.png){kind=icon}
.png){kind=icon}
Return
Topic: regression-based machine learning approaches for visible light positioning under light dimming conditions
.png){kind=icon}
PRE
AI
DOI：10.1007/s12596-026-03128-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Visible Light Positioning (VLP) is an emerging indoor positioning technology that utilizes the Received Signal Strengths (RSSs) of Light Emitting Diodes (LEDs) to calculate a receiver's location. Unlike traditional radio-frequency-based positioning solutions, VLP is immune to electromagnetic interference, making it particularly valuable in areas such as hospitals, smart homes, and industrial automation. However, one of the most notable challenges in real VLP implementations is the dynamic behavior of illumination particularly due to light dimming which significantly affects RSS measurements and, consequently, increases the localization error. This paper examines the impact of light dimming on VLP performance by analyzing machine-learning (ML) regression models that predict three-dimensional position coordinates from dimming-affected RSS values. Four regression models Multilayer Perceptron (MLP), Support Vector Regression (SVR), Random Forest Regression (RFR), and Gaussian Process Regression (GPR) are trained and tested under dimming factors alpha is an element of {0.2, 0.4, 0.6, 0.8, 1.0} in a simulated indoor environment. Localization performance is quantified using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) in meters. The results show that GPR consistently provides the most accurate and robust positioning across all dimming levels, achieving MAE = 0.15 m at alpha = 0.2 and MAE = 0.14 m at alpha =0.8-1.0, while maintaining RMSE = 0.36 m at alpha =0.2 and RMSE = 0.35 m for alpha = 0.4-1.0. These findings indicate that GPR is a reliable approach for VLP systems operating under light dimming and fluctuating illumination conditions.
Keywords:
Visible light positioning (VLP)
Light dimming
Regression models
Received signal strengths (RSSs)
Journal
J
IF:
2.5
Papers:
216
Citations:
0