Scholarlink

<p>Efficient integration of autonomous vehicles with human-driven vehicles on shared roads necessitates the development of driver models that emulate human behavior. Recognizing the nuanced emotional aspects inherent in human-like driving, it becomes imperative to adopt a qualitative approach for design and evaluation of driver models. In response to the predominant reliance on quantitative analysis in previous studies, which often compare against recorded human driving data, our research emphasizes a qualitative evaluation framework. Through experience-based experiments, we identify and incorporate several qualitative factors of awkward maneuvers influencing human-likeness into the design process of our human-like driver model. Verification of our designed human-like driver model’s resemblance to human behavior is achieved through statistical analysis of qualitative ratings provided by participants. Results indicate a superior level of human similarity compared to alternative driver models, highlighting the efficacy of our qualitative approach.</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

Enhancing Human-Likeness in Autonomous Vehicle Driving by Reducing Awkward Maneuvers

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Jemin Woo,
  • Changsun Ahn

摘要

Efficient integration of autonomous vehicles with human-driven vehicles on shared roads necessitates the development of driver models that emulate human behavior. Recognizing the nuanced emotional aspects inherent in human-like driving, it becomes imperative to adopt a qualitative approach for design and evaluation of driver models. In response to the predominant reliance on quantitative analysis in previous studies, which often compare against recorded human driving data, our research emphasizes a qualitative evaluation framework. Through experience-based experiments, we identify and incorporate several qualitative factors of awkward maneuvers influencing human-likeness into the design process of our human-like driver model. Verification of our designed human-like driver model’s resemblance to human behavior is achieved through statistical analysis of qualitative ratings provided by participants. Results indicate a superior level of human similarity compared to alternative driver models, highlighting the efficacy of our qualitative approach.