A study of driver taking-over control behavior in automated vehicles
MetadataShow full item record
Automated vehicles (AVs) have received great attention, since they offer the possibility of significantly increasing traffic safety, mobility and driver comfort, and reducing driver workload, congestion and fuel emissions. Current automation technology is still imperfect. Therefore, there will still be situations which the automation will not be able to handle and will request the driver to take over control of the automated vehicle in a limited period. AV design significantly affects driver taking over behavior and influences driving safety and user acceptance accordingly. However, only a few studies have provided information in understanding taking over behavior. The overall objective of this study was to explore how major factors in AV design, including lead time, realistic non-driving related task, human-AV interface, and reliability affect human taking over behavior. At first, a simulated AV platform was built which was able to automatically take over vehicle control and deliver multi-modal take-over request to the driver when it is necessary. With this platform, the majority of properties of take-over request could be investigated. Three experiments were conducted focusing on different aspects of AV design. Results suggested the optimal lead time of the take-over request was no less than 10s across general non-driving related tasks. However, a longer lead time (e.g., no less than 15s) was necessary to achieve optimal driver trust and acceptance even though drivers could successfully take over control with shorter lead time. In addition, when more sensorial modalities were occupied (e.g., watching the video), the non-driving task was mental and/or physically demanding (e.g., playing games), or when the cognitive load was very low (e.g., taking a nap), driver take-over performance was impaired, especially when the take-over request was too late (e.g., lead time was 3s). What is more, the optimal vibration pattern of the tactile interface (back-back-seat-seat) was obtained which could generate the fastest takeover response. And this finding was consistent across all types of non-driving related tasks. Moreover, worse taking over behavior was observed with lower warning reliability and in driving sessions involving misses. Driver response bias was observed, which was, the lower beta value generated better taking over behavior and user acceptance. Results from this study suggest that AV design significantly affects human taking over behavior. The findings of this study provide new findings and recommendations for the design of AVs, which will improve automated driving safety and user acceptance. This study will also drive the development of intelligent transportation systems, infotainment systems, and driver state identifying systems. The simulated automated vehicle platform can be used to address other factors in AV design in future studies.