(A) cache update strategy and bandwidth allocation to minimize the average latency in wireless VR

Abstract

Recently, the popularity of virtual reality (VR) contents, which enables users to experience the virtualized world, has been increased. However, ultra-low latency is needed for providing the users with high-quality VR experience. In order to satisfy the latency requirement, the wireless VR delivery framework based on mobile edge computing (MEC) networks, where the VR device can store some of 2D or 3D field of views (FoVs) and locally compute the projection from 2D FoV to 3D FoV, was studied. In this paper, we propose a cache update strategy which updates the cache contents of the VR device by predicting the user's viewpoint in the above wireless VR delivery framework. The proposed cache update strategy is based on the bandwidth allocation, which divides the bandwidth into two parts: part for downloading the current viewpoint of the user, part for updating the cache with predicted viewpoint. We consider two cache update scenarios, 2D FoV update scenario and 3D FoV update scenario, and model both scenarios as a Markov chain, respectively. Furthermore, we formulate the bandwidth allocation optimization problem to minimize the average latency at steady-state for both cases, and numerically find the optimal bandwidth allocation. Numerical results show that the proposed cache update strategy can significantly reduce the average latency than tight latency constraint in wireless VR.