Prediction-Based Cooperative Cache Discovery in VANETs for Social Networking

Abstract

Social media traffic is increasingly pervading the Internet. Such traffic is mostly generated by mobile devices, which creates immense traffic load on backhaul links in 5G networks. This load can be mitigated by using vehicular networks as a traffic offloading platform, where intelligent vehicles can act as a valuable asset that can bring the data closer to the requester via edge caching. In this paper, we propose the Prediction-Assisted Cooperative Content Discovery (PACD) scheme. PACD exploits the static and mobile nature of parked and moving vehicles, respectively, to promulgate cached content information into the network via bloom filters. PACD is the first scheme that leverages such information to perform cooperative cache discovery to locate closer replicas to the requester by dynamically predicting the location of caching nodes. The Any Relation Clustering Algorithm (ARCA) is employed to cluster trips based on their route similarity using the XXDice similarity coefficient. Each cluster is then trained using the Mixture Transition Distribution-Probit (MTD-Probit) model to predict the remaining trajectory of vehicles. Using these predictions, PACD tracks all possible data providers and ranks them based on their proximity to the requester, as well as their prediction entropy. Extensive evaluations show that PACD yields significant improvements of up to 86%, 30%, 42%, and 16% in terms of delay, packet delivery ratio, cache hit ratio, and prediction accuracy, respectively, compared to prominent caching and prediction schemes in vehicular networks.