Bitrate adaptation for video streaming using RL

We will implement a system from an existing paper called Pensieve. Pensieve is a system that generates adaptive bitrate algorithms using reinforcement learning. It trains a neural network model that selects bitrates for future video chunks based on observations collected by client video players. Currently, most video players use policy based algorithms to decide the bitrate for the next chunks, however, with the ever changing network conditions and with the development of new network protocols like QUIC, these policies become obsolete/ outdated. So, there is a need for an algorithm that can capture not only the diverse network protocols and conditions into its decision making process, but can also update its policies automatically over time. To address these issues, Pensieve generates ABR algorithms using observations of the resulting performance of past decisions across a large number of video streaming experiments. This allows Pensieve to optimize its policy for different network characteristics and QoE metrics directly from experience.

Built With

• python
• pytorch