Multi-stage Congestion Games for Wireless Real-time Streaming

This work mainly proposes to model real-time streaming in congestion game form: at each stage, given the prevailing content distribution, peers (or players) choose who to ask for an additional content unit. In this way, (stage- and peer-wise) congestion is measured precisely by counting how many requests to forward are addressed to each peer (and to the broadcaster). While in the wired setting all nodes (or peers, together with the broadcaster) are assumed to directly communicate with each other, in the wireless framework an ad hoc network formalizes transmission constraints due to spatial positions: the broadcaster and each peer can send content only to those at unitary hop distance from them in such a network. For the wired (or fully-connected) setting, we develop a strategy restriction mechanism whose equilibrium outcomes minimize both streaming length (or number of stages needed to spread the whole content over the whole population) and congestion, while also fulfilling a fairness condition. As this mechanism heavily relies upon connectivity among nodes, in general it does not apply to the wireless (or non-fully connected) case. Accordingly, the focus next turns on equilibrium outcomes with no strategy restriction apart from communication constraints due to the ad hoc network. In the wired setting, simulations allow to compare equilibrium outcomes with and without the proposed strategy restriction mechanism, while in the wireless one they are used for evaluating how performance deteriorates, at equilibrium, as the size of the given ad hoc network decreases.