Reproduce “Towards Timely Video Analytics Services at the Network Edge” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xishuo Li, Shan Zhang, Yuejiao Huang, Xiao Ma, Zhiyuan Wang, Hongbin Luo
Researchers from Beihang University introduce the Age of Processed Information (AoPI) as a unified metric for quantifying the timeliness and accuracy of real-time video analytics services at the network edge. They propose a Lyapunov-based Block Coordinate Descent (LBCD) algorithm that minimizes AoPI by adaptively managing edge server selection, video configuration, and resource allocation, reducing AoPI by up to 10.94X compared to baselines in dynamic environments while meeting accuracy constraints.
Key insights
- The Age of Processed Information (AoPI) provides a unified metric that quantifies overall video analytics performance by integrating recognition accuracy, transmission, and computation efficiency.
- The choice between First-Come-First-Serve (FCFS) and Last-Come-First-Serve-with-Preemption (LCFSP) processing policies impacts AoPI; LCFSP performs better when recognition accuracy and processing load are higher due to beneficial preemption, while FCFS is superior at lower accuracy.
- A Lyapunov-based online optimization framework can effectively manage dynamic edge environments, achieving asymptotically optimal long-term performance without requiring future knowledge of varying conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.