Federated Learning and Edge Collaboration Empowered Service Provision Method for Mobile Terminals

Abstract

With the rapid development of 5G technology and mobile edge computing (MEC) technology, it is possible to realize intelligent assisted operating based on artificial intelligence technology in the mobile terminal scene. To solve the problem of high service response delay, this research studies a federated learning and edge collaboration empowered content service provision method, which effectively reduces latency and improves service efficiency for vehicles by deploying services on the edge of the network closer to the terminal. First, we construct the Edge Collaborative Caching Domains (ECCDs) to allow edge nodes (ENs) with similar features and meeting the geographical constraint serve to the vehicles jointly. Then, we make user behavior prediction with long short-term memory networks (LSTMs) based on federated learning framework to analyze the user’s future preferences under the condition of ensuring user privacy. Furthermore, based on user behavior analysis, we also propose a content pre-caching strategy based on Deep Q-Network to improve the hit rate. The simulation results show that comparing to the mechanism without user behavior prediction and pre-cache deployment the mechanism proposed in this project can improve the hit rate, time delay and resource utilization.