Notice: This paper has been recommended as a "Distinguished Paper" in the IEEE ComSoc MMTC Reviewer Letter in April 2013.
The rate at which data can be sent over a wireless link is inherently variable because of the variable link conditions. This variability can be further aggravated by bursty co-channel interference. Low power access points (APs) like femtocells can be a major source of bursty interference, as they are sporadically active due to fewer users served by them. Increasing the number of such APs can lead to further variations in throughput across the network. These throughput variations lead to large quality variations specifically for real-time video, in which case the throughput variations cannot be smoothed out through halting playback and buffering, resulting in a degraded quality of experience (QoE) for the user.
In this paper, the authors propose and analyze a network-level resource management algorithm called interference shaping to smooth out these throughput variations, and hence improve the QoE of video users by reducing the variability of interference. Interference shaping operates by decreasing the transmission power, and hence peak rate, of co-channel APs serving bursty data (best-effort) users. This smooths their transmit power profile and hence the interference caused by them to the video user link, at the cost of a modest rate decrease for best-effort users. The proposed technique is analyzed by mapping the throughput variations for video users to the corresponding video quality fluctuations and packet loss rate. For video users, QoE is quantified by bench-marking against a metric, which incorporates the strong dependence of the current QoE (which is subjective) on the recent past. The QoE of data users is evaluated using a framework, which quantifies the response of human sensory system to an external stimulus.
The proposed technique increases mean video QoE and reduces the QoE variability over time, with a net perceptual increase of about 2-3x in illustrative settings, while incurring insignificant decrease in the QoE for co-channel data users. The presented algorithm introduces a trade-off between the QoE of video users and other data users, and an optimal operating regime depends on the context. A load-aware cellular model with randomly located interfering APs transmitting bursty data is also developed. Using this model, it is shown that aside from smoothing the throughput of the video link, interference shaping may also increase the mean capacity in scenarios where each interfering AP serves a small number of data users. Increased capacity allows a higher video encoding rate and hence results in higher mean quality. Interference shaping can be applied to both unicast and multicast real-time video streaming with gains proportional to the number of video users sharing the same broadcast and interferers in the latter.