Federated Learning-based Vehicle Edge Caching Optimization Strategy

Vehicle edge caching can better meet the real-time computing and data access needs of end users and improve user experience. However, unreasonable cache resource allocation strategy will affect the performance of the vehicle-linked edge cache. This paper proposes a Mobile-aware Edge Caching scheme based on federated learning (MAE-FL). Firstly, the federal learning framework is used to select vehicles according to their dynamic parameters, train local models with historical data requests, upload model parameters to the roadside unit (RSU) for global model aggregation, and accurately predict the popular content in the Internet of vehicles. Then, a moving-aware cache replacement strategy is proposed, which dynamically updates the cache content in the roadside unit according to the real-time mobility and location information of the vehicle to maximize the cache hit rate and reduce the communication overhead. Simulation results show that the proposed scheme is effective in improving cache efficiency and reducing communication cost.