Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Augmenting Channel Simulator and Semi- Supervised Learning for Efficient Indoor Positioning

Yupeng Li, Xinyu Ning, Shijian Gao, Yitong Liu, Zhi Sun, Qixing Wang, Jiangzhou Wang

TL;DR

This work tackles indoor positioning by integrating semi-supervised learning with a biased teacher (SSLB) and an Updated Channel Simulator (UCHS) to efficiently leverage unlabeled data. By extracting channel features such as angle spread and delay spread, and by using KDE-based confidence to weight pseudo-labels, the approach reduces measurement overhead and hyperparameter tuning while achieving strong positioning accuracy. The key innovations are UCHS for environment-specific unlabeled data generation and the biased-teacher SSLB framework that assigns data-driven weights to unlabeled samples in regression. The results demonstrate substantial data-efficiency gains (≈40% reduction in labeling/measurement overhead) and robust performance across ablations, enabling practical deployment in AI-enabled indoor factories.

Abstract

This work aims to tackle the labor-intensive and resource-consuming task of indoor positioning by proposing an efficient approach. The proposed approach involves the introduction of a semi-supervised learning (SSL) with a biased teacher (SSLB) algorithm, which effectively utilizes both labeled and unlabeled channel data. To reduce measurement expenses, unlabeled data is generated using an updated channel simulator (UCHS), and then weighted by adaptive confidence values to simplify the tuning of hyperparameters. Simulation results demonstrate that the proposed strategy achieves superior performance while minimizing measurement overhead and training expense compared to existing benchmarks, offering a valuable and practical solution for indoor positioning.

Augmenting Channel Simulator and Semi- Supervised Learning for Efficient Indoor Positioning

TL;DR

This work tackles indoor positioning by integrating semi-supervised learning with a biased teacher (SSLB) and an Updated Channel Simulator (UCHS) to efficiently leverage unlabeled data. By extracting channel features such as angle spread and delay spread, and by using KDE-based confidence to weight pseudo-labels, the approach reduces measurement overhead and hyperparameter tuning while achieving strong positioning accuracy. The key innovations are UCHS for environment-specific unlabeled data generation and the biased-teacher SSLB framework that assigns data-driven weights to unlabeled samples in regression. The results demonstrate substantial data-efficiency gains (≈40% reduction in labeling/measurement overhead) and robust performance across ablations, enabling practical deployment in AI-enabled indoor factories.

Abstract

This work aims to tackle the labor-intensive and resource-consuming task of indoor positioning by proposing an efficient approach. The proposed approach involves the introduction of a semi-supervised learning (SSL) with a biased teacher (SSLB) algorithm, which effectively utilizes both labeled and unlabeled channel data. To reduce measurement expenses, unlabeled data is generated using an updated channel simulator (UCHS), and then weighted by adaptive confidence values to simplify the tuning of hyperparameters. Simulation results demonstrate that the proposed strategy achieves superior performance while minimizing measurement overhead and training expense compared to existing benchmarks, offering a valuable and practical solution for indoor positioning.
Paper Structure (22 sections, 12 equations, 5 figures, 2 tables)

This paper contains 22 sections, 12 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Problem description
  3. Adapting semi-supervised learning for indoor positioning
  4. Channel feature aided data augmentation
  5. Wireless channel feature extracting method
  6. Angle spread
  7. Delay spread
  8. Unlabeled data generation via UCHS
  9. Semi-Supervised Pseudo-Labeling with a Biased Teacher
  10. The biased teacher training
  11. Confidence-based pseudo-labeling
  12. Weighted loss for student training
  13. Experiments and Results
  14. Scenario deployment
  15. Experimental setup
  16. ...and 7 more sections

Figures (5)

  • Figure 1: The DL model for the indoor positioning.
  • Figure 2: The pipeline of semi-supervised learning with a biased teacher.
  • Figure 3: Distribution of predicted position bias and confidence weight.
  • Figure 4: Positioning accuracy comparison with different labeled samples.
  • Figure 5: Impact of weight magnitudes on the test loss.