Comparison of Direct and Indirect Identification of a Human Driver Model

Abstract

This paper focuses on recent advancements in the identification of human dynamics in vehicle steering tasks. The field of human driver identification encompasses a plethora of distinct models and methods for parameter identification. This study experimentally compares the most prominent identification approaches in terms of model efficiency (ME). The advantages of indirect identification methods are theoretically discussed and experimentally verified using a dataset comprising extensive measurements from 23 individual drivers. Both theoretical analysis and numerical evaluation of ME corroborate that the indirect identification method proposed by the author is superior to the direct methods employed in many previous studies. The suggested approach is expected to provide a more reliable estimate of drivers’ dynamic parameters, which are desirable for further statistical analyses or machine learning applications in this research domain.