Scholarlink

<p>Vehicular Edge Network (VEN) is a dynamic network that utilizes mobile vehicles as nodes and opportunistic data delivery to achieve efficient exchange and sharing of large-scale data. This network has the characteristics of rapid topology changes and rapid data growth. In recent years, opportunity routing has adopted deep reinforcement learning (DRL) to optimize the decision-making process under dynamic network conditions. However, despite the effective improvement of the performance of opportunistic routing algorithms by DRL, it still faces challenges such as slow convergence speed, insufficient real-time link data processing, and weak adaptability to dynamic network environments. In this regard, a Community-aware Routing algorithm based on Deep Neural Network(DNN) and Proximal Policy Optimization (PPO), named CRDP, is introduced. Firstly, DNN is used to divide the community and calculate the node encounter probability, and then PPO algorithm is used to optimize the routing decision. Extensive simulations have been conducted on NS-2 simulator to evaluate the performance of CRDP algorithm and other algorithms. Experimental results show that CRDP algorithm has a significant improvement of 8–20% in the delivery success ratio compared with other algorithms, highlighting its performance advantages.</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

CRDP: Community-Aware Routing Algorithm Based on DNN and PPO

  • 期刊论文

  • 发表日期：

    2025年1月1日

查看全文
  • Huahong Ma,
  • Jingyun You,
  • Honghai Wu,
  • Ling Xing,
  • Xiaohui Zhang

摘要

Vehicular Edge Network (VEN) is a dynamic network that utilizes mobile vehicles as nodes and opportunistic data delivery to achieve efficient exchange and sharing of large-scale data. This network has the characteristics of rapid topology changes and rapid data growth. In recent years, opportunity routing has adopted deep reinforcement learning (DRL) to optimize the decision-making process under dynamic network conditions. However, despite the effective improvement of the performance of opportunistic routing algorithms by DRL, it still faces challenges such as slow convergence speed, insufficient real-time link data processing, and weak adaptability to dynamic network environments. In this regard, a Community-aware Routing algorithm based on Deep Neural Network(DNN) and Proximal Policy Optimization (PPO), named CRDP, is introduced. Firstly, DNN is used to divide the community and calculate the node encounter probability, and then PPO algorithm is used to optimize the routing decision. Extensive simulations have been conducted on NS-2 simulator to evaluate the performance of CRDP algorithm and other algorithms. Experimental results show that CRDP algorithm has a significant improvement of 8–20% in the delivery success ratio compared with other algorithms, highlighting its performance advantages.