Hybrid Eyes: Design and Evaluation of the Prediction-Level Cooperative Driving with a Real-World Automated Driving System

While automated driving systems (ADS) have progressed fast in recent years, there are still various situations in which an ADS cannot perform as well as a human driver. Being able to anticipate situations, particularly when it comes to predicting the behaviour of surrounding traffic, is one of the key elements for ensuring safety and comfort. As humans are still surpassing state-of-the-art ADS in this task, this led to the development of a new concept, called prediction-level cooperation, in which the human can help the ADS to better anticipate the behaviour of other road users. Following this concept, we implemented an interactive prototype, called Prediction-level Cooperative Automated Driving system (PreCoAD), which allows human drivers to intervene in an existing ADS that has been validated on the public road, via gaze-based input and visual output. In a driving simulator study, 15 participants drove different highway scenarios with plain automation and with automation using the PreCoAD system. The results show that the PreCoAD concept can enhance automated driving performance and provide a positive user experience. Follow-up interviews with participants also revealed the importance of making the system's reasoning process more transparent.