"Can you rely on me?" Evaluating a Confidence HMI for Cooperative, Automated Driving

Abstract

First concepts for cooperative driving already illustrate the potential of the cooperation between human drivers and automated vehicles. However, the main influencing factors that determine an efficient and effective cooperation still need to be investigated. We therefore examined the effects of displaying the vehicle's intended decision in a critical situation in combination with a confidence value. In a driving simulator study (N=49), the automated vehicle communicated uncertainty in predicting the behavior of a pedestrian and the participants could support the automation by entering their own decision to stop or drive through. The results show that the time until entering the own decision (system override) was longer when a pending system decision was indicated with a confidence value in the HMI. A low confidence value resulted in the longest interaction times. In addition, trust in automation and usability were lower compared to the baseline cooperative HMI without a confidence value.