Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Dynamic Indoor Fingerprinting Localization based on Few-Shot Meta-Learning with CSI Images

Jiyu Jiao, Xiaojun Wang, Chenpei Han, Yuhua Huang, Yizhuo Zhang

TL;DR

This work tackles the high data costs and environmental dynamics of indoor fingerprinting localization by introducing a data-efficient few-shot meta-learning framework. It leverages historical localization tasks under a Learning-to-Learn paradigm and incorporates a Wasserstein-distance-based task weighting (W-Dis) to prioritize relevant tasks for knowledge transfer. The system maps CSI-derived planar images to 2D coordinates using a four-convolution-plus-one-full connected network, with an inner/outer optimization scheme designed for rapid adaptation. Empirical results across multiple indoor scenarios demonstrate substantial improvements in Mean Euclidean Distance (MED) and robust cross-area transfer, highlighting the method's practicality for low-data, dynamic environments.

Abstract

While fingerprinting localization is favored for its effectiveness, it is hindered by high data acquisition costs and the inaccuracy of static database-based estimates. Addressing these issues, this letter presents an innovative indoor localization method using a data-efficient meta-learning algorithm. This approach, grounded in the ``Learning to Learn'' paradigm of meta-learning, utilizes historical localization tasks to improve adaptability and learning efficiency in dynamic indoor environments. We introduce a task-weighted loss to enhance knowledge transfer within this framework. Our comprehensive experiments confirm the method's robustness and superiority over current benchmarks, achieving a notable 23.13\% average gain in Mean Euclidean Distance, particularly effective in scenarios with limited CSI data.

Dynamic Indoor Fingerprinting Localization based on Few-Shot Meta-Learning with CSI Images

TL;DR

This work tackles the high data costs and environmental dynamics of indoor fingerprinting localization by introducing a data-efficient few-shot meta-learning framework. It leverages historical localization tasks under a Learning-to-Learn paradigm and incorporates a Wasserstein-distance-based task weighting (W-Dis) to prioritize relevant tasks for knowledge transfer. The system maps CSI-derived planar images to 2D coordinates using a four-convolution-plus-one-full connected network, with an inner/outer optimization scheme designed for rapid adaptation. Empirical results across multiple indoor scenarios demonstrate substantial improvements in Mean Euclidean Distance (MED) and robust cross-area transfer, highlighting the method's practicality for low-data, dynamic environments.

Abstract

While fingerprinting localization is favored for its effectiveness, it is hindered by high data acquisition costs and the inaccuracy of static database-based estimates. Addressing these issues, this letter presents an innovative indoor localization method using a data-efficient meta-learning algorithm. This approach, grounded in the ``Learning to Learn'' paradigm of meta-learning, utilizes historical localization tasks to improve adaptability and learning efficiency in dynamic indoor environments. We introduce a task-weighted loss to enhance knowledge transfer within this framework. Our comprehensive experiments confirm the method's robustness and superiority over current benchmarks, achieving a notable 23.13\% average gain in Mean Euclidean Distance, particularly effective in scenarios with limited CSI data.
Paper Structure (9 sections, 14 equations, 7 figures, 2 tables, 1 algorithm)

This paper contains 9 sections, 14 equations, 7 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Description
  3. Problem Formulation
  4. Enhancing Meta-Learning with Task Importance Weighting via Wasserstein Distance (W-Dis)
  5. Image construction
  6. Experiment Results
  7. Experimental Setup
  8. Performance Analysis
  9. Concluding Remarks

Figures (7)

  • Figure 1: Proposed few-shot meta-learning schematic for indoor localization.
  • Figure 2: Structure of inner model.
  • Figure 3: Planar Layout of the Experimental Scenario B(RPs denoted by green dots).
  • Figure 4: CDF of Localization Errors for Different Algorithms (area 3).
  • Figure 5: Comparison of MED Across Different Algorithms (1-Input 1-Output).
  • ...and 2 more figures