Distributed Reinforcement Learning for Privacy-Preserving Dynamic Edge Caching

Mobile edge computing (MEC) is a prominent paradigm which expands the application fields of wireless communication. Due to limitation capacities user equipments and MEC servers, caching (EC) optimization crucial effective utilization resources in MEC-enabled networks. However, dynamics complexities content popularities over space time as well privacy preservation users pose significant challenges EC optimization. In this paper, privacy-preserving distributed deep deterministic policy gradient (P2D3PG) algorithm proposed maximize cache hit rates devices Specifically, we consider fact that are dynamic, complicated unobservable, formulate maximization on problems under constraints preservation. particular, convert optimizations into model-free Markov decision process then introduce federated learning method for popularity prediction. Subsequently, P2D3PG developed based reinforcement solve problems. Simulation results demonstrate superiority approach improving rate baseline methods while preserving privacy.