Chen Yongkang, Wang Jianmin, Jia Fusheng, Wu Xingting, Xiao Qi, Wang Zhaodong, You Fang
College of Design and Innovation, Tongji University, Shanghai, China.
College of Arts and Media, Tongji University, Shanghai, China; Shenzhen Research Institute, Sun Yat-Sen University, Shenzhen, China.
Accid Anal Prev. 2025 Jul;217:108067. doi: 10.1016/j.aap.2025.108067. Epub 2025 Apr 25.
Before full achieving automation, Autonomous Vehicle(AV) must undergo a transitional phase of human-machine collaborative driving. Therefore, designing appropriate Human-Machine Interface (HMI) modes of collaboration is key to ensuring both driving safety and user experience. However, existing research has rarely considered the design of human-machine collaboration modes under different Hazard Visibility scenarios. In this study, we conducted a simulated driving experiment (N = 28) to explore the effects of three HMI-based collaboration modes (HMI1, HMI2, and HMI3) on driving behavior and subjective perception under two hazard visibility scenarios (visible and invisible hazard). The designs of the three collaboration modes were primarily based on varying levels of explainability and control. The results show that in the invisible hazard scenario, drivers exhibited significantly lower situation awareness and preference compared to the visible hazard scenario. The design of HMI in different collaboration modes significantly influences drivers' situation awareness, cognitive workload, trust, and attention distribution, with the highest satisfaction reported for HMI2 (high explainability, AV-led decision-making). Particularly in the invisible hazard scenario, HMI2 significantly improved drivers' situation awareness and attention while minimizing cognitive workload. The study also indicates that during autonomous driving, drivers require a certain sense of control, though this does not necessarily mean they need to directly participate in decision-making. Instead, a sense of control can be fostered by augmenting the explainability of the HMI. These findings provide valuable insights for the design of human-machine interfaces in AV to enhance driving safety.
在完全实现自动化之前，自动驾驶汽车必须经历人机协同驾驶的过渡阶段。因此，设计合适的人机协作界面（HMI）模式是确保驾驶安全和用户体验的关键。然而，现有研究很少考虑不同危险可见性场景下的人机协作模式设计。在本研究中，我们进行了一项模拟驾驶实验（N = 28），以探究三种基于HMI的协作模式（HMI1、HMI2和HMI3）在两种危险可见性场景（可见危险和不可见危险）下对驾驶行为和主观感知的影响。三种协作模式的设计主要基于不同程度的可解释性和控制权。结果表明，在不可见危险场景中，与可见危险场景相比，驾驶员的态势感知和偏好显著降低。不同协作模式下的HMI设计对驾驶员的态势感知、认知工作量、信任度和注意力分配有显著影响，其中HMI2（高可解释性，自动驾驶主导决策）的满意度最高。特别是在不可见危险场景中，HMI2显著提高了驾驶员的态势感知和注意力，同时将认知工作量降至最低。该研究还表明，在自动驾驶过程中，驾驶员需要一定的控制感，尽管这不一定意味着他们需要直接参与决策。相反，可以通过增强HMI的可解释性来培养控制感。这些发现为自动驾驶汽车人机界面的设计提供了有价值的见解，以提高驾驶安全性。
