¹Department of Mathematics
University of Wisconsin, Whitewater
Whitewater, Wisconsin
²Department of Industrial Engineering & Management
Oklahoma State University
Stillwater, Oklahoma
Title
A Machine Learning Framework for Driver Behavior Analysis Using CARLA Simulation Data
Synopsis
This paper analyzes the driving patterns of seven drivers in the CARLA simulation environment. We performed data preprocessing and feature engineering to extract acceleration, gyroscope, and jerk magnitudes, as well as rolling variability to capture driver behavioral change over time. The unsupervised k-means clustering identified smooth, moderate, and aggressive driving behavior categories, and a Pearson chi-square test confirmed a statistically significant association between driver identity and cluster membership, though Cramér’s V indicated a modest effect size.