Scholarlink

<p>Reinforcement learning (RL) has emerged as a promising paradigm for enabling intelligent and adaptive resource management in future vehicular networks. This paper surveys recent advancements in RL-based optimization schemes across multiple domains, reflecting the evolving architecture and demands of intelligent transportation systems. RL-based resource allocation strategies are reviewed in the context of vehicular networks, including centralized, decentralized, and hybrid coordination models, while applications in platoon driving and traffic signal control are examined for their potential to optimize traffic flow. The survey also highlights RL-enabled scheduling and routing techniques designed to enhance communication efficiency in highly dynamic environments, as well as task offloading and content caching in vehicular edge computing to meet next-generation network requirements. In addition, RL-based resource sharing frameworks in the internet of vehicles and unmanned aerial vehicles are explored, underscoring their role in extending network capacity and resilience. Despite these advances, real-world deployment remains constrained by gaps such as safety guarantees, robustness under heterogeneous conditions, limited standardization, and the computational overhead of real-time training. Addressing these challenges within the context of 6G networks, edge intelligence, and multi-access edge computing integration is crucial to bridging the gap between theoretical potential and practical implementation. This survey synthesizes current trends, identifies open research challenges, and outlines future directions, offering valuable insights for developing scalable, efficient, and reliable RL-based solutions for intelligent 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

Reinforcement Learning-Based Resource Allocation for Smart Vehicular Networks: A Review

  • 期刊论文

  • 发表日期：

    2026年1月12日

查看全文
  • Edwin M. Kataka,
  • Thomas O Olwal,
  • Karim Djouani,
  • Prosper Z Sotenga

摘要

Reinforcement learning (RL) has emerged as a promising paradigm for enabling intelligent and adaptive resource management in future vehicular networks. This paper surveys recent advancements in RL-based optimization schemes across multiple domains, reflecting the evolving architecture and demands of intelligent transportation systems. RL-based resource allocation strategies are reviewed in the context of vehicular networks, including centralized, decentralized, and hybrid coordination models, while applications in platoon driving and traffic signal control are examined for their potential to optimize traffic flow. The survey also highlights RL-enabled scheduling and routing techniques designed to enhance communication efficiency in highly dynamic environments, as well as task offloading and content caching in vehicular edge computing to meet next-generation network requirements. In addition, RL-based resource sharing frameworks in the internet of vehicles and unmanned aerial vehicles are explored, underscoring their role in extending network capacity and resilience. Despite these advances, real-world deployment remains constrained by gaps such as safety guarantees, robustness under heterogeneous conditions, limited standardization, and the computational overhead of real-time training. Addressing these challenges within the context of 6G networks, edge intelligence, and multi-access edge computing integration is crucial to bridging the gap between theoretical potential and practical implementation. This survey synthesizes current trends, identifies open research challenges, and outlines future directions, offering valuable insights for developing scalable, efficient, and reliable RL-based solutions for intelligent vehicular networks.