Scholarlink

<p>Alcohol consumption and involvement in non-driving related tasks (NDRT) are the main causes of human error in driving. Identifying and controlling these factors can improve human-autonomous driving systems by increasing their ability to support the driver and reducing the risk of accidents. However, research on takeover performance during autonomous driving has largely examined the effects of alcohol consumption and NDRT in isolation. Thirty young and middle-aged participants were recruited for this study and divided into alcohol and non-alcohol groups (BAC:50&#xa0;mg/100&#xa0;ml vs. 0&#xa0;mg/100&#xa0;ml). A mixed-factor experiment was conducted involving three NDRTs (monitoring the road, watching videos and editing information) and two takeover scenarios (vehicle breakdown and marker disappearance). During the takeover, indicators of lateral and longitudinal driving behavior and indicators of eye-movement attention were measured, along with a post-experimental assessment of trust in the automated driving system by these participants. Cluster analysis revealed that alcohol, video watching, information editing, and vehicle breakdown scenarios were major contributors to hazardous takeovers, which may lead to human takeover errors. The combined effects of alcohol consumption and NDRTs amplified driving impairments and impeded the transition of attention from the NDRTs to the takeover task, thereby reducing takeover quality. Notably, at low drinking levels (0.05 BAC), the interaction between alcohol and NDRTs showed less sensitivity than expected. However, a significant interaction was observed between takeover scenarios and the combined effects of alcohol and NDRTs. Sober drivers were least affected by the takeover scenarios, while alcohol consumption significantly exacerbated their adverse effects on takeover effectiveness. Similarly, increased distraction from NDRTs led to more aggressive lateral driving behaviors under higher situational demands. Trust assessment results indicated a decrease in driver confidence in the automated driving system after alcohol consumption and NDRT engagement. This decrease in trust may reduce reliance on automation. These insights emphasize the inadequacy of driver coping mechanisms when faced with the dual disruptions of alcohol and NDRTs. This study calls for a more detailed exploration and how alcohol, NDRTs, and takeover scenarios (i.e., factors that characterize human error) jointly influence takeover performance. Finally, the theoretical and practical implications of this study are discussed. The findings of this research contribute to advancing effective interventions for preventing human error and fostering the development of human-machine harmony design strategies, thereby enhancing the effectiveness and safety of autonomous vehicle takeovers while promoting greater harmony between vehicle and human operators.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Analyzing the effects of human error on automated driving takeovers

  • 期刊论文

  • 发表日期：

    2025年3月27日

查看全文
  • Zhenhua Yu,
  • Xinyi Zhao,
  • Kang Jiang,
  • Qingqing Deng,
  • Shuangyu Yu,
  • Zhipeng Huang,
  • Jiajia Chen,
  • Yunhong Chen,
  • Linjing Zhai

摘要

Alcohol consumption and involvement in non-driving related tasks (NDRT) are the main causes of human error in driving. Identifying and controlling these factors can improve human-autonomous driving systems by increasing their ability to support the driver and reducing the risk of accidents. However, research on takeover performance during autonomous driving has largely examined the effects of alcohol consumption and NDRT in isolation. Thirty young and middle-aged participants were recruited for this study and divided into alcohol and non-alcohol groups (BAC:50 mg/100 ml vs. 0 mg/100 ml). A mixed-factor experiment was conducted involving three NDRTs (monitoring the road, watching videos and editing information) and two takeover scenarios (vehicle breakdown and marker disappearance). During the takeover, indicators of lateral and longitudinal driving behavior and indicators of eye-movement attention were measured, along with a post-experimental assessment of trust in the automated driving system by these participants. Cluster analysis revealed that alcohol, video watching, information editing, and vehicle breakdown scenarios were major contributors to hazardous takeovers, which may lead to human takeover errors. The combined effects of alcohol consumption and NDRTs amplified driving impairments and impeded the transition of attention from the NDRTs to the takeover task, thereby reducing takeover quality. Notably, at low drinking levels (0.05 BAC), the interaction between alcohol and NDRTs showed less sensitivity than expected. However, a significant interaction was observed between takeover scenarios and the combined effects of alcohol and NDRTs. Sober drivers were least affected by the takeover scenarios, while alcohol consumption significantly exacerbated their adverse effects on takeover effectiveness. Similarly, increased distraction from NDRTs led to more aggressive lateral driving behaviors under higher situational demands. Trust assessment results indicated a decrease in driver confidence in the automated driving system after alcohol consumption and NDRT engagement. This decrease in trust may reduce reliance on automation. These insights emphasize the inadequacy of driver coping mechanisms when faced with the dual disruptions of alcohol and NDRTs. This study calls for a more detailed exploration and how alcohol, NDRTs, and takeover scenarios (i.e., factors that characterize human error) jointly influence takeover performance. Finally, the theoretical and practical implications of this study are discussed. The findings of this research contribute to advancing effective interventions for preventing human error and fostering the development of human-machine harmony design strategies, thereby enhancing the effectiveness and safety of autonomous vehicle takeovers while promoting greater harmony between vehicle and human operators.