Scholarlink

<p>This paper presents the Diagonal-Intersection-Routing Protocol with Reinforcement Learning (DIRP-RL), which addresses the problem of dependable packet delivery in Vehicular Ad hoc Networks (VANETs). Using RL, DIRP-RL optimizes routing decisions to reduce latency and improve transmission efficiency by dynamically learning traffic information for specific road segments. The protocol incorporates geographical and off-road forwarding techniques, such as Store-Carry-Forward, to reduce delay and packet loss even in the event of abrupt route changes. In practical VANET applications, performance benchmarks show that DIRP-RL performs better than similar routing protocols as TDRL-RP, APRRL, DRL-ML, and QGrid. In particular, DIRP-RL produces a 98% packet delivery rate, a 5% drop in end-to-end latency, a 95% increase in hop count, a 4% reduction in routing overhead, and a 96% improvement in total performance. These findings demonstrate how reinforcement learning can improve the efficiency and dependability of VANET connectivity.</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

VANETs Assisted Diagonal-Intersection-Routing Using a Reinforcement Learning Approach

  • 期刊论文

  • 发表日期：

    2026年3月12日

查看全文
  • Rajiv Kumar,
  • Srinath Doss,
  • S. Baskar

摘要

This paper presents the Diagonal-Intersection-Routing Protocol with Reinforcement Learning (DIRP-RL), which addresses the problem of dependable packet delivery in Vehicular Ad hoc Networks (VANETs). Using RL, DIRP-RL optimizes routing decisions to reduce latency and improve transmission efficiency by dynamically learning traffic information for specific road segments. The protocol incorporates geographical and off-road forwarding techniques, such as Store-Carry-Forward, to reduce delay and packet loss even in the event of abrupt route changes. In practical VANET applications, performance benchmarks show that DIRP-RL performs better than similar routing protocols as TDRL-RP, APRRL, DRL-ML, and QGrid. In particular, DIRP-RL produces a 98% packet delivery rate, a 5% drop in end-to-end latency, a 95% increase in hop count, a 4% reduction in routing overhead, and a 96% improvement in total performance. These findings demonstrate how reinforcement learning can improve the efficiency and dependability of VANET connectivity.