Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Unified Deep Transfer Learning Model for Accurate IoT Localization in Diverse Environments

Abdullahi Isa Ahmed, Yaya Etiabi, Ali Waqar Azim, El Mehdi Amhoud

TL;DR

This work tackles IoT localization across indoor and outdoor settings by proposing a unified deep transfer learning framework that uses RSSI fingerprints to bridge environments. It introduces an encoder-based transfer learning approach to align representations between environments and a multitask unified MLP (U-MLP) to predict environment and coordinates. The encoder-TL yields significant improvements in mean distance error—roughly $17.18\%$ indoors and $9.79\%$ outdoors—with MDEs around $6.65$ m indoors and $361.21$ m outdoors, while U-MLP achieves $9.61$ m indoors and $341.94$ m outdoors. By enabling a single scalable model for both environments, the approach reduces deployment complexity and data requirements for smart-city IoT localization using Wi‑Fi and LoRa RSSI data.

Abstract

Internet of Things (IoT) is an ever-evolving technological paradigm that is reshaping industries and societies globally. Real-time data collection, analysis, and decision-making facilitated by localization solutions form the foundation for location-based services, enabling them to support critical functions within diverse IoT ecosystems. However, most existing works on localization focus on single environment, resulting in the development of multiple models to support multiple environments. In the context of smart cities, these raise costs and complexity due to the dynamicity of such environments. To address these challenges, this paper presents a unified indoor-outdoor localization solution that leverages transfer learning (TL) schemes to build a single deep learning model. The model accurately predicts the localization of IoT devices in diverse environments. The performance evaluation shows that by adopting an encoder-based TL scheme, we can improve the baseline model by about 17.18% in indoor environments and 9.79% in outdoor environments.

A Unified Deep Transfer Learning Model for Accurate IoT Localization in Diverse Environments

TL;DR

This work tackles IoT localization across indoor and outdoor settings by proposing a unified deep transfer learning framework that uses RSSI fingerprints to bridge environments. It introduces an encoder-based transfer learning approach to align representations between environments and a multitask unified MLP (U-MLP) to predict environment and coordinates. The encoder-TL yields significant improvements in mean distance error—roughly indoors and outdoors—with MDEs around m indoors and m outdoors, while U-MLP achieves m indoors and m outdoors. By enabling a single scalable model for both environments, the approach reduces deployment complexity and data requirements for smart-city IoT localization using Wi‑Fi and LoRa RSSI data.

Abstract

Internet of Things (IoT) is an ever-evolving technological paradigm that is reshaping industries and societies globally. Real-time data collection, analysis, and decision-making facilitated by localization solutions form the foundation for location-based services, enabling them to support critical functions within diverse IoT ecosystems. However, most existing works on localization focus on single environment, resulting in the development of multiple models to support multiple environments. In the context of smart cities, these raise costs and complexity due to the dynamicity of such environments. To address these challenges, this paper presents a unified indoor-outdoor localization solution that leverages transfer learning (TL) schemes to build a single deep learning model. The model accurately predicts the localization of IoT devices in diverse environments. The performance evaluation shows that by adopting an encoder-based TL scheme, we can improve the baseline model by about 17.18% in indoor environments and 9.79% in outdoor environments.
Paper Structure (13 sections, 7 equations, 5 figures, 2 tables)

This paper contains 13 sections, 7 equations, 5 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. PROPOSED SYSTEM ARCHITECTURE
  3. Encoder Network
  4. Transfer Learning in Diverse Environment
  5. RSSI Data Collection
  6. Data Preprocessing
  7. System Evaluation Metrics
  8. PERFORMANCE EVALUATION AND DISCUSSION
  9. Simulation Setup and Parameters
  10. Data Exploration and Characterization
  11. Evaluation of the Encoder-based TL framework
  12. Evaluation of the U-MLP framework
  13. CONCLUSION

Figures (5)

  • Figure 1: Multi-Environment model-based indoor and outdoor localization system.
  • Figure 2: Frequency of RSSI Values of Indoor-Outdoor database.
  • Figure 3: Performance of Encoder-based TL Models in Indoor-Outdoor environment.
  • Figure 4: RMSE of MLP Model on Indoor, Outdoor, and Combined Datasets.
  • Figure 5: CDFs of RMSE for Indoor, Outdoor and Combined Datasets.