Cooperative Caching, Rendering, and Beamforming for RIS-Assisted Wireless Virtual Reality Networks
Cooperative Caching, Rendering, and Beamforming for RIS-Assisted Wireless Virtual Reality Networks
复制标题
DOI：
10.1109/tvt.2023.3345354
复制
发表时间：
2024-05
影响因子：
6.8
通讯作者：
Jian Chen;Jiale Xia;Jie Jia;Leyou Yang;Xingwei Wang
中科院分区：
文献类型：
--
作者：
Jian Chen;Jiale Xia;Jie Jia;Leyou Yang;Xingwei Wang
无线虚拟现实网络（WVRNs）提供了虚拟现实设备之间的无缝连接，具有巨大的应用和商业价值。然而，制约其发展的主要问题是VR设备上3D视频渲染的高能耗和计算量消耗。为了解决这个问题，我们提出了一种新的协调多点（CoMP）和可重构智能表面（RISs）辅助系统，其中视频由多个协作移动边缘计算（MEC）服务器同时呈现。此外，与这些MEC服务器相关联的BSs组成一个CoMP集群，以实现高数据速率。本文旨在通过联合优化MEC服务器上的视频缓存和渲染以及BSs和RIS的波束形成，最大限度地降低长期功耗。我们提出了一个在线混合学习框架，该框架结合了用于视频缓存和渲染的深度强化学习（DRL），以及用于所有BSs和RIS的波束形成的交替优化。特别地，DRL算法中每个动作的奖励是通过提出的交替优化问题来计算的，从而减少了动作空间，加快了收敛速度。数值结果和对比实验表明，该方法可以有效降低系统的长期平均功耗，满足低计算复杂度的3D视频传输要求，优于不使用CoMP和RIS技术的方法。
Wireless virtual reality networks(WVRNs) provide seamless connectivity between virtual reality devices with colossal application and commercial value. However, the main problem restricting its development is the high energy and computational consumption in 3D video rendering on VR devices. To address this issue, we propose a novel coordinated multi-point (CoMP) and reconfigurable intelligent surfaces (RISs) assisted system, where the video is rendered by multiple collaborative mobile edge computing (MEC) servers simultaneously. Besides, BSs associated with these MEC servers are formed as a CoMP cluster to achieve a high data rate. This paper aims to minimize long-term power consumption by jointly optimizing the video caching and rendering at the MEC servers and the beamforming for both BSs and RIS. We propose an online, hybrid learning framework that combines deep reinforcement learning (DRL) for video caching and rendering, and an alternating optimization for the beamforming of all BSs and the RIS. In particular, the reward of each action in the DRL algorithm is calculated by the proposed alternating optimization problem, thus reducing the action space and accelerating convergence speed. Numerical results and comparison experiments show that our proposed method can effectively reduce the long-term average power consumption of the system, satisfy the requirement of 3D video transmission with low computational complexity, and outperform that without CoMP and RIS techniques.