Resource Scheduling And Allocation Based On Deep Learning For Internet Of Vehicles

The advancement of information technology has driven the rapid development of society.The Internet of Vehicles(Io V)is an emerging technology that has developed rapidly in recent years.It combines the Internet,wireless communication technology,and vehicle electronic technology to enable the exchange and intelligence of vehicles through the interconnection of information between vehicles and between vehicles and roadside facilities.This greatly improves the efficiency and safety of the transportation system.However,due to the high mobility of vehicle radio terminals and the combined effects of some surrounding scatterers,the channels of vehicle-to-vehicle and vehicle-to-infrastructure communications fluctuate greatly in transmission bandwidth and frame duration which leads to multipath effect and doppler effect of wireless communication in vehicular networks,and all these factors can negatively affect vehicle communication.Resource scheduling techniques of vehicular networks focus on improving the transmission quality of communication links by starting from the communication regime,such as modulation selection and load length allocation.Accordingly,this paper combines channel estimation technology,adaptive modulation and coding technology,and effective payload allocation in the research direction of resource scheduling for the Io V,and proposes a deep learning-based adaptive vehicle network resource scheduling algorithm.The algorithm corrects the channel data collected through the receiver using spectral temporal averaging channel estimation techniques.After extracting the channel state characteristics through the deep learning algorithm,it is classified and fed back to the vehicle transmitter.The transmitter then decides on the preset modulation and coding and effective payload length selection based on these classifications,further improving the communication performance between vehicles.On the other hand,vehicles have the characteristics of high-frequency,group sending,group receiving,and periodic communication.The limited system capacity in the area leads to serious conflicts in the allocation of wireless resources for the Io V.In addition,road and driving safety require extremely strict low latency and high reliability communication.Ensuring these safety measures while improving user experience is also crucial for the future development of the industry.Our work transforms these difficulties into two specific indicators: the total throughput of entertainment data transmission and the success rate of periodic safety information transmission.Based on the Manhattan grid layout of the urban virtual environment model,this paper proposes a resource allocation method based on multi-agent deep reinforcement learning that combines Dueling DQN and Double DQN with low-dimensional feature fingerprints and soft update architecture.The algorithm takes the communication link between vehicles as an agent and reuses the spectrum resources of the link between vehicles and infrastructure,maintaining a good total throughput of entertainment data transmission while ensuring a high success rate of periodic safety information transmission.