Scholarlink

The Internet of Vehicles (IoV) integrates vehicle communication and IoT technologies to enhance driving safety with low-latency, and support compute-intensive applications such as autonomous driving and traffic management. Mobile Edge Computing (MEC) processes tasks near data sources to reduce latency and power consumption, but the dynamic nature of vehicular networks complicates the resource allocation required by data sources. Deep Reinforcement Learning (DRL) offers a promising solution, managing complex state spaces effectively. This systematic mapping study reviews 16 studies published between 2013 and 2024 on DRL-based resource allocation in IoV. It categorizes the algorithms, evaluation criteria, and software environments used, highlighting the advantages of DRL over traditional methods. Techniques like Deep Q-Learning (DQL) and Deep Deterministic Policy Gradient (DDPG) have shown strong performance in various IoV scenarios. The study recommends future research in areas such as developing sophisticated algorithms, integrating with 5G networks, and exploring federated learning to support decentralized training and preserve data privacy. This comprehensive overview of DRL applications in IoV resource allocation provides valuable insights for future advancements and practical implementations in this emerging field, aiming to enhance the efficiency and performance of IoV systems.

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

Deep Reinforcement Learning-Based Resource Allocation for the Internet of Vehicles: A Systematic Mapping Study

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ramzi Bouzoubia,
  • Sofiane Zaidi,
  • Lazhar Khamer,
  • Abdelhabib Bourouis,
  • Mostafa Ogab

摘要

The Internet of Vehicles (IoV) integrates vehicle communication and IoT technologies to enhance driving safety with low-latency, and support compute-intensive applications such as autonomous driving and traffic management. Mobile Edge Computing (MEC) processes tasks near data sources to reduce latency and power consumption, but the dynamic nature of vehicular networks complicates the resource allocation required by data sources. Deep Reinforcement Learning (DRL) offers a promising solution, managing complex state spaces effectively. This systematic mapping study reviews 16 studies published between 2013 and 2024 on DRL-based resource allocation in IoV. It categorizes the algorithms, evaluation criteria, and software environments used, highlighting the advantages of DRL over traditional methods. Techniques like Deep Q-Learning (DQL) and Deep Deterministic Policy Gradient (DDPG) have shown strong performance in various IoV scenarios. The study recommends future research in areas such as developing sophisticated algorithms, integrating with 5G networks, and exploring federated learning to support decentralized training and preserve data privacy. This comprehensive overview of DRL applications in IoV resource allocation provides valuable insights for future advancements and practical implementations in this emerging field, aiming to enhance the efficiency and performance of IoV systems.