Multi-connection Based Scalable Video Streaming in UDNs: A Multi-armed Bandit Approach.

Scalable video coding (SVC) has been proposed as a promising paradigm for video transmission over wireless communications due to its flexibility. In this paper, we study the SVC-based video streaming in ultra-dense networks (UDNs), which allows each user to download different layers of its demanded video block from different base stations (BSs), rather than a single one in most existing work. Specifically, an optimization problem is formulated aiming to maximize the quality of experience (QoE) for each user. To this end, we first formulate a subproblem of choosing the optimal connection strategy as a multi-armed bandit (MAB) problem with no information exchange among users. In addition, each user is enabled to adapt its connection strategy in a self-learning process. For obtaining the optimal arm for the MAB problem, a best-arm-selecting upper confidence bound (UCB) algorithm is proposed. Based on this, we further develop an SVC-based video downloading scheme which provides an approximately optimal solution to the original optimization problem. Finally, simulations and comparisons are conducted to show the feasibility and superiority of our proposed schemes.