Video Streaming Caching and Transcoding for Heterogeneous Mobile Users.

With the increasing prevalence of video streaming media, users have higher requirements for the quality of video streams, which can be quantified as Quality of Experience (QoE). Caching and transcoding video chunks on edge servers are effective methods to improve users’ QoE and reduce the cost of retrieving video chunks. In this paper, we investigate how to cache and transcode video chunks for mobile users to maximize their QoE, while taking into consideration the cooperation between edge servers, users’ mobility, and heterogeneous preferences, simultaneously. We employ a Multi-Agent Reinforcement Learning (MARL) framework and propose an MARLbased Cache replacement and Transcoding (MACT) mechanism. More specifically, we formulate the problem as a multi-agent game and prove this game is a state-based potential game. Then, we decompose this multi-agent game into individual agent decision problems to be solved. Through extensive simulations, we demonstrate that the proposed MACT mechanism achieves about 19.3% performance improvement, compared to the state-of-the-art works.