Designing for Prediction-Level Collaboration between a Human Driver and an Automated Driving System

Abstract

Although automated driving (AD) systems progress fast in recent years, there are still various corner cases that such systems cannot handle well especially for predicting the behavior of surrounding traffic. This may result in discomfort or even dangerous situations. Results from a previous Wizard-of-OZ study suggest that the collaboration between human and system at the prediction level can effectively enhance the experience and comfort of automated driving. For an in-depth investigation of the confluence between AD and driver, a prototype was implemented in a driving simulator driven by a functional AD system that has been partially validated on the public road. Furthermore, we designed and implemented a gaze-button input for intuitive vehicle referencing and a graphical user interface (GUI) for enhancing the explainability of the AD system. Three typical driving scenarios in which an AD could take advantage of the human driver's anticipation to drive more comfortable and personalized were created for subsequent evaluation.