Scholarlink

<p>Vehicular Edge Computing Network (VECN) plays an important role in the future intelligent transportation system, providing distributed, multi-level, and fine-grained resources for vehicles by relocating services from cloud down to edge server. However, the diverse types of application services, limited edge node resources, and the unpredictable task demands resulting from high-speed vehicle mobility present VECN with challenges such as irrational service caching decisions, increased service latency, and elevated computational energy consumption. To tackle these challenges, this work introduces a collaborative service caching and computation offloading scheme, called CSC2O, which integrates generative VECN with reinforcement learning to reduce computational latency and energy consumption through accurate task demand prediction. By leveraging the predictive capabilities of Generative Adversarial Networks (GANs), the accuracy of task demand prediction is significantly improved, assisting edge server nodes in optimizing service caching and improving the Edge Offloading Success Rate (EOSR). In addition, to accommodate the dynamic nature of VECN, a hybrid reinforcement learning-based collaborative service caching and computation offloading framework is developed to optimize task offloading decisions. Experimental results show that, in comparison to existing Deep Reinforcement Learning-based algorithms, CSC2O significantly enhances the quality of service (QoS), mitigates the computational latency by 20.9<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, and decreases computational energy consumption by 26.9<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>.</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

CSC2O: Collaborative Service Caching and Computation Offloading Approach Based on GAN-Powered VECN

  • 期刊论文

  • 发表日期：

    2025年5月24日

查看全文
  • Jianhang Liu,
  • Bingyin Jiang,
  • Xuerong Cui,
  • Tingpei Huang,
  • Haibo Wu

摘要

Vehicular Edge Computing Network (VECN) plays an important role in the future intelligent transportation system, providing distributed, multi-level, and fine-grained resources for vehicles by relocating services from cloud down to edge server. However, the diverse types of application services, limited edge node resources, and the unpredictable task demands resulting from high-speed vehicle mobility present VECN with challenges such as irrational service caching decisions, increased service latency, and elevated computational energy consumption. To tackle these challenges, this work introduces a collaborative service caching and computation offloading scheme, called CSC2O, which integrates generative VECN with reinforcement learning to reduce computational latency and energy consumption through accurate task demand prediction. By leveraging the predictive capabilities of Generative Adversarial Networks (GANs), the accuracy of task demand prediction is significantly improved, assisting edge server nodes in optimizing service caching and improving the Edge Offloading Success Rate (EOSR). In addition, to accommodate the dynamic nature of VECN, a hybrid reinforcement learning-based collaborative service caching and computation offloading framework is developed to optimize task offloading decisions. Experimental results show that, in comparison to existing Deep Reinforcement Learning-based algorithms, CSC2O significantly enhances the quality of service (QoS), mitigates the computational latency by 20.9 \(\%\) % , and decreases computational energy consumption by 26.9 \(\%\) % .