Multimedia Streaming over P2P networks using Markov Decision Processes

Abstract

The aim of this project is to overcome the limitations of Foresighted Resource Reciprocation strategy (FRR) [11] for distributing multimedia content while providing a better performance for resource reciprocation than other solutions proposed. In this project, we build on the Foresighted Resource Reciprocation strategy by explicitly considering the timing constraints for continuous display of the multimedia data and the importance of each multimedia data segment for the multimedia quality. In particular, we incorporate data priority functions into the reward function in order to adapt to the specific characteristics of media streaming applications. As a result, the peers exchange video packets with a strategy that ensures that the most important packets have a higher probability to reach the decoder on time for proper decoding.