Hybrid Content Caching Empowered By AIGC in Wireless Networks

Abstract

Content caching at base stations (BS) can reduce backhaul traffic delays to deliver the requested files to users, but its effectiveness is limited by BS storage capacity. We propose a novel approach that integrates artificial intelligence-generated content (AIGC) into the BS operations. Instead of caching entire files, our AIGC-enhanced BS can cache smaller prompts, allowing files to be reconstructed on demand. We explore the challenge of jointly optimizing hybrid caching, AIGC computation, and communication resource allocation with the goal of minimizing average system latency. Given the non-convex nature and the complexity of mixed integer non-linear programming involved, we propose a divide-and-conquer algorithm that breaks down the problem into two timescale levels. Theoretical analysis and simulations confirms that our AIGC-enhanced hybrid content caching outperforms the conventional content caching.