This thesis aims at providing an innovative contribution to the definition of the Future Internet Core Platform, in the frame of the "La Sapienza" University research activities on the EU FP7 FI-WARE project. The thesis introduces and designs an innovative "Cognitive Application Interface" in charge of deriving key parameters driving the Network Control elements to meet personalised Application Quality of Experience Requirements. The thesis proposes the innovative concept of a dynamic association between Applications and Classes of Service. A Reinforcement Learning based approach is followed. A solution based on a standard Q-learning algorithm is proposed. Simulation results obtained using the OPNET simulation tool are described. Preliminary work on an alternative solution based on a Foe Q-Learning algorithm is also illustrated. The proposed framework is very flexible, allows QoE personalization, requires low processing capabilities and entails a very limited signalling overhead.
A Reinforcement Learning based Cognitive Approach for Quality of Experience Management in the Future Internet / Fogliati, Laura. - (2012 Apr 11).
A Reinforcement Learning based Cognitive Approach for Quality of Experience Management in the Future Internet
FOGLIATI, LAURA
11/04/2012
Abstract
This thesis aims at providing an innovative contribution to the definition of the Future Internet Core Platform, in the frame of the "La Sapienza" University research activities on the EU FP7 FI-WARE project. The thesis introduces and designs an innovative "Cognitive Application Interface" in charge of deriving key parameters driving the Network Control elements to meet personalised Application Quality of Experience Requirements. The thesis proposes the innovative concept of a dynamic association between Applications and Classes of Service. A Reinforcement Learning based approach is followed. A solution based on a standard Q-learning algorithm is proposed. Simulation results obtained using the OPNET simulation tool are described. Preliminary work on an alternative solution based on a Foe Q-Learning algorithm is also illustrated. The proposed framework is very flexible, allows QoE personalization, requires low processing capabilities and entails a very limited signalling overhead.File | Dimensione | Formato | |
---|---|---|---|
PHD Thesis Fogliati.pdf
accesso aperto
Tipologia:
Tesi di dottorato
Licenza:
Creative commons
Dimensione
2.88 MB
Formato
Adobe PDF
|
2.88 MB | Adobe PDF |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.