Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Timely Video Analytics Services at the Network Edge

Xishuo Li, Shan Zhang, Yuejiao Huang, Xiao Ma, Zhiyuan Wang, Hongbin Luo

TL;DR

This work addresses timely video analytics at the network edge by introducing AoPI, a metric that jointly captures transmission, computation efficiency, and recognition accuracy. It derives closed-form AoPI expressions under FCFS and LCFSP policies, showing how policy choice and system rates shape timeliness. An online Lyapunov-based block coordinate descent (LBCD) method is proposed to jointly optimize edge server selection, video configuration, and resource allocation with provable asymptotic optimality. Extensive simulations and testbed experiments demonstrate substantial AoPI reductions (up to 10.94X in some settings) while satisfying long-term accuracy constraints. The approach enables scalable, real-time video analytics at the edge under dynamic network and content conditions.

Abstract

Real-time video analytics services aim to provide users with accurate recognition results timely. However, existing studies usually fall into the dilemma between reducing delay and improving accuracy. The edge computing scenario imposes strict transmission and computation resource constraints, making balancing these conflicting metrics under dynamic network conditions difficult. In this regard, we introduce the age of processed information (AoPI) concept, which quantifies the time elapsed since the generation of the latest accurately recognized frame. AoPI depicts the integrated impact of recognition accuracy, transmission, and computation efficiency. We derive closed-form expressions for AoPI under preemptive and non-preemptive computation scheduling policies w.r.t. the transmission/computation rate and recognition accuracy of video frames. We then investigate the joint problem of edge server selection, video configuration adaptation, and bandwidth/computation resource allocation to minimize the long-term average AoPI over all cameras. We propose an online method, i.e., Lyapunov-based block coordinate descent (LBCD), to solve the problem, which decouples the original problem into two subproblems to optimize the video configuration/resource allocation and edge server selection strategy separately. We prove that LBCD achieves asymptotically optimal performance. According to the testbed experiments and simulation results, LBCD reduces the average AoPI by up to 10.94X compared to state-of-the-art baselines.

Towards Timely Video Analytics Services at the Network Edge

TL;DR

This work addresses timely video analytics at the network edge by introducing AoPI, a metric that jointly captures transmission, computation efficiency, and recognition accuracy. It derives closed-form AoPI expressions under FCFS and LCFSP policies, showing how policy choice and system rates shape timeliness. An online Lyapunov-based block coordinate descent (LBCD) method is proposed to jointly optimize edge server selection, video configuration, and resource allocation with provable asymptotic optimality. Extensive simulations and testbed experiments demonstrate substantial AoPI reductions (up to 10.94X in some settings) while satisfying long-term accuracy constraints. The approach enables scalable, real-time video analytics at the edge under dynamic network and content conditions.

Abstract

Real-time video analytics services aim to provide users with accurate recognition results timely. However, existing studies usually fall into the dilemma between reducing delay and improving accuracy. The edge computing scenario imposes strict transmission and computation resource constraints, making balancing these conflicting metrics under dynamic network conditions difficult. In this regard, we introduce the age of processed information (AoPI) concept, which quantifies the time elapsed since the generation of the latest accurately recognized frame. AoPI depicts the integrated impact of recognition accuracy, transmission, and computation efficiency. We derive closed-form expressions for AoPI under preemptive and non-preemptive computation scheduling policies w.r.t. the transmission/computation rate and recognition accuracy of video frames. We then investigate the joint problem of edge server selection, video configuration adaptation, and bandwidth/computation resource allocation to minimize the long-term average AoPI over all cameras. We propose an online method, i.e., Lyapunov-based block coordinate descent (LBCD), to solve the problem, which decouples the original problem into two subproblems to optimize the video configuration/resource allocation and edge server selection strategy separately. We prove that LBCD achieves asymptotically optimal performance. According to the testbed experiments and simulation results, LBCD reduces the average AoPI by up to 10.94X compared to state-of-the-art baselines.
Paper Structure (24 sections, 7 theorems, 57 equations, 16 figures, 1 table, 3 algorithms)

This paper contains 24 sections, 7 theorems, 57 equations, 16 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. System Model
  4. Frame Uploading Model
  5. Frame Processing Model
  6. Problem Formulation
  7. AoPI Analysis
  8. Average AoPI under FCFS Policy
  9. Average AoPI under LCFSP Policy
  10. Comparison between Two Policies
  11. Online Algorithm Design
  12. Lyapunov Framework
  13. Subproblem 1: Video Configuration Adaptation and Resource Allocation
  14. Subproblem 2: Edge Server Selection
  15. Analysis of LBCD
  16. ...and 9 more sections

Key Result

Theorem 1

If the transmission and computation delay of video frames follow exponential distributions with mean value $1/\lambda$ and $1/\mu$ respectively, the average AoPI of camera $n$ in slot $t$ under FCFS policy is: where $p$ denotes the recognition accuracy.

Figures (16)

  • Figure 1: An illustration of the edge-assisted video analytics system.
  • Figure 2: AoPI evolution under FCFS computation policy.
  • Figure 3: Minimum transmission rate (a) and computation rate (b) required to keep the average AoPI lower than 0.5s under the FCFS policy.
  • Figure 4: AoPI evolution under LCFSP computation policy.
  • Figure 5: Minimum transmission rate (a) and computation rate (b) required to keep the average AoPI lower than 0.5s under the LCFSP policy.
  • ...and 11 more figures

Theorems & Definitions (7)

  • Theorem 1
  • Corollary 4.1
  • Corollary 4.2
  • Theorem 2
  • Theorem 3
  • Lemma 1
  • Theorem 4