Modeling of human behaviors in real driving situations

Abstract

A model of human behavior could be used to produce improved human-machine systems and, in particular, an automated highway system to improve the safety of current highways. Pentland and Liu have demonstrated that they can categorize human driving actions very soon after the beginning of the action using dynamic Markov models. The experiment was conducted within a driving simulator instrumented to record driver control input such as steering wheel angle, brake position, and accelerator position. Our work is an attempt to assess the classification accuracy of the modeling approach in real driving situations.