Although the literature is rich in numerous approaches for driver modeling, it has been lately discovered that relatively few investigations have been done in replicating driving behaviors using unsupervised driver clustering techniques. This paper suggests a novel driver categorization approach and uses a Model Predictive Controller (MPC) to replicate the driver behavioral patterns. The driver datasets from 34 participants have been gathered through a driving simulator using a cosimulation setup (IPG CarMaker, MATLAB/Simulink, and SUMO) for parameterization and evaluation. The driving data is cleaned using various data processing techniques. The interquartile range is utilized to separate the driving characteristics, and a voting procedure is used to segregate the driver types. Further analysis reveals that the MPC-Driver can significantly match the actual driver in the simulator by accurately replicating the dominant behaviors. The proposed methodology can aid in testing newly developed driver assistance systems with various categories of drivers and contribute towards driver-specific parameter-tuning of multiple components.
A Novel Approach to Classify and Replicate Human Drivers using Model Predictive Control
Proceedings
International Commercial Vehicle Technology Symposium ; 2022 ; Kaiserslautern, Deutschland September 13, 2022 - September 15, 2022
2023-02-28
16 pages
Article/Chapter (Book)
Electronic Resource
German
Human Expertise Helps Computer Classify Images
NTRS | 1991
|Formation Flight Control Using Model Predictive Approach
British Library Conference Proceedings | 2009
|Formation Flight Control Using Model Predictive Approach
AIAA | 2009
|OPTICAL COATINGS: Antireflection coatings replicate moth-eye features
British Library Online Contents | 2007