Indoor Positioning with CNN and Path-Loss Model Based on Multivariable Fingerprints in 5G Mobile Communication...

Area of Science:

• Wireless communication
• Indoor positioning systems
• Signal processing

Background:

• Fifth generation (5G) systems necessitate advanced indoor positioning.
• Traditional received signal strength-based fingerprinting suffers from poor accuracy due to non-line-of-sight and multipath propagation.
• Existing methods struggle with the complexities of 5G indoor environments.

Purpose of the Study:

• To develop a novel indoor positioning method for 5G systems.
• To enhance positioning accuracy by addressing limitations of traditional signal strength-based approaches.
• To leverage multivariable fingerprints (MVF) for improved location classification and path-loss modeling.

Main Methods:

• Proposed a multivariable fingerprint (MVF) approach using secondary synchronization signals (SSS).
• Developed a convolutional neural network (CNN) for location classification using MVF.
• Integrated MVF with a path-loss model to establish distance-measurement relationships.
• Created a hybrid positioning model combining CNN and path-loss estimations.

Main Results:

• All machine learning algorithms utilizing MVF demonstrated improved accuracy over Reference Signal Receiving Power (RSRP)-only fingerprinting.
• The convolutional neural network (CNN) model exhibited superior performance in experimental settings.
• The hybrid positioning model achieved an average positioning error of 1.47 m.
• The hybrid model offered a 9.26% accuracy improvement compared to the CNN model alone.

Conclusions:

• Multivariable fingerprints (MVF) derived from secondary synchronization signals (SSS) enhance indoor positioning accuracy in 5G networks.
• Machine learning models, particularly CNNs, effectively utilize MVF for improved location estimation.
• The proposed hybrid model integrating CNN and path-loss modeling offers a significant advancement in 5G indoor positioning accuracy.