Machine Learning-Enhanced Visible Light Positioning in IRS-Assisted Indoor Environments with a Single LED Transmitter
Date:
This study explores visible light positioning (VLP) using a single LED transmitter and an intelligent mirror array in confined spaces with obstructed paths. Evaluated algorithms include Maximum Likelihood Estimation (MLE), K-Nearest Neighbors (KNN) Regression, and Neural Networks. MLE showed superior accuracy, while intelligent reflecting surface orientations enhanced receiver localization.
Abstract - This research paper explores the use of a single LED transmitter paired with an intelligent mirror array for power- based visible light positioning (VLP) algorithms in a confined space. The study compares the performance of various algorithms in scenarios where the line-of-sight (LoS) path is obstructed, leaving only non-line-of-sight (NLoS) power available for location estimation. To localize the receiver with just one LED transmitter, we conducted multiple measurements at the investigation point using various intelligent reflecting surface (IRS) orientations. The effectiveness of these orientations was evaluated under different channel noise conditions, comparing their root mean squared error (RMSE) values in estimating the location of the visible light communication (VLC) receiver. Our method- ology incorporates both classical and machine learning-based algorithms, including Maximum Likelihood Estimation (MLE), K-Nearest Neighbors (KNN) Regression, and Fully Connected Neural Networks (FCNN), to process the power measurements from the mirror arrays. The MLE approach’s performance is benchmarked against the Cramer-Rao lower bound to evaluate its precision and reliability. Simulations were conducted to assess the effectiveness of the proposed classical and machine learning- based methods. It was observed that certain IRS orientations, particularly those capable of focusing light on specific room areas, showed enhanced performance in locating the receiver. While the KNN and FCNN algorithms underperformed compared to MLE, they still achieved a level of accuracy influenced by the spatial resolution of their training data. A key advantage of these algorithms is their ability to function without prior knowledge of the channel model, offering increased flexibility in their application.
Index Terms - Intelligent Reflective Surfaces, Mirror Ar- rays, Visible Light Positioning, Location Estimation, Cram ́er- Rao Lower Bound, KNN Regression, Fully Connected Neural Networks.
Sample Figures: