Zhang Ruifo, Tan Zhengyu, Lin Zemin, Zhang Ruiying, Liu Chenhui
School of Design, Hunan University, Changsha 410082, China.
School of Design, Hunan University, Changsha 410082, China; Lushan Lab, Changsha 410082, China.
Accid Anal Prev. 2025 Jul;217:108071. doi: 10.1016/j.aap.2025.108071. Epub 2025 Apr 29.
Trust in automation is crucial for the optimal utilization of advanced driver assistance systems (ADAS). While previous studies have examined trust in automated driving (TiAD) and its impact on behavior, there remains a need to explore how experienced drivers interact with partially automated systems in real-world contexts. This study investigates the trust and behavior of 34 experienced ADAS drivers, divided into trustful and distrustful groups, during on-road driving encompassing six typical scenarios. This study evaluates the initial and final TiAD, situational trust across six driving scenarios; and behaviors, including hands-off the steering wheel, engagement in non-driving-related activities (NDRAs), and visual behavior. Results reveal no significant change in TiAD between pre- and post-driving evaluations, but there are significant differences in TiAD and situational trust across six scenarios between the trustful and distrustful groups. Regarding behavior, trustful drivers exhibit more hands-off events and delay responses to warnings. Both groups engage in risky NDRAs with different patterns, while trustful drivers showing a higher tendency for high-risk NDRAs. Visual behavior analysis shows that trustful drivers spend less time monitoring the driving environment, particularly in complex scenarios such as lane addition/reduction, but more time focusing on the human-machine interface (HMI) overall compared to distrustful drivers. The study also explores the impact of ADAS type and mileage, showing that drivers with advanced functionality exhibit higher trust and reduced monitoring, while mileage influence trust with a turning point at around 3,000 km. With these findings, this study highlights safety risks and proposes strategies to address them. This study is expected to provide insights into trust research and ADAS optimization, enhancing driving safety and user experience.
对自动化的信任对于高级驾驶辅助系统（ADAS）的优化利用至关重要。虽然先前的研究已经考察了对自动驾驶的信任（TiAD）及其对行为的影响，但仍有必要探索经验丰富的驾驶员在现实世界环境中如何与部分自动化系统进行交互。本研究调查了34名经验丰富的ADAS驾驶员在包括六种典型场景的道路驾驶过程中的信任和行为，这些驾驶员被分为信任组和不信任组。本研究评估了初始和最终的TiAD、六种驾驶场景下的情境信任；以及行为，包括双手离开方向盘、参与非驾驶相关活动（NDRAs）和视觉行为。结果显示，驾驶前和驾驶后评估之间的TiAD没有显著变化，但信任组和不信任组在六种场景下的TiAD和情境信任存在显著差异。在行为方面，信任组驾驶员表现出更多双手离开方向盘的情况以及对警告的反应延迟。两组都以不同模式参与有风险的NDRAs，而信任组驾驶员进行高风险NDRAs的倾向更高。视觉行为分析表明，与不信任组驾驶员相比，信任组驾驶员用于监测驾驶环境的时间更少，尤其是在诸如车道增加/减少等复杂场景中，但总体上关注人机界面（HMI）的时间更多。该研究还探讨了ADAS类型和里程数的影响，表明具有高级功能的驾驶员表现出更高的信任并减少了监测，而里程数对信任的影响在约3000公里处有一个转折点。基于这些发现，本研究突出了安全风险并提出了应对策略。预计本研究将为信任研究和ADAS优化提供见解，提高驾驶安全性和用户体验。
