Scholarlink

With the rapid development of Connected Autonomous Vehicles (CAVs), various applications requiring high computation resources have emerged. To resolve these issues, the integration of edge computing in CAVs can offer a scalable solution by distributing computation across edge nodes. Using edge computing in CAVs enables each vehicle to make local decisions based on data available at the edge, reducing the need for centralized systems. However, computing on an edge device is still challenging due to its limited computational power and battery life. To address these problems, we first modeled the task offloading issue using the Markov Decision Process (MDP), and then we applied the Proximal Policy Optimization (PPO)-based Deep reinforcement learning (DRL) technique. Extensive simulation results show that compared to the existing RRB and GBA schemes, our proposed DRL-based task offloading technique make significant improvements with reduction in average latency by 26.64, 13.33% and amount of energy consumption by 27.02, 15.44%. Finally, we explored open issues and research directions for task offloading in the CAVs system.

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 for Connected Autonomous Vehicles (CAVs) Task Offloading

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Aditya Bhardwaj

摘要

With the rapid development of Connected Autonomous Vehicles (CAVs), various applications requiring high computation resources have emerged. To resolve these issues, the integration of edge computing in CAVs can offer a scalable solution by distributing computation across edge nodes. Using edge computing in CAVs enables each vehicle to make local decisions based on data available at the edge, reducing the need for centralized systems. However, computing on an edge device is still challenging due to its limited computational power and battery life. To address these problems, we first modeled the task offloading issue using the Markov Decision Process (MDP), and then we applied the Proximal Policy Optimization (PPO)-based Deep reinforcement learning (DRL) technique. Extensive simulation results show that compared to the existing RRB and GBA schemes, our proposed DRL-based task offloading technique make significant improvements with reduction in average latency by 26.64, 13.33% and amount of energy consumption by 27.02, 15.44%. Finally, we explored open issues and research directions for task offloading in the CAVs system.