Age of Information Optimization for Cache Update with Dynamic Content Popularity

Abstract

In edge caching systems, maintaining content freshness is critical for optimizing user experience, especially under dynamic content popularity. This paper proposes a novel Dynamic Cache Update Algorithm (DCUA) that leverages Age of Information (AoI) as a performance metric to optimize cache updates. To address the challenge that the content popularity changes over time which is modeled as a finite-state Markov chain, AoI minimization problem is formulated as a Markov decision process (MDP). By carefully designing the state-dependent features and reward, DCUA integrates Proximal Policy Optimization (PPO) to learn the cache update policy. By employing Exponential Moving Averages (EMA) to track shifting popularity trends, DCUA adapts the cache update frequency to match the dynamic peaks and troughs of user demands. Numerical simulation results demonstrate that DCUA consistently achieves superior cache freshness and robust convergence compared with baseline strategies by accurately capturing temporal popularity trends across diverse dynamic scenarios.