Enhancing Highway Safety Analysis through Statistical Techniques and On-Board Vehicular Data

Highway safety remains a paramount concern, with many accidents attributed to suboptimal pavement conditions. Acquiring essential condition data for a thorough safety analysis often proves challenging, given the associated data collection costs and data quality constraints. To tackle this challenge, this study harnessed in-vehicle sensor data, commonly known as Onboard Diagnostics (OBD), obtained from Interstate 35 in Oklahoma. The research objective is to utilize onboard vehicle diagnostic data, including metrics such as wheel speed, engine power, acceleration, roll, pitch, yaw, alongside latitude and longitude coordinates, to pinpoint accident-prone locations. Disparities were observed in the speeds of the four distinct wheels on vehicles, exerting a consequential impact on the distance traveled and thereby influencing wheel slip and the geometric curvature of roadways – two pivotal factors in highway crashes. To forge a direct link between the OBD data and pavement surface skid resistance and real crash incidents, skid data were provided by the Oklahoma Department of Transportation's (ODOT) and crash data was collected from the Oklahoma highway collision database. Subsequently, the analytical approach encompassed an initial step, aligning location data from OBD records with corresponding crash locations. Multivariant statistical models were then developed to unveil the intricate relationships between OBD parameters and crash data. This research could provide a valuable opportunity to enhance the comprehension of highway crashes through the information ingrained within vehicle sensor data.