Scholarlink

With the evolution of 6G technology and Internet of Vehicles (IoV), automobiles have become increasingly interconnected and intelligent. Advanced autonomous decision-making in intelligent vehicles, given limited computing and communication resources, as well as strict task deadlines, demands attention to balancing energy consumption and latency of computing tasks. To address these challenges, this paper proposes a vehicle edge computing (VEC) task offloading model based on deep reinforcement learning (DRL) for multiple vehicles within the VEC system, targeting partial computation offloading schemes. Due to constraints in vehicle computing power, timely task completion may be limited, and such tasks can be offloaded to base station (BS) VEC servers with more robust computing capabilities. To enhance convergence efficiency and achieve superior system performance, we introduce the DRL algorithm based on the Actor-Critic framework to expedite model training. Simulation results indicate that our proposed Actor-Critic-based DRL algorithm effectively accelerates convergence, improves performance, and reduces the overall system cost.

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 Task Offloading for 6G-Enabled Intelligent Vehicles

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Bingxin Wang,
  • Dan Tu,
  • Jie Wang

摘要

With the evolution of 6G technology and Internet of Vehicles (IoV), automobiles have become increasingly interconnected and intelligent. Advanced autonomous decision-making in intelligent vehicles, given limited computing and communication resources, as well as strict task deadlines, demands attention to balancing energy consumption and latency of computing tasks. To address these challenges, this paper proposes a vehicle edge computing (VEC) task offloading model based on deep reinforcement learning (DRL) for multiple vehicles within the VEC system, targeting partial computation offloading schemes. Due to constraints in vehicle computing power, timely task completion may be limited, and such tasks can be offloaded to base station (BS) VEC servers with more robust computing capabilities. To enhance convergence efficiency and achieve superior system performance, we introduce the DRL algorithm based on the Actor-Critic framework to expedite model training. Simulation results indicate that our proposed Actor-Critic-based DRL algorithm effectively accelerates convergence, improves performance, and reduces the overall system cost.