Machine-Learning-Based Indoor Localization under Shadowing Condition for P-NOMA VLC Systems

Area of Science:

• Wireless Communication
• Machine Learning
• Indoor Localization

Background:

• Effective agent localization is vital for maintaining communication quality in collaborative tasks.
• Power-domain Non-Orthogonal Multiple Access (P-NOMA) systems require accurate environmental information for efficient power allocation.
• Dynamic environments and signal shadowing pose challenges for real-time agent positioning in P-NOMA.

Purpose of the Study:

• To develop a real-time indoor localization method for agents using visible light communication (VLC).
• To enhance power allocation strategies in P-NOMA systems by integrating accurate agent position estimates.
• To address signal loss due to shadowing using the Euclidean Distance Matrix (EDM).

Main Methods:

• Utilizing a two-way VLC link to receive agent signal power at the base station.
• Employing machine learning algorithms for real-time indoor agent position estimation.
• Implementing the Simplified Gain Ratio Power Allocation (S-GRPA) scheme with a look-up table for resource allocation.
• Applying the Euclidean Distance Matrix (EDM) for localization when signals are shadowed.

Main Results:

• Machine learning algorithms achieved a localization accuracy of 0.19 m.
• The proposed method enables effective power allocation to agents in a dynamic indoor environment.
• Successful estimation of agent locations even in the presence of signal shadowing.

Conclusions:

• The integration of machine learning with VLC provides a robust solution for real-time indoor agent localization.
• Accurate localization significantly improves the performance of P-NOMA systems by enabling precise power allocation.
• The developed approach enhances communication reliability for collaborative tasks in dynamic indoor settings.