Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Prediction based computation offloading and resource allocation for multi-access ISAC enabled IoT system

Duc-Thuan Le

TL;DR

This work tackles energy-efficient computation offloading for ISAC-enabled IoT under mobility by developing ClusterMan, a prediction-driven framework that uses trajectory- and velocity-related attributes to guide offloading decisions. It couples a greedy baseline (IterGAlg) with a data-driven enhancement (ClusPreAlg) that uses mutual-information feature selection and per-cluster predictive models to better approximate real-world dynamics. The system model integrates local computation, OTFS/velocity-based spectral efficiency (calSE), and an optimization objective to minimize total energy across all devices and tasks. Numerical results on a large VED dataset show that clustering-prediction can reduce prediction errors (MAE by ~10%, MSE to ~3%) and that strong feature subsets yield high predictive accuracy (up to 97%), highlighting practical potential for adaptive, low-latency ISAC-enabled MEC. The study also notes gaps such as lack of a formal convergence bound for the greedy method and sensitivity to initial points, which future work could address to strengthen theoretical guarantees.

Abstract

In the new era of the Internet of Things (IoT), tasks are now being migrated to edge sites closer to data generators. Mobile devices inherently encounter limitations in terms of energy and computational processing capabilities. In high mobility paradigm, ISAC provides a promising foundation for integrating deployment management within dynamic spatial settings. We are interested in applying prediction mechanism to resource allocation management by extracting data attributes, focusing on ISAC related contexts of the trajectory and velocity and making the allocating decision. We present a system design, a theoretical framework and an implementation of the ClusterMan software package. The numerical suggests that the strong clustering subset of feature may yield high accuracy up to 97\% in the prediction results.

Prediction based computation offloading and resource allocation for multi-access ISAC enabled IoT system

TL;DR

This work tackles energy-efficient computation offloading for ISAC-enabled IoT under mobility by developing ClusterMan, a prediction-driven framework that uses trajectory- and velocity-related attributes to guide offloading decisions. It couples a greedy baseline (IterGAlg) with a data-driven enhancement (ClusPreAlg) that uses mutual-information feature selection and per-cluster predictive models to better approximate real-world dynamics. The system model integrates local computation, OTFS/velocity-based spectral efficiency (calSE), and an optimization objective to minimize total energy across all devices and tasks. Numerical results on a large VED dataset show that clustering-prediction can reduce prediction errors (MAE by ~10%, MSE to ~3%) and that strong feature subsets yield high predictive accuracy (up to 97%), highlighting practical potential for adaptive, low-latency ISAC-enabled MEC. The study also notes gaps such as lack of a formal convergence bound for the greedy method and sensitivity to initial points, which future work could address to strengthen theoretical guarantees.

Abstract

In the new era of the Internet of Things (IoT), tasks are now being migrated to edge sites closer to data generators. Mobile devices inherently encounter limitations in terms of energy and computational processing capabilities. In high mobility paradigm, ISAC provides a promising foundation for integrating deployment management within dynamic spatial settings. We are interested in applying prediction mechanism to resource allocation management by extracting data attributes, focusing on ISAC related contexts of the trajectory and velocity and making the allocating decision. We present a system design, a theoretical framework and an implementation of the ClusterMan software package. The numerical suggests that the strong clustering subset of feature may yield high accuracy up to 97\% in the prediction results.
Paper Structure (26 sections, 1 theorem, 13 equations, 8 figures, 2 algorithms)

This paper contains 26 sections, 1 theorem, 13 equations, 8 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. System design
  3. ClusterMan architecture design
  4. ClusterMan interface and interaction specification
  5. System Model
  6. System notations
  7. Task model
  8. Local computation model
  9. Communication model
  10. Problem modeling statement
  11. Proposed Algorithm
  12. The Greedy Algorithm - IterGAlg
  13. Step FI - Feasible Initialization
  14. Step FS - Feasible Solution
  15. Step GIm - Greedy Improvement
  16. ...and 11 more sections

Key Result

proposition thmcounterproposition

The error bound of clustering prediction The boundness of the probability of reward $r_{t+1}$ where it employs the bound of $P(c)$, i.e. $logP(c) \leq \beta$

Figures (8)

  • Figure 1: System Architecture Diagram
  • Figure 2: Pipeline structure of the proposed mechanism
  • Figure 3: Clustering examination
  • Figure 4: The moving distant vs signal radius distant
  • Figure 5: Total energy consumption in various modulation setting
  • ...and 3 more figures

Theorems & Definitions (3)

  • remark thmcounterremark
  • proposition thmcounterproposition
  • proof