Survey on Reinforcement Learning Approaches for Cache-Assisted Video Streaming System

Abstract

This paper surveys the state-of-the-art studies on video caching and delivery especially based on the reinforcement learning (RL) approach. Even though there have been many researches on the RL-based content caching and delivery, caching and delivery of video contents should be more carefully designed to reflect the unique characteristics of video contents. We focus on the following characteristics of videos files: i) a video can be encoded into multiple quality levels, ii) provision of streaming services for mobile users is necessary, and iii) a video consists of multiple sequential chunks.