ASTRA: Predictive Prefetching for Edge AR

Abstract

ASTRA addresses the latency bottlenecks in edge-assisted Augmented Reality (AR) by modeling user intent through field-of-view and spatial data.

Note: This research received the Best Research Paper Award at IEEE IC2E 2025.

Methodology & Results

Built utilizing C++ and SimCpp, the predictive framework dynamically anticipates data needs.

• Increased cache hit rates by 35% compared to baseline strategies[cite: 35].
• Reduced end-to-end latency by 11%[cite: 35].