A study on incident detection model applying APID model, fuzzy logic and traffic pattern

Abstract

The study has designed an incident detection model by utilizing fuzzified APID Model along with traffic pattern, for realizing incident detection that is more efficient and suitable for road environment with lamps. Traffic data used by the model are traffic volume, occupancy and speed collected for 3 months in 5 minute interval by the loop detector installed at the Seoul Ring Road. The APID Model uses upper and lower threshold data detected by the loop detector. However, these data do not reflect road environment selected by the study. Therefore, the study has fuzzified variables to suit single site loop detector configuration, adopted MIN-MAX centroid method for inference method and adopted center of gravity method for defuzzification method. Also, the study has designed an incident pattern by considering compressional wave test, confirmation of road incident dissolution, day of the week and varying traffic occupancy shown by different links. In addition, incident rate and differences in incident pattern values of finally modified APID Model had been used to derive the incident rate threshold value for judging occurrence of an incident. The result of detection rate and false alarm rate test, conducted to comparatively analyze conventional APID Model, fuzzified APID Model and proposed integration model, has shown improved performance by the model suggested by the study.