Field of View Aware Proactive Caching for Mobile Augmented Reality Applications.

Mobile Augmented Reality (MAR) applications require significant computational and storage resources at the end devices or at edge clouds (EC) to, inter alia, support for the Augmented Reality Objects (AROs) that are amalgamated to the physical world. In this work, a MAR service is considered under the lenses of microservices where MAR service components can be decomposed and anchored at different locations ranging from the end device to different ECs to optimize the overall service and network efficiency. The novel content-aware aspect of the proposed solution allows for proactive caching of high probability 2D field of views (FoVs) of the AROs to be stored instead of caching the significantly larger and complex 3D original AROs. To this end, a joint optimization scheme (Optim) considering mobility and the trade-off between delay, storage capacity and FoV allocation is proposed. A nominal Long Short Term Memory ( LSTM) deep neural network is further explored to provide efficient pro- active decision making in real-time. More specifically, the LSTM deep neural network is trained with optimal solutions derived from a mathematical programming formulation in an offline manner. A set of numerical investigations reveal that optimal decisions manage to outperform recently proposed schemes by 24.4% to 67.5% in delay under different weights, whilst the LSTM deep neural network is effective in providing competitive solutions as well as being amenable for real-time decision making.