Optimal Scalable Video Scheduling Policies for Real-Time Single- and Multiuser Wireless Video Networks

In this paper, we present optimal video scheduling policies for layered scalable video streaming in delay-sensitive real-time fading mobile wireless video networks. Toward this end, we formulate the optimal video scheduling paradigm as a Markov decision process (MDP) for long-term video reward maximization, employing the scheduler state that is comprised of the fading channel, scalable video buffer status, decoder display slack, and large-scale wireless radio channel conditions. Further, we demonstrate structural properties of the optimal scheduling policy that significantly reduce the optimal policy computation. Subsequently, the above optimal video scheduling paradigm is extended to the context of a multiuser-fading-channel-based wireless video streaming scenario, where we develop an optimal index-based policy for user video scheduling, additionally incorporating fairness among users through a starvation-based approach. The presented optimal video scheduling policies are based on practical parameter values extracted from the Joint Scalable Video Model (JSVM)-based H.264 scalable video codec and, hence, are applicable in practice. Simulation results demonstrate improved video quality performance of the proposed optimal scalable video scheduling policies for single-user and multiuser scenarios.