Ultra-Low-Power, High-Accuracy 434 MHz Indoor Positioning System for Smart Homes Leveraging Machine Learning Models.

Area of Science:

• Electrical Engineering
• Computer Science
• Machine Learning

Background:

• Global Navigation Satellite Systems (GNSS) offer reliable outdoor positioning but lack accuracy indoors.
• Existing indoor positioning systems struggle with low accuracy (5m), high cost, and power consumption.
• Smart homes require cost-effective, high-accuracy (<5m), and low-power indoor localization solutions for mobile and long-term use.

Purpose of the Study:

• To propose and implement an intelligent, accurate, and low-power indoor positioning system for smart homes.
• To leverage Gaussian Process Regression (GPR) with information-theoretic gain for enhanced positioning accuracy.
• To develop a system with dual functionality for both positioning and telemetry data reception.

Main Methods:

• Utilized Time Difference of Arrival (TDOA) with ultra-low-power 434 MHz radio transceivers.
• Implemented a Gaussian Process Regression (GPR) model incorporating information-theoretic gain (reduction in differential entropy).
• Integrated differential circuitry to process time difference pulses for radio frequency (RF) transmitter node localization.

Main Results:

• Achieved high positioning accuracy of 0.68m outdoors and 1.08m indoors using GPR models.
• Demonstrated dual functionality, receiving telemetry data at 250 Kbauds alongside positioning.
• Enabled low-power battery operation (<200 mW) using ultra-low-power CC1101 transceivers and differential amplifiers.

Conclusions:

• The proposed system offers a low-cost, low-power, and high-accuracy solution for indoor localization.
• This technology is crucial for enabling advanced features and public well-being in future smart homes.
• The dual functionality enhances the system's utility and cost-effectiveness for smart home applications.