Research On Mobility-aware Cooperative Caching In Vehicular Content Centric Network

Vehicular Content Centric Network is an architecture that applies the concept of content center network to vehicular network environment.As its basic feature,in-network caching makes the sharing of cached data among multiple nodes an efficient way of data access.However,due to the mobility of vehicles and the sparsity of connections,lack of coordination between vehicles leads to high redundancy of cache content and poor utilization of cache space.Road Side Unit(RSU)can only passively process requests from vehicles,and is hard to actively provide services,therefore can not make full use of caching ability.In order to solve the problems above and improve the efficiency of caching service,Cooperative Caching based on Mobility Prediction and Consistent Hashing(CCMPCH)is firstly designed.Prediction by Partial Matching(PPM)is adopted to forecast each vehicle’s path,and vehicles with similar mobile characteristics are clustered.In each cluster,the consistent hashing algorithm is used to achieve content allocation on the cooperative nodes,and a cache replacement policy based on popularity is proposed which improves the priority of cooperative content by preference constant.For further improving the user experience,a scheme named Pre-Caching based on Federated Learning(PCFL)is designed to better activate the service ability of RSU,where Deep Reinforcement Learning is used to model and learn the RSU cache environment and the agent deployed on RSU makes the cache decision.Further,federated learning architecture is used to aggregate the models trained by RSU on remote server.To mitigate neighboring RSU cache duplication,a mechanism is designed based on Q values.This paper evaluates the proposed cooperative caching scheme and RSU pre-caching scheme through a large number of simulations.According to the cache hit rate,content access delay,hop count and universal cached content number,the two schemes are analyzed comprehensively under different network environment configuration.Simulation results show that the two schemes proposed in this paper can effectively improve the performance of cache service and user experience compared with other state-of-the-art schemes.