Federated Deep Learning for Immersive Virtual Reality over Wireless Networks
Federated Deep Learning for Immersive Virtual Reality over Wireless Networks
复制标题
DOI：
10.1109/globecom38437.2019.9013419
复制
发表时间：
2019-12
期刊：
影响因子：
--
通讯作者：
Mingzhe Chen;Omid Semiari;W. Saad;Xuanlin Liu;Changchuan Yin
中科院分区：
文献类型：
--
作者：
Mingzhe Chen;Omid Semiari;W. Saad;Xuanlin Liu;Changchuan Yin
在本文中，增强无线用户的虚拟现实（VR）体验的问题进行了研究，通过最大限度地减少中断的存在（BIP），可以脱离他们的虚拟世界的用户的发生。为了测量无线VR用户的BIP，提出了一种新的模型，该模型联合考虑了VR应用、传输延迟、VR视频质量和用户对虚拟环境的感知。在所开发的模型中，基站（BS）使用定向传输链路将VR视频传输到无线VR用户，以提高VR用户的数据速率，从而减少每个用户的BIP数量。因此，在最小化BIP时必须考虑VR用户的移动性和方向，因为VR用户的身体移动可能会导致其无线链路的阻塞。BIP问题被公式化为优化问题，其联合考虑用户移动性模式、方向及其BS关联的预测。为了预测VR用户的方向和移动模式，提出了一种基于深度回声状态网络（ESN）机器学习框架的分布式学习算法。所提出的算法使用来自联合学习的概念，使多个BS能够使用其收集的数据在本地训练其深度ESN，并协作构建学习模型来预测整个用户的移动模式和方向。使用这些预测，最大限度地减少BIP的用户关联策略。仿真结果表明，与集中式ESN和深度学习算法相比，所开发的算法分别将用户的BIP降低了16%和26%。
In this paper, the problem of enhancing the virtual reality (VR) experience for wireless users is investigated by minimizing the occurrence of breaks in presence (BIPs) that can detach the users from their virtual world. To measure the BIPs for wireless VR users, a novel model that jointly considers the VR applications, transmission delay, VR video quality, and users’ awareness of the virtual environment is proposed. In the developed model, the base stations (BSs) transmit VR videos to the wireless VR users using directional transmission links so as to increase the data rate of VR users, thus, reducing the number of BIPs for each user. Therefore, the mobility and orientation of VR users must be considered when minimizing BIPs, since the body movements of a VR user may result in blockage of its wireless link. The BIP problem is formulated as an optimization problem which jointly considers the predictions of users’ mobility patterns, orientations, and their BS association. To predict the orientation and mobility patterns of VR users, a distributed learning algorithm based on the machine learning framework of deep echo state networks (ESNs) is proposed. The proposed algorithm uses concept from federated learning to enable multiple BSs to locally train their deep ESNs using their collected data and cooperatively build a learning model to predict the entire users’ mobility patterns and orientations. Using these predictions, the user association policy that minimizes BIPs is derived. Simulation results demonstrate that the developed algorithm reduces the users’ BIPs by up to 16% and 26%, respectively, compared to centralized ESN and deep learning algorithms.