Poster 19: Human Factor Modeling of Driver Speed Assistance Using Game Engine: A Learning-Based Approach
(Room Palms Ballroom)
16 Jul 19
5:30 PM
-
7:00 PM
Tracks:
Vehicle Automation Technology
Driver speed assistance (DSA), as a good example of advanced driving-assistance systems (ADAS), aims to improve the driving safety and/or energy efficiency by showing the suggested vehicle speed to the driver of an intelligent vehicle. However, since it is impossible for the driver to control the vehicle and track the suggested speed perfectly, speed tracking errors are inevitably generated, which would degrade the effectiveness of the DSA. In this study, we propose a learning-based approach to model the human factor of DSA, aiming to predict and compensate the speed tracking error in real time. Different drivers are categorized into different groups according to their driving styles by principal component analysis and hierarchical clustering, and nonlinear autoregressive neural network is modeled to predict the speed tracking error generated by each driver. A scenario of cooperative on-ramp merging is built in the game engine, where DSA is designed as a head-up display (HUD) on the intelligent vehicle. Human-in-the-loop simulation is conducted on the driving simulator platform by various volunteer drivers, showing the effectiveness of the proposed learning-based human factor model on predicting and compensating the speed tracking error.