Related Experiment Video
Updated: Aug 10, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
600
A TinyML Deep Learning Approach for Indoor Tracking of Assets
Diego Avellaneda1, Diego Mendez1, Giancarlo Fortino2
1School of Engineering, Electronics Engineering Department, Pontificia Universidad Javeriana, Bogotá 110231, Colombia.
Sensors (Basel, Switzerland)
|February 11, 2023
Summary
This study introduces a TinyML-based indoor positioning system using machine learning to classify radio frequency signals. The edge-deployed neural network achieves 88% accuracy, improving indoor location tracking for resource-constrained devices.
Area of Science:
- Computer Science
- Electrical Engineering
- Embedded Systems
Background:
- Traditional Global Positioning System (GPS) lacks accuracy and scalability for indoor environments.
- Radio Frequency (RF) fingerprinting offers an alternative by analyzing signal characteristics for location recognition.
- Machine learning (ML) and TinyML are emerging solutions for resource-constrained embedded devices.
Purpose of the Study:
- To design, implement, and deploy embedded devices for accurate indoor positioning.
- To leverage TinyML and deep learning for classifying Received Signal Strength Indicator (RSSI) data.
- To overcome limitations of traditional positioning systems in indoor settings.
Main Methods:
- Utilized a machine learning approach, specifically a neural network, for RSSI data classification.
- Employed TinyML for implementing ML models on resource-constrained embedded devices.
- Leveraged the Edge Impulse platform for data processing, model training, and deployment.
Main Results:
- Achieved an 88% classification accuracy in identifying object locations using real-time TAG data.
- Demonstrated the feasibility of deploying deep learning models on edge devices for positioning.
- Showcased a 94% accuracy improvement with the addition of a post-processing stage.
Conclusions:
- The developed TinyML-based system provides an accurate and scalable solution for indoor positioning.
- Edge-deployed deep learning effectively addresses external error factors affecting traditional algorithms.
- This approach offers a viable alternative for applications requiring precise indoor localization.
