Research On Edge Cooperative Caching Strategy Based On The Internet Of Vehicles

With the increasingly mature urban information network and the rapid growth of vehicle ownership,many cutting-edge application technologies and software services in the Internet of Vehicles may rely on superior computing,communication and storage resources,which have relatively strict requirements on content access delay and network bandwidth.Under such a trend,the research on edge cooperative caching strategy based on the Internet of Vehicles can shorten the system content access delay,alleviate the load of the backhaul link and improve the caching efficiency,so as to build up the performance of the entire network.However,the existing edge caching strategies in the Internet of Vehicles are still limited to too many ideal conditions,making it difficult to deploy the caching system in the Internet of Vehicles.The traditional cooperative caching strategy relies heavily on global information,is difficult to adapt to the dynamic environment of the Internet of Vehicles and lacks the pertinence of vehicles,and the centrally deployed intelligent caching strategy tends to waste too much network resources when dealing with repetitive tasks.This paper proposes a decentralized edge cooperative caching strategy to further improve the performance and ubiquity of the caching system deployed in the Internet of Vehicles.The main contributions are as follows:1.By integrating the federated learning framework into the edge cooperative caching strategy based on deep reinforcement learning,and utilizing t he opportunistic vehicle-to-vehicle communication and data sharing formed by the concept of vehicle social network,a decentralized edge cooperative caching strategy based on social awareness is proposed.The strategy further solves the problem of minimizing longterm average content access latency.In addition,the caching system can achieve faster global model training by placing the original training data locally in the vehicle,and can also protect the privacy information of vehicle users.At the same t ime,it integrates a variety of in-vehicle communication methods,which alleviates the problem of duplicate content caching.2.By applying the attention mechanism to the weighted aggregation of different local models,and using the deep deterministic policy gradient algorithm to dynamically control the decision process of caching node selection and caching update based on vehicle performance and local historical data,an attention-weighted decentralized edge cooperative caching strategy is proposed.The strategy trains deep deterministic policy gradient models in a decentralized approach and saves the training data locally on the vehicle,which fully addresses the problem of dealing with continuous action spaces and model aggregation between heterogeneous vehicles.The attention mechanism is used to control the weights of different local models,which avoids the problem of invalid aggregation due to the difference precision of local models.