Analytic Hierarchy Process in Discomfort Cause Recognition With Geospatial Data Correlation

This article presents an Analytic Hierarchy Process (AHP) for the automatic discomfort cause recognition (DCR) in real-life vehicle driving conditions. The measurement system, developed as an Android application, using a three-axis accelerometer and a Global Positioning System (GPS) module, generates 14 Vibro-Related (VR) parameters for each 10s driving interval. Passengers marked discomfort locations using a Geographic Information System (GIS). This data was analyzed together with the additional weather and vehicle data to prioritize and weigh the different factors contributing to discomfort. During the study, 162 rides were performed and almost 11000 driving intervals were categorized as comfortable or un-comfortable. For each uncomfortable interval, a Discomfort Causes (DC) was assigned to one of four classes (class 1: static -road works, poor gravel; class 2: static -pothole, bump (speed bump); class 3: static -old asphalt, cobblestone, better gravel; class 4: dynamic -sudden event, braking, sliding). Several Artificial Intelligence (AI) models were trained and tested for DCR, with top-performing models achieving an overall accuracy of 94%. Afterward, geospatial analyzes confirmed a correlation between specific DC and static road problems. The selected AI model was successfully tested for DC prediction at a new location. This study, for the first time, introduces an integrated AHP approach for DCR, combining transport comfort monitoring, AI for DCR, GIS for spatial data correlations, and AI in GIS for DC prediction at a new location, paving the way for further progress in geospatial data analysis and transportation systems. Tests across diverse real-world cases confirmed the accuracy of the proposed method.