafi Bedewi,
  • Ali Daud

摘要

The rapid growth of Vehicular Big Data (VBD) generated by Autonomous Vehicles (AVs) has massive challenges for meeting hard latency requirements, efficient data transmission, and adaptive workload management. Conventional offloading strategies that rely on edge or cloud computing tiers often struggle to adapt in dynamic vehicular environments, where fluctuating workload and network congestion are frequently observed. Moreover, existing approaches focus on single-objective optimization and fail to consider diverse performance metrics and intermediate computing layers. To address these limitations, this paper presents a multi-tier, multi-objective offloading framework using Proximal Policy Optimization (PPO). The proposed approach dynamically offloads VBD tasks across vehicular edge, regional, and cloud layers. It utilizes real-time system feedback, including delay, congestion, cost, CPU utilization, and memory availability, to make adaptive offloading decisions. We introduce a Regional Computing tier, strategically positioned between the edge and cloud, which serves as a latency-aware buffer capable of absorbing excess load and mitigating cloud dependency. To validate the model, we developed RegionalEdgeSimPy simulator. This modular Python simulator enables experimentation with customizable task generation, hierarchical tier modeling, and Artificial Intelligence (AI) scheduling integration. Simulation results demonstrate that the proposed PPO model significantly performed in reducing delay, cost, congestion, CPU utilization, and storage usage while improving load distribution across tiers. The proposed framework demonstrates strong potential for real-time, VBD processing in future Intelligent Transportation Systems (ITS) and vehicular networks.