The integration of Internet of Vehicles (IoV) and edge computing has emerged as an efficient computational paradigm for reducing latency in vehicular applications. By offloading computational tasks to edge servers instead of data centers, vehicles can significantly decrease application response times. However, vehicular mobility may result in scenarios where a vehicle leaves the coverage area of the current base station before the completion of its computational task, rendering it unable to receive the results in a timely manner. To address this challenge, we propose a trajectory-based task offloading decision-making method. By analyzing vehicular trajectories, the method identifies suitable roadside units (RSUs) to enhance the efficiency and reliability of task offloading, thereby mitigating the communication time constraints caused by vehicular movement. Our approach is built upon a deep reinforcement learning (DRL) algorithm, incorporating vehicular trajectory information as a core consideration in the task offloading decision process. Furthermore, we comprehensively account for multidimensional factors such as vehicular resources, network conditions, and task characteristics as inputs to the deep reinforcement learning model. Supported by this information, the model intelligently determines whether tasks should be offloaded to RSUs or processed locally. Simulation experiments conducted in vehicular edge computing scenarios demonstrate that the proposed method effectively reduces the computational latency and significantly improves the efficiency and stability of task processing.
Deep Reinforcement Learning for Efficient Task Offloading in Internet of Vehicles
23.05.2025
2005264 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Adaptive task offloading in V2X networks based on deep reinforcement learning
British Library Conference Proceedings | 2022
|