Joint Content Placement and Storage Allocation Based on Federated Learning in F-RANs

Area of Science:

• Computer Science
• Wireless Communication Networks
• Artificial Intelligence

Background:

• Explosive growth in wireless data traffic necessitates efficient network architectures like Fog Radio Access Networks (F-RAN).
• Content caching in F-RAN reduces network traffic and latency but requires accurate popularity prediction due to cache capacity limits.
• Traditional prediction methods raise privacy concerns by centralizing user data.

Purpose of the Study:

• To propose an intelligent F-RAN framework using federated learning (FL) for privacy-preserving user demand prediction.
• To develop an integrated model for optimizing storage resource allocation and content placement to minimize network traffic costs.
• To address the computational complexity of the optimization problem with efficient heuristic algorithms.

Main Methods:

• Implementation of a federated learning (FL) framework for decentralized user demand prediction.
• Formulation of an Integer Linear Programming (ILP) model for integrated resource allocation and content placement.
• Design of two heuristic algorithms to solve the computationally complex ILP problem.

Main Results:

• The federated learning approach accurately predicts content popularity distribution without centralizing user data, ensuring privacy.
• The integrated model effectively optimizes storage and content placement to minimize network traffic costs.
• Simulation results demonstrate that the heuristic algorithms achieve near-optimal performance.

Conclusions:

• The proposed FL-based F-RAN framework offers a privacy-preserving solution for content caching and network optimization.
• The developed heuristic algorithms provide efficient methods for solving complex resource allocation problems in F-RAN.
• This approach enhances network performance by reducing traffic and latency while safeguarding user privacy.