Tile-Based Edge Caching for 360° Live Video Streaming

Abstract

360 o video is becoming an increasingly popular technology on commercial social platforms and vital part of emerging Virtual Reality/Augmented Reality (VR/AR) applications. However, the delivery of 360 o video content in mobile networks is challenging because of its size. The encoding of 360 o video into multiple quality layers and tiles and edge cacheassisted video delivery have been proposed as a remedy to the excess bandwidth requirements of 360 o video delivery systems. Existing works using the above tools have shown promising performance for Video-on-Demand (VoD) 360 o delivery, but they cannot be straightforwardly extended in a live-streaming setup. Motivated by the above, we study edge cache-assisted 360 o live video streaming to increase the overall quality of the delivered 360 o videos to users and reduce the service cost. We employ Long Short-Term Memory (LSTM) networks to forecast the evolution of the content requests and prefetch content to caches. To further enhance the delivered video quality, users located in the overlap of the coverage areas of multiple Small Base Stations (SBSs) are allowed to receive data from any of these SBSs. We evaluate and compare the performance of our algorithm with Least Frequently Used (LFU), Least Recently Used (LRU), and First In First Out (FIFO) algorithms. The results show the superiority of the proposed approach in terms of delivered video quality, cache-hitratio and backhaul link usage. Index Terms-Tile-encoding, 360 o live video streaming, edgecaching, LSTM networks.

Machine Learning for Multimedia Communications

“…Different machine learning algorithms have been proposed to solve the content placement and update problem at caches (and in some cases the delivery problem), such as Transfer learning [ 136 , 137 ], deep Q-learning [ 34 , 138 , 139 , 140 ], Actor-Critic [ 141 , 142 ], multi-agent multi-armed bandits (MMBAs) [ 143 , 144 ], 3D-CNN [ 145 ], LSTM networks [ 146 , 147 ], among other methods. Reinforcement learning algorithms are often preferred, as cache updates can be modelled as a Markov Decision Process [ 34 , 138 , 139 , 140 , 141 , 142 , 143 , 144 ], while deep learning supervised learning methods [ 145 , 146 , 147 ] are used to capture the trends in the evolution of the video requests. These trends are then used to optimize the content placement and the cache updates.…”

“…In such coded caching schemes, users should download a file with size of at least equal size to the original video file prior to displaying it and hence cannot respect time delivery deadlines. Only a few works in the literature consider timing constraints and can be used for video-on-demand [ 126 ] and live streaming [ 140 , 141 , 147 ].…”

“…Performance gains can be noticed by clustering the users using machine learning to decide how to optimally serve the content [ 145 ] from the base stations and by taking into account QoE metrics [ 141 ]. Further performance improvements can be noticed by considering the content importance [ 126 , 140 , 147 ].…”

“…Therefore, caching algorithms for 360-degree videos should not only consider the video popularity but also take into account the viewport’s popularity or the tile’s popularity. Motivated by this change, the works in [ 140 , 147 ] proposed machine learning algorithms that optimize which tiles to cache and on what quality. Caching systems designed for regular videos [ 34 , 138 , 139 , 140 , 141 , 142 , 143 , 144 , 145 ] perform poorly for 360-degree video, as they do not take into account the users’ consumption model and the content importance.…”

A Cooperative Caching and Computing-Offloading Method for 3C Trade-Off in VR Video Services

“…[10] studied the cache-assisted 360 • video streaming to increase the overall quality of the delivered 360 • videos to users and reduce the service cost. But the entire panoramic video needed to be transmitted in the network of [10]. Since each user only watched a viewpoint of the 360 • VR video at any given time, the requested FOV was chosen to be transmitted instead of the entire panoramic video, thereby saving transmission resource significantly [4], [11], [12].…”