ANFIS-Based Estimation of Average Luminance Values for Different Road Lighting Systems

The primary criterion in road lighting is to maintain the average luminance (Lavg) value within the ranges specified in the standards; in fact, insufficiency of these levels reduce visibility, whereas excessive levels cause glare, thereby compromising traffic safety. In current inspection processes, the necessity of stopping traffic and performing measurements at multiple points using a luminance meter makes the process time-consuming and operationally challenging. In this study, in order to overcome these limitations, a novel method based on the Adaptive-Network Based Fuzzy Inference System (ANFIS) which estimates the average luminance using only edge-of-road measurements and physical parameters without fully closing the road to traffic, or by applying partial closure when necessary is proposed. To validate the effectiveness of the proposed method, Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models were used as baseline performance references. It was demonstrated that the ANFIS model achieved superior prediction performance with a Mean Absolute Percentage Error (MAPE) of 0.71% on the test dataset, compared to ANN (1.62%) and MLR (4.11%). Within the scope of the current dataset parameters, this study provides a comprehensive preliminary validation for Lavg estimation with high accuracy. In other words, the proposed model represents a road-specific calibration framework and requires revalidation or recalibration before being applied to different road conditions. Additionally, the generalization capability of the model across different road geometries is planned to be evaluated in future studies using larger datasets.