A CSI-Based Indoor Positioning System Using Single UWB Ranging Correction

Area of Science:

• Robotics and Automation
• Wireless Communication Systems
• Sensor Fusion

Background:

• Traditional fingerprint-based indoor localization is time-consuming and labor-intensive.
• Ubiquitous Wireless Fidelity (Wi-Fi) and Ultra-Wideband (UWB) technologies offer potential for cost-effective indoor positioning.
• Existing systems often struggle with accuracy and long-term stability in real-world environments.

Purpose of the Study:

• To develop a ready-to-use indoor localization system that is both accurate and economical.
• To overcome the limitations of traditional fingerprint collection in time and effort.
• To achieve stable, long-term, high-accuracy indoor positioning using minimal UWB hardware.

Main Methods:

• A two-stage approach: initial training and continuous positioning.
• Utilizing Inertial Measurement Unit (IMU) and Pedestrian Dead Reckoning (PDR) for initial positioning and fingerprint collection.
• Training a Convolutional Neural Network (CNN) for position prediction and a Support Vector Regression (SVR) model to adapt UWB ranging to Non-Line-of-Sight (NLoS) conditions.
• Calibrating CNN predictions with corrected UWB ranging data during the positioning stage.

Main Results:

• The proposed system demonstrates stable localization results over extended periods.
• Experimental validation in real environments confirms the system's effectiveness.
• Achieved high accuracy in indoor positioning through continuous calibration.

Conclusions:

• The developed system provides a stable, low-cost, and anti-noise solution for indoor localization.
• The integration of IMU, Wi-Fi, and a single UWB anchor significantly improves positioning efficiency and accuracy.
• This approach effectively addresses the challenges of traditional fingerprint-based localization systems.