ProCAVE cuts edge video latency with deep RL caching

ProCAVE (Proactive Caching with Adaptive Video Experience) is a self-adaptive framework that unifies three components: a lightweight Transformer that forecasts short-term throughput, a PPO-based adaptive bitrate (ABR) agent that selects video quality, and a DDPG-based continuous cache controller that manages a high-dimensional global state. This full-lifecycle approach replaces the reactive ABR heuristics and loosely coupled cache policies used by existing systems like FlyCache, enabling more coordinated decisions under real-world wireless network fluctuations.

In experiments using MovieLens user preference traces and Ghent 4G bandwidth measurements, ProCAVE delivered measurable gains over FlyCache and other baselines: higher byte hit rates, lower backhaul traffic, and improved quality of experience (QoE). By making edge caching proactively aware of both network conditions and user preferences, ProCAVE demonstrates that DRL-coordinated control can significantly boost efficiency for mobile video delivery—a key competitive area as edge computing infrastructure scales.