Anticipating Resource Management and QoE for Mobile Video Streaming under Imperfect Prediction

Imen Triki , Rachid El-Azouzi , Majed Haddad (1)

1 LIA - Laboratoire Informatique d'Avignon

By leveraging geolocation and contextual information for mobile users, the prediction of the future throughput becomes more and more feasible. Many approaches on contextaware content delivery have been explored to balance the operators' limited resources with users' requirements. However, the perfect knowledge of the future context cannot be easily performed in real world, which represents a hurdle for most context-aware approaches. In this paper, we address a contextaware delivery algorithm for adaptive video streaming (NEW-CAST) that have already been explored in [1] under perfect knowledge of future capacity, to balance the user's perception of the video and the cost of network usage. In order to make NEWCAST more resistant to eventual throughput prediction errors and adapt it to short-term horizons, we propose 4 algorithms that efficiently reduce the number of stalls by at least 75%.

Domaines: Réseaux et télécommunications [cs.NI]; Intelligence artificielle [cs.AI]; Théorie de l'information [cs.IT]

Imen Triki, Rachid El-Azouzi, Majed Haddad. Anticipating Resource Management and QoE for Mobile Video Streaming under Imperfect Prediction. 2016 IEEE International Symposium on Multimedia (ISM), Dec 2016, San Jose, United States. pp.93-98, ⟨10.1109/ISM.2016.0026⟩. ⟨hal-03485117⟩