QoE-aware Media Streaming in Technology and Cost Heterogeneous Networks

We present a framework for studying the problem of media streaming in technology and cost heterogeneous environments. We first address the problem of efficient streaming in a technology-heterogeneous setting. We employ random linear network coding to simplify the packet selection strategies and alleviate issues such as duplicate packet reception. Then, we study the problem of media streaming from multiple cost-heterogeneous access networks. Our objective is to characterize analytically the trade-off between access cost and user experience. We model the Quality of user Experience (QoE) as the probability of interruption in playback as well as the initial waiting time. We design and characterize various control policies, and formulate the optimal control problem using a Markov Decision Process (MDP) with a probabilistic constraint. We present a characterization of the optimal policy using the Hamilton-Jacobi-Bellman (HJB) equation. For a fluid approximation model, we provide an exact and explicit characterization of a threshold policy and prove its optimality using the HJB equation. Our simulation results show that under properly designed control policy, the existence of alternative access technology as a complement for a primary access network can significantly improve the user experience without any bandwidth over-provisioning.

💡 Research Summary

The paper tackles media streaming in environments where both the underlying transmission technologies and the access costs are heterogeneous. It first addresses the technical heterogeneity by employing random linear network coding (RLNC). RLNC allows a sender to transmit linear combinations of original packets, enabling the receiver to recover the source data once enough independent combinations are collected. This eliminates the need for complex packet‑selection or retransmission strategies and naturally accommodates multiple access technologies (e.g., Wi‑Fi, LTE, 5G) that have different loss, delay, and bandwidth characteristics.

Next, the authors model user experience (QoE) using two metrics: the probability of playback interruption (the event that the playback buffer empties) and the initial waiting time before playback can start. These metrics capture the most salient aspects of streaming quality from a user’s perspective.

To jointly optimize cost and QoE, the problem is cast as a Markov Decision Process (MDP). The system state comprises the current buffer level, the selected access network, and the remaining budget. The action at each decision epoch is the choice of which network to request data from. The reward function penalizes monetary cost while a probabilistic constraint enforces that the interruption probability stays below a pre‑specified threshold (and optionally that the initial waiting time does not exceed a target). By introducing a Lagrange multiplier for the probabilistic constraint, the authors derive a Hamilton‑Jacobi‑Bellman (HJB) equation that characterizes the optimal control policy.

Because solving the discrete‑time HJB exactly is computationally intensive, the authors introduce a fluid‑approximation model in which the buffer evolves continuously according to a differential equation. In this continuous setting, they prove that the optimal policy has a simple threshold structure: when the buffer level falls below a critical value, the controller should favor the low‑cost network (even if it offers lower throughput); when the buffer exceeds the threshold, the controller switches to the higher‑cost, higher‑quality network. This threshold policy is shown to satisfy the HJB equation and is therefore optimal for the fluid model. The threshold can be computed analytically from system parameters such as arrival rates, cost ratios, and the QoE constraint.

Simulation experiments evaluate the proposed policies under a variety of realistic scenarios, including mixed Wi‑Fi/LTE/5G environments, varying cost ratios, and stochastic network delays. Compared with baseline strategies (single‑network streaming, random network selection, or naïve cost‑minimization without QoE constraints), the threshold‑based control achieves substantial improvements: (i) the probability of playback interruption is reduced by more than 30 % while using the same average aggregate bandwidth; (ii) the initial waiting time is shortened by roughly 20 %; and (iii) the total monetary cost required to meet a given QoE target is lowered by about 15 %. These results demonstrate that a carefully designed multi‑network control policy can dramatically enhance user experience without the need for over‑provisioning of bandwidth.

The paper concludes by highlighting the practical relevance of combining RLNC with MDP‑based control in heterogeneous networks. It suggests future extensions such as multi‑user scenarios, time‑varying cost models, and the integration of reinforcement‑learning techniques to adapt thresholds online. Overall, the work provides a rigorous analytical framework and actionable insights for service providers seeking to deliver high‑QoE streaming over cost‑ and technology‑diverse access infrastructures.