To cite this version:Alice Combernoux, Cyrile Delestre, Nesrine Changuel, Bessem Sayadi, Michel Kieffer. Cross-layer optimization of a multimedia streaming system via dynamic programming. The problem is cast in the framework of Markov Decision Processes (MDP). This formalism provides efficient tools to compute a foresighted control policy maximizing some longterm discounted sum of rewards linked to the video quality received by the user. Experimental results illustrate the benefits in terms of average PSNR of this approach compared to a short-term (myopic) policy. The robustness of the proposed control policy to variations of the transmitted contents is also illustrated.

No abstract