Proactive Edge Cloud Optimization for Mobile Augmented Reality Applications

Enhanced LSTM-based Service Decomposition for Mobile Augmented Reality

“…The benefit of applying such decomposition can be further revealed by a toy example shown in Fig. (2). When there is no mobility like the first case showing in Fig.…”

“…To this end, Edge clouds (ECs) assistance allows the deployment of service at the edge of the network and hence is widely embedded in the design for orchestrating and offloading MAR services. In [2], we have outlined how a MAR application can be decomposed into a series of granular micro-services together with an optimization framework to provide optimal pro-active mobility-aware decision making in terms of EC assignment. As illustrated in Fig.…”

“…When there is no mobility like the first case showing in Fig. (2), it could be possible to deploy the MAR application at a single server. However, it is clear that, without decomposition, the complete MAR application uses significant levels of CPU and cache memory resources.…”

“…Thus, the given prediction not only contains where to set functions and send requests, but also indicates what AR content should be allocated to the target server. In the Optim scheme from [2], constructing and solving a complex mixed integer linear problem is quite time-consuming and cannot respond to network changes in time. Therefore, the previous scheme should be improved to take into account both highquality solutions and computational efficiency.…”

“…As already eluded in this work optimal decision making is utilized to train the deep neural network. The mathematical programming formulation that is used to provide optimal decision making is detailed in [2];in that case, an edge cloud resource allocation optimization framework is proposed to proactively allocate resources and to satisfy related requirements of MAR applications. The user mobility is considered explicitly and the overall delay is minimized as multi-objective optimization problem.…”