Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic-Relevance Based Sensor Selection for Edge-AI Empowered Sensing Systems

Zhiyan Liu, Kaibin Huang

TL;DR

A novel framework for semantic-relevance-aware sensor selection to achieve optimal end-to-end (E2E) task performance under heterogeneous sensor relevance and channel states is presented.

Abstract

The sixth-generation (6G) mobile network is envisioned to incorporate sensing and edge artificial intelligence (AI) as two key functions. Their natural convergence leads to the emergence of Integrated Sensing and Edge AI (ISEA), a novel paradigm enabling real-time acquisition and understanding of sensory information at the network edge. However, ISEA faces a communication bottleneck due to the large number of sensors and the high dimensionality of sensory features. Traditional approaches to communication-efficient ISEA lack awareness of semantic relevance, i.e., the level of relevance between sensor observations and the downstream task. To fill this gap, this paper presents a novel framework for semantic-relevance-aware sensor selection to achieve optimal end-to-end (E2E) task performance under heterogeneous sensor relevance and channel states. E2E sensing accuracy analysis is provided to characterize the sensing task performance in terms of selected sensors' relevance scores and channel states. Building on the results, the sensor-selection problem for accuracy maximization is formulated as an integer program and solved through a tight approximation of the objective. The optimal solution exhibits a priority-based structure, which ranks sensors based on a priority indicator combining relevance scores and channel states and selects top-ranked sensors. Low-complexity algorithms are then developed to determine the optimal numbers of selected sensors and features. Experimental results on both synthetic and real datasets show substantial accuracy gain achieved by the proposed selection scheme compared to existing benchmarks.

Semantic-Relevance Based Sensor Selection for Edge-AI Empowered Sensing Systems

TL;DR

A novel framework for semantic-relevance-aware sensor selection to achieve optimal end-to-end (E2E) task performance under heterogeneous sensor relevance and channel states is presented.

Abstract

The sixth-generation (6G) mobile network is envisioned to incorporate sensing and edge artificial intelligence (AI) as two key functions. Their natural convergence leads to the emergence of Integrated Sensing and Edge AI (ISEA), a novel paradigm enabling real-time acquisition and understanding of sensory information at the network edge. However, ISEA faces a communication bottleneck due to the large number of sensors and the high dimensionality of sensory features. Traditional approaches to communication-efficient ISEA lack awareness of semantic relevance, i.e., the level of relevance between sensor observations and the downstream task. To fill this gap, this paper presents a novel framework for semantic-relevance-aware sensor selection to achieve optimal end-to-end (E2E) task performance under heterogeneous sensor relevance and channel states. E2E sensing accuracy analysis is provided to characterize the sensing task performance in terms of selected sensors' relevance scores and channel states. Building on the results, the sensor-selection problem for accuracy maximization is formulated as an integer program and solved through a tight approximation of the objective. The optimal solution exhibits a priority-based structure, which ranks sensors based on a priority indicator combining relevance scores and channel states and selects top-ranked sensors. Low-complexity algorithms are then developed to determine the optimal numbers of selected sensors and features. Experimental results on both synthetic and real datasets show substantial accuracy gain achieved by the proposed selection scheme compared to existing benchmarks.

Paper Structure

This paper contains 32 sections, 5 theorems, 42 equations, 8 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. Models and Metrics
  3. Sensing Model
  4. Semantic Matching Model
  5. Feature Pruning-and-Fusion Model
  6. Communication Model
  7. Inference Model and Metrics
  8. Semantic-Relevance Based Sensor Selection Protocol
  9. Query broadcast
  10. Semantic matching and feedback
  11. Sensor selection and RRM
  12. Feature uploading
  13. Feature fusion and inference
  14. E2E Sensing Accuracy Analysis
  15. Sensing Accuracy Conditioned on Sensor Relevance
  16. ...and 17 more sections

Key Result

Theorem 1

(Lower Bound on Conditional Accuracy) Given selected sensors, $\mathcal{S}$ and selected features, $\tilde{\mathcal{D}}$, the classification accuracy conditioned on the sensor relevance, $A(\mathcal{S},\tilde{\mathcal{D}}|\{I_m\})$, is lower bounded by where $\tilde{G}_{\min}\triangleq\min_{\ell\neq\ell'}\Vert\tilde{{\boldsymbol{\mu}}}_{\ell}-\tilde{{\boldsymbol{\mu}}}_{\ell'}\Vert^2_{\tilde{\mat

Figures (8)

  • Figure 1: An ISEA system with semantic-relevance based sensor selection.
  • Figure 2: The proposed semantic-relevance based sensor selection protocol.
  • Figure 3: Illustration of classification margin reduction incurred by irrelevant views in a binary-class case. Therein, the red arrow reflects the margin reduction, which is quantified by $2(1-\rho)\tilde{\delta}_{\max}$ in Theorem 1.
  • Figure 4: Distributions of relevance scores on the ModelNet training dataset fitted by Gaussian probability density functions.
  • Figure 5: Empirical and theoretical accuracy w.r.t. the number of selected sensors. The sensors are ranked by priority indicators designed shortly in Sec. \ref{['sec: optimal_sel_algo']}. The number of classes is set as $L=10$ and the number of features $D=20$.
  • ...and 3 more figures

Theorems & Definitions (8)

  • Theorem 1
  • proof
  • Lemma 1
  • proof
  • Lemma 2
  • Lemma 3
  • Lemma 4
  • proof