Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RE-POSE: Synergizing Reinforcement Learning-Based Partitioning and Offloading for Edge Object Detection

Jianrui Shi, Yong Zhao, Zeyang Cui, Xiaoming Shen, Minhang Zeng, Xiaojie Liu

TL;DR

The paper addresses real-time object detection on edge devices under tight resource and latency constraints. It introduces RE-POSE, a framework that combines RL-driven non-uniform frame partitioning (RL-DCA) with a DP-based parallel offloading strategy to multiple edge servers. Key contributions include a PPO-based clustering agent with a tailored reward function, a partition-level MCKP formulation for per-block model selection, and a DP-based offloading pipeline that preserves high accuracy within latency budgets. Empirical results on four Jetson Orin NX devices using the PANDA dataset show substantial improvements in mAP under latency constraints, demonstrating practical potential for scalable, accurate edge video analytics in dense scenes.

Abstract

Object detection plays a crucial role in smart video analysis, with applications ranging from autonomous driving and security to smart cities. However, achieving real-time object detection on edge devices presents significant challenges due to their limited computational resources and the high demands of deep neural network (DNN)-based detection models, particularly when processing high-resolution video. Conventional strategies, such as input down-sampling and network up-scaling, often compromise detection accuracy for faster performance or lead to higher inference latency. To address these issues, this paper introduces RE-POSE, a Reinforcement Learning (RL)-Driven Partitioning and Edge Offloading framework designed to optimize the accuracy-latency trade-off in resource-constrained edge environments. Our approach features an RL-Based Dynamic Clustering Algorithm (RL-DCA) that partitions video frames into non-uniform blocks based on object distribution and the computational characteristics of DNNs. Furthermore, a parallel edge offloading scheme is implemented to distribute these blocks across multiple edge servers for concurrent processing. Experimental evaluations show that RE-POSE significantly enhances detection accuracy and reduces inference latency, surpassing existing methods.

RE-POSE: Synergizing Reinforcement Learning-Based Partitioning and Offloading for Edge Object Detection

TL;DR

The paper addresses real-time object detection on edge devices under tight resource and latency constraints. It introduces RE-POSE, a framework that combines RL-driven non-uniform frame partitioning (RL-DCA) with a DP-based parallel offloading strategy to multiple edge servers. Key contributions include a PPO-based clustering agent with a tailored reward function, a partition-level MCKP formulation for per-block model selection, and a DP-based offloading pipeline that preserves high accuracy within latency budgets. Empirical results on four Jetson Orin NX devices using the PANDA dataset show substantial improvements in mAP under latency constraints, demonstrating practical potential for scalable, accurate edge video analytics in dense scenes.

Abstract

Object detection plays a crucial role in smart video analysis, with applications ranging from autonomous driving and security to smart cities. However, achieving real-time object detection on edge devices presents significant challenges due to their limited computational resources and the high demands of deep neural network (DNN)-based detection models, particularly when processing high-resolution video. Conventional strategies, such as input down-sampling and network up-scaling, often compromise detection accuracy for faster performance or lead to higher inference latency. To address these issues, this paper introduces RE-POSE, a Reinforcement Learning (RL)-Driven Partitioning and Edge Offloading framework designed to optimize the accuracy-latency trade-off in resource-constrained edge environments. Our approach features an RL-Based Dynamic Clustering Algorithm (RL-DCA) that partitions video frames into non-uniform blocks based on object distribution and the computational characteristics of DNNs. Furthermore, a parallel edge offloading scheme is implemented to distribute these blocks across multiple edge servers for concurrent processing. Experimental evaluations show that RE-POSE significantly enhances detection accuracy and reduces inference latency, surpassing existing methods.
Paper Structure (29 sections, 20 equations, 5 figures, 1 algorithm)

This paper contains 29 sections, 20 equations, 5 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Overview
  3. RL-DCA Partitioning
  4. Parallel Edge Offloading
  5. Proposed Methodology
  6. Partitioning Scheme
  7. Coarse Detection
  8. Initial Clustering
  9. PPO-based Cluster Adjustment
  10. Algorithm Overview
  11. State Representation
  12. Action Space
  13. Reward Function
  14. Inference Stage
  15. Task Offloading Scheme
  16. ...and 14 more sections

Figures (5)

  • Figure 1: RE-POSE framework overview
  • Figure 2: Experimental setup with 4 Jetson Orin
  • Figure 3: Inference latency and mAP50:95 on PANDA dataset
  • Figure 4: Performance using different ways of coarse detection
  • Figure 5: Performance using simple MeanShift and RL-DCA in RE-POSE