为了满足车联网中不同类型的应用对服务质量的不同需求,提出了一种基于深度强化学习的车联网资源分配方案.根据车联网服务对时延和数据传输速率的要求,把车联网服务分成两种切片方式.切片1是要求低时延的V2V链路的服务;切片2是要求高数据传输速率的V2I链路的服务.通过设计动作空间、状态空间、奖励函数,训练出一个可以在车联网动态环境下实时调整,给出当下最优资源分配策略的网络,使得设计出的网络能够在保证切片1时延约束的前提下提高切片2的数据传输速率,以解决基于V2V通信和V2I通信的车联网资源分配问题.
To meet the diverse Quality of Service(QoS)requirements of different applications in the Internet of Vehicles(IoV),a resource allocation scheme is proposed based on deep reinforcement learning.According to the requirements of IoV services on the latency and data transmission rate,the IoV services are categorized into two slicing modes.Slice 1 corresponds to the services on the Vehicle-to-Vehicle(V2V)links with low-latency requirements,while Slice 2 corresponds to the services on the Vehicle-to-Infrastructure(V2I)links with high data transmission rate requirements.By designing the action space,state space,and reward function,a network is trained to dynamically adjust in real-time within the IoV environment to provide the current optimal resource allocation strategy.The designed network can enhance the data transmission rate for Slice 2 while ensuring the latency constraints for Slice 1.This addresses the resource allocation problems in the IoV based on V2V and V2I communications.