Access-Point Centered Window-Based Radio-Map Generation Network.

Area of Science:

• Robotics
• Computer Science
• Electrical Engineering

Background:

• Indoor radio-mapping positioning technologies, like fingerprinting, are crucial for real-time object tracking.
• Generating accurate radio-maps requires extensive measurement processes and workflows, increasing costs and complexity.
• Existing methods face challenges in robustness to access-point (AP) location changes and efficient utilization of similar indoor structures.

Purpose of the Study:

• To propose a novel radio-map generation network (RGN) centered around access-point (AP) coordinates using an access-point (AP)-centered window (APCW) approach.
• To minimize costs and enhance the usability and prediction accuracy of radio-maps for indoor positioning.
• To develop a robust and efficient method for generating complete radio-maps from partitioned indoor space learning.

Main Methods:

• Development of an access-point (AP)-centered window (APCW) radio-map generation network (RGN).
• Utilizing adversarial learning methods with APCW as input for partitioned indoor space learning.
• Combining radio-maps from individual APs to generate a comprehensive map, ensuring robustness to AP location changes.

Main Results:

• The proposed APCW-based RGN achieved the highest accuracy among four tested learning models.
• The model demonstrated enhanced radio-map prediction accuracy by extracting radio-map parts based on AP floor plan coordinates.
• A root-mean-square error (RMSE) of 4.01 dBm was achieved, indicating high precision in radio-map generation.

Conclusions:

• The APCW-based RGN is a highly accurate and efficient method for generating indoor radio-maps.
• This approach significantly reduces training time and improves prediction accuracy compared to traditional methods.
• The proposed technique offers robustness and enhanced utilization of similar structures for indoor positioning systems.