Genetic Fuzzy Logic Controller-Based Freeway Automatic Incident Detection Models with Selected/Extracted Factors

Abstract

This paper aims to develop automatic incident detection models based on a genetic fuzzy logiccontroller (GFLC). Two approaches are used to overcome the problem that GFLC can not consider too many state variables simultaneously. The first approach is to partially select three variables from all available traffic information (nine variables) as state variables of four GFLC models. The second approach is to extract first three principal components from the original nine variables as state variables of GFLC model, namely the Components model. For comparison, artificial neural network (ANN) incident detection models are also developed. To investigate the applicability of the proposed models, three commonly used indices: detection rate (DR), false alarm rate (FAR) and mean time to detect (MTD) are used to measure their performances. The results show that the Components model outperforms the other incident detection models with DR