Game theory based bitrate adaptation for Dash.Js reference player

Abstract

Most existing DASH adaptive bitrate (ABR) schemes are designed to behave in their own self-interest and do not perform consistently in all network environments. In this work, we provide a practical implementation, materials and demonstration of a game theoretical ABR scheme, termed GTA [1], in the dash.js reference player. The GTA approach optimizes the viewer experience across multiple players without requiring explicit communication, and maintains a high playback bitrate while reducing startup delay, and minimizing quality switches and stalls.