Mobility-Aware Cooperative Caching in Vehicular Edge Computing Based on Asynchronous Federated and Deep Reinforcement Learning

Vehicular edge computing (VEC) can learn and cache most popular contents for vehicular users (VUs) in the roadside units (RSUs) to support real-time applications. Federated learning (FL) protect VUs' privacy by sharing vehicles' local models instead of data. In traditional FL, global model is periodically updated aggregating all models. However, vehicles may frequently drive out coverage area VEC before they finish training thus FL cannot upload as expected, which would degrade accuracy model. The asynchronous be performed without models, more uploaded improve vehicle mobility significantly impacts FL. There no published work considering design cooperative caching based on addition, capacity RSU limited size predicted usually exceeds RSU. Hence, should different RSUs while content transmission delay. this paper, we consider propose a scheme federated deep reinforcement (CAFR) predict further obtain optimal location contents. Extensive experimental results have demonstrated that CAFR outperforms other baseline schemes.