An Efficient Cooperative Active Caching Strategy in Vehicular Edge Network Based on Asynchronous Federated Learning
Pang Liu, Zhuofan Liao, Xiaoyong Tang · 2024
Edge caching is a promising technique for effectively reducing backhaul pressure and content access latency in the Internet of Vehicles (IoV). However, the high mobility of vehicles and dynamic user requests often lead to outdated cached content. Expired cache wastes the limited storage space and transmission in vehicular edge computing. Improving the cache hit rate is an effective approach to address these issues. In this work, we propose a Cooperative Active Caching Strategy (CACS) which works as follows. First, user preferences are analyzed from the historical content data of vehicle users. After that, multiple vehicles cooperate to learn the global model under an asynchronous federated learning framework, and get the content popularity from user preferences and content features. To explore the problem of local models being outdated in asynchronous federated learning, CACS integrates model compression algorithms, enhancing system efficiency and prediction accuracy. Finally, a heuristic cooperative caching content placement algorithm is proposed based on a greedy policy to minimize average access latency. Simulation results show that the CACS can improve the cache hit rate by 15% at most compared to existing state-of-the-art caching strategies.