Delivering the right content at the right time is one of the main challenges in designing smart information delivery systems. Predicting the user’s preferences in the future and caching the required content in advance to improve the quality of service has been proposed and investigated before for different applications.

In this paper, we explicitly consider a scenario with very high bandwidth and speed requirements, an augmented/virtual reality-enabled home, where the user interacts with an AR/VR device. Our goal is to deliver information in the maximum quality and optimize the delivery cost by predicting user’s requests from the AR/VR system. We first generate synthetic user requests from two real datasets and two bigger simulated datasets of users’ daily activities. We then propose a method based on LSTMs and a probability-based approach for prediction and content caching by considering the quality of the content, user satisfaction, and caching cost. Our experiments suggest that our method works better than baseline caching strategies in terms of cost for caching and user satisfaction.

在正确的时间交付正确的内容是设计智能信息交付系统的主要挑战之一。之前已经针对不同的应用提出并研究了预测用户的喜好并预先缓存所需的内容以提高服务质量。

在本文中，我们明确考虑了具有非常高的带宽和速度要求的场景，即启用了增强/虚拟现实的家庭，其中用户与AR / VR设备进行交互。我们的目标是通过预测来自AR / VR系统的用户请求，以最高质量交付信息并优化交付成本。我们首先从两个真实的数据集和两个更大的用户日常活动模拟数据集生成综合用户请求。然后，我们考虑到内容的质量，用户满意度和缓存成本，提出了一种基于LSTM的方法和基于概率的预测和内容缓存方法。我们的实验表明，就缓存成本和用户满意度而言，我们的方法比基线缓存策略更好。