Characterizing Distributed Mobile Augmented Reality Applications at the Edge

What is it about?

The growing interests in mixed reality (XR), from both academia and industry, such as Apple and Meta, have paved the way for new use cases leveraging immersive perception and interaction platforms. Augmented reality (AR), as one of the prominent pillars in XR, enables real-time interactive user experiences on mobile devices by superimposing digital content onto physical environments. Our study marks one of the first attempts to examine the operational challenges of deploying distributed stream processing (DSP) based AR applications on edge-cloud infrastructure.

Featured Image

Photo by julien Tromeur on Unsplash

Photo by julien Tromeur on Unsplash

Why is it important?

Most existing AR applications are monolithic. A client captures a camera feed and offloads computation to a remote server. Even explorations that leverage edge computing to improve performance are typically limited to single-edge servers, primarily due to strong dependencies between different pipeline stages. While existing research suggests that distributing AR functions across multiple edge servers can improve performance, existing solutions ignore effective ways to distribute and replicate AR services while considering resource contention, virtualization techniques, and hardware heterogeneity.