Perceptual Quality-aware Active Queue Management for Video Transmission

Reducing the queue length and queuing delay in the buffer which in turn reduces congestion and latency in the network are of great importance for real-time video services. In this paper, we propose a perceptual quality-aware queue management mechanism to shorten the queue length and waiting time of video traffic in a finite buffer. Our proposed approach prioritizes video layers in a stream in accordance with their hierarchical structure and selectively drops packets from different layers under a target user perceived quality constraint. This in turn reduces network load and latency and increases the capacity to serve more concurrent streams while satisfying users' perceptual quality requirements. We develop a statistical queuing model to express the delay requirements of video layers with an equivalent cross-layer constraint. We then formulate the queue management problem as a minimization of queuing delay using a selective packet drop approach under perceptual quality and queue expiry time provisioning. Numerical results show that our proposed approach outperforms the widely deployed random early detection (RED) and tail drop (TD) queue management schemes in terms of reducing queuing delay and queue length in the buffer, while increasing the fraction of perceived quality-satisfied users.