Authors: Eduardo Cordova*, San Diego State University, Arash Jahangiri, Department of Civil, Construction, and Environmental Engineering San Diego State University, Atsushi Nara, Associate Professor of Geography at San Diego State University, Ming-Hsiang Tsou, Director, Big Data Analytics Program At San Diego State University
Topics: Transportation Geography, Geographic Information Science and Systems
Keywords: Aggressive, Behavior, Dangerous, Driving, Detecting
Session Type: Paper
Start / End Time: 9:35 AM / 10:50 AM
Room: Plaza Court 5, Sheraton, Concourse Level
Presentation File: No File Uploaded
Aggressive driving (AD) is a widespread complicated issue and has lead to many vehicular accidents and deaths. Studying AD from recorded driving data can aid in highlighting unseen details that facilitate this behavior, such as certain road conditions, traffic, and locations that are susceptible to this behavior. The data size of recorded driving data is the main obstacle and focus of this study. An automated approach is necessary as detailed recorded driving data can contain vasts amount of information. The safety pilot model deployment program (SPMD) dataset used in this study consists of over two billion driving records from Ann Arbor, Michigan, with each record containing data such as location, time, speed, heading, and more. The solution to this issue is the Aggressive Driving Index (ADI), which is designed to automatically and systematically detect AD, score, and isolate the flagged driving trajectory and location. The ADI functions by examining the vehicle’s speed, turning, lane changing, braking, and acceleration. These attributes are used to detect and score a segment for AD. The more severe the action, and the more actions deemed to be severe, the higher the ADI will score the driving segment. The advantage of the ADI is its ability to condense several dimensions of measurement into a single score, which in turn allows for the condensing of billions of driving records into thousands of AD instances. The ADI allows for efficient detection, which leaves more time to commit to studying AD behavior.