Modeling Drivers' Distracted Behavior: A Hidden Markov Model Approach

• Activity Number: 23
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Shan Bao and Huimin Xiong*+ and James Sayer
• Companies: University of Michigan Transportation Research Institute and University of Michigan Transportation Research Institute and University of Michigan Transportation Research Institute
• Keywords: Hidden Markov Model , Driver Distraction, Driver Behavior Modeling ; Driver distraction ; Driver behavior modeling ; Cell phone use

Abstract

Signals are generally used to characterize the real-world processes. Hidden Markov Model is a stochastic method that can characterize the statistical properties of these signals.The method has been successfully used in many research domains such as social sciences, psychology, speech recognition, and genetics for pattern recognition. Given the complex dynamic situations of driving, characterizing drivers' behavior is still very challenging. This study presents an analysis of using Hidden Markov Models to model drivers' distracted behavior under naturalistic driving conditions. Vehicle sensor and driver behavior data from a naturalistic driving study were used to build up the model. A total of 108 drivers participated in the driving study. Their behavior and performance with and without cell phone use while driving were extracted and used in this analysis. The hidden Markov model was applied to categorize the driving situations and drivers' activities. Different driving patterns have been observed. The findings suggest that the safety consequences of cell phone use were quite different across drivers and were indeed related to some kind of self-limiting behavior.