Bandwidth estimation and rate adaptation for 360-degree mobile video streaming

The popularity of 360-degree (panoramic) videos is evident in recent years. Support from device manufacturers and commercial video platforms made the streaming of 360-degree video content available for a large number of users. The majority of these users are connected through cellular networks usually using their smartphones, which have limited resources shared among multiple users at the same time. Therefore, users usually receive a small amount of bandwidth that can vary according to other users requests. Unfortunately, 360-degree videos have very high bandwidth requirements which make their streaming challenging on cellular networks. In this thesis, we address two problems: capacity estimation and rate adaptation. For capacity estimation, we propose a new method, referred to as Phy360, that leverages information from the network structure and transmission patterns to get a precise estimation of the network current capacity. For the second problem, we propose a rate adaptation algorithm for tiled 360-degree videos, which results in smooth and high quality viewports. We implemented the proposed algorithms in an LTE testbed. Our evaluation shows that Phy360 achieves more accurate bandwidth estimation and is robust to different channel conditions, and our rate adaptation algorithm improves the overall 360-degree video quality, compared to the closest works in the literature.