A Hybrid Proactive Caching System in Vehicular Networks Based on Contextual Multi-Armed Bandit Learning

Proactive edge caching has been regarded as an effective approach to satisfy user experience in mobile networks by providing seamless content transmission and reducing network delay. This is particularly useful rapidly changing vehicular networks. paper addresses the proactive (at roadside unit (RSU)) problem mobility prediction, i.e., next RSU prediction. Specifically, proposes a distributed Hybrid cMAB Caching System where RSUs act independent learners that implement two parallel online reinforcement learning-based prediction algorithms between which they can adaptively finalize their predictions for RSU. The are based on xmlns:xlink="http://www.w3.org/1999/xlink">Contextual Multi-armed bandit (cMAB) learning, called xmlns:xlink="http://www.w3.org/1999/xlink">Dual-context cMAB xmlns:xlink="http://www.w3.org/1999/xlink">Single-context . hybrid system further developed into variants: xmlns:xlink="http://www.w3.org/1999/xlink">Vehicle-Centric xmlns:xlink="http://www.w3.org/1999/xlink">RSU-Centric In addition, also conducts comprehensive simulation experiments evaluate performance of proposed system. They include three traffic scenarios: xmlns:xlink="http://www.w3.org/1999/xlink">Commuting traffic, Random traffic xmlns:xlink="http://www.w3.org/1999/xlink">Mixed Las Vegas, USA Manchester, UK. With different road layouts urban areas, aims generalize application Simulation results show Vehicle-Centric reach nearly 95% cumulative accuracy Commuting scenario outperform other methods used comparison reaching 80% Mixed scenario. Even completely scenario, it guarantees minimum 60%.