LAUSR

Social media traffic constitutes the highest percentage of Internet traffic. Such traffic is largely facilitated by mobile devices, which imposes a huge traffic load on backhaul links in 5G networks, and can in turn affect the quality of service. This traffic load can be alleviated by using vehicular networks as a traffic offloading platform. In particular, vehicles can act as a resourceful asset for edge caching, thus enabling data acquisition from nearby caching nodes rather than the remote backhaul servers. In this paper, we propose the Predictive Proactive Caching Framework (PPCF), which exploits the daily driving routine and predictable behavior of users to pre-cache the data at parked vehicles for the requesters to proactively acquire as they pass by. PPCF is composed of a cache placement module and a prediction module. The former aims at maximizing cache hits by assigning replicas to caching spots that yield maximum certainty in their spatiotemporal availability for requesters. To estimate such availability, the prediction module aims at making accurate travel time predictions by proposing the use of a Long Short-Term Memory (LSTM) network trained using particle swarm optimization (PSO-LSTM). The predicted average travel time is then used to estimate a personalized travel time for users by considering their different driving behaviors. Extensive simulations demonstrate the ability of the proposed framework to achieve significant improvements in its targeted objectives compared to other prominent caching and prediction schemes in vehicular networks.