Hierarchical Forwarding Resource Allocation With Proactive Frame Dropping for VR Video Streaming

In this paper, a hierarchical forwarding resource allocation scheme with proactive frame dropping is proposed for multi-user virtual reality (VR) video streaming, aiming to enhance video quality in the presence of network congestion caused by multiple VR flows sharing a certain hop while considering the impact of network fluctuation. Firstly, we assess the queuing delay bound of frames, allowing bottleneck node to preemptively discard frames that are anticipated to expire. Secondly, we model the priority of different frames based on encoding referencing relationship and users' viewpoints, which helps to prioritize frames of each flow. Subsequently, we formulate an optimization problem to minimize the long-term quality loss caused by frame dropping, while adhering to per-flow quality assurance and total resource constraints. Considering the distinct timescales of network fluctuation and frame dropping, we propose a two-timescale hierarchical resource allocation scheme. On the long timescale, a resource allocation algorithm is designed based on queuing theory to calculate the resource demand of each flow. On the short timescale, to accommodate network fluctuations, a low-complexity heuristic algorithm is devised to adjust allocation results based on frame priority. Comprehensive simulation results validate the effectiveness of the proposed scheme in improving the quality of VR flow under diverse network conditions.