社会与工程心理学研究室知识库

有人驾驶中，驾驶员可以通过和行人互动来减少通行权的不确定性，但自动驾驶汽车(Autonomous vehicle, AV)本身不具备与行人的沟通能力，需要通过车外交互界面 eHMI (external Human-Machine Interface)来与行人等其他道路使用者发生互动。设计良好的车外交互能够有效弥补行人-车辆通信不足的问题，从而提高交通安全性和公众对自动驾驶汽车的接受度。但现有 eHMI 存在适用场景少、可理解性差等问题。为了解决这些问题，本研究旨在开发一套高可理解性的车外交互图标体系，研究内容包括三部分：提取车外交互人车场景及需要表达的内容、开发 eHMI 图标原型、对已开发的图标原型进行系统性评估。

研究一，系统地提取有关车外交互的场景及交互内容。研究 1a 进行文献分析，总结了 4 个场景共 7 个交互内容。研究 1b 基于焦点小组(N = 7)，获取用户视角下日常生活中的车外交互场景及信息。根据用户的讨论结果，整理出 20 个 交互场景及信息。研究 1c 通过专家访谈(N = 10)，结合研究 1a 与研究 1b 产出的交互场景及信息，根据交互内容的重要性及常见性，整理并提取了一套（共7 个场景13个交互内容）服务于自动驾驶车辆外部界面开发和评估的人车交互场景及交互内容。

研究二，基于研究一产出的车外交互场景及内容，开发若干套具有风格一致性的 eHMI 图标原型。研究 2a 进行文献分析，整理、筛选车外交互相关图标，汇总为图标库。研究 2b 基于焦点小组(N = 7)，从用户角度收集图标方案，鼓励绘图表达。研究 2c 用户、专家、设计人员组成小组，进行参与式图标设计(N = 10)。结合文献分析汇总的图标库与访谈过程中产生的图标概念草图，反复迭代多次、优化表意方式，完成终稿。每个场景内容产出 3 个静态图标 1 个动态图 标，共 52 个图标。

研究三，对研究二产出的 52 个图标进行系统的可理解性评估，产出可理解性最高的一套 eHMI 图标，并对图标进行可记忆性评估。研究 3a 基于文献分析建立了主客观相结合的、多维度的可理解性评估体系。首先以在线问卷的形式进行预实验(N = 34)，从《道路交通标志国家标准 GB5768-2009》中选取常见的交通标志作为实验材料。其中警示类选取 47 个、禁令类选取 48 个、指示类选取 36 个，共 131 个交通标志。根据交通标志熟悉程度与重要程度的平均得分，将交通标志分为了三组：熟悉重要度高分组、熟悉重要度中等组、熟悉重要度低分组。随后从各组中选出 4 个图标共 12 个作为基线图标。正式试验(N = 132) 为在线实验方式，邀请被试使用自由文本填写 52 个设计图标的含义，随后使用 7 分量表从准确(Cao et al., 2021)、生动(Bartneck et al., 2008)、易懂(Laugwitz et al., 2008)、有趣(Laugwitz et al., 2008)、有用(Laugwitz et al., 2008)、令人愉悦 (Laugwitz et al., 2008)、智能(Bartneck et al., 2008)、自然(Bartneck et al., 2008)、 新颖(Laugwitz et al., 2008)这九个维度进行主观评价打分。完成设计图标的可理解性试验后，对 12 个交通图标重复该评估过程。最后选出每个场景中可理解性最高的图标，并对图标可理解性不足的原因进行了分析。研究 3b 为确定图标是否容易被用户记住，在可理解性研究结束后，邀请相同的被试(N = 132)在 20 分 钟、1 天、7 天对图标的可理解性进行复测。发现设计图标正确率始终维持在高于 90%的水平，表明设计图标在记忆性方面性能出色。

本研究结合用户和专家视角，系统分析了车外交互所适用的场景和所要表达的信息内容，在此基础上，我们系统设计了具有风格一致性的 eHMI 图标原型，并对其进行了系统的可理解性评估。本研究的研究成果增强了我们对于图标可理解性的认识，为车厂和政策制定者在选用车外交互图标或设计标准时提供了参考。

In human driving, drivers can reduce right-of-way uncertainty by interacting with pedestrians, but autonomous vehicles (AV) do not have the ability to communicate with pedestrians per se, and need to interact with other roadway users such as pedestrians through an external Human-Machine Interface (eHMI). It needs to interact with other road users such as pedestrians through the eHMI (external Human- Machine Interface). A well-designed external human-machine interface can effectively make up for the problem of insufficient pedestrian-vehicle communication, thus improving traffic safety and public acceptance of self-driving cars. However, existing eHMIs have problems such as few applicable scenarios and poor comprehensibility. In order to solve these problems, this study aims to develop a set of highly comprehensible icon system for eHMI interaction, which consists of three parts: extracting the eHMI interaction human-vehicle scenarios and what needs to be expressed, developing a prototype of eHMI icons, and systematically evaluating the developed prototype icons.

In Study 1, systematically extracts the scenarios and interaction information about eHMI interaction. Study 1a conducted a literature analysis and summarized four scenarios with a total of seven interaction messages. Study 1b was based on focus groups (N = 7) to obtain scenes and information about eHMI interactions in daily life from users' perspectives. Based on the users' discussion results, 20 interaction scenarios and information were organized. Study 1c combines the interaction scenarios and information output from Study 1a and Study 1b through expert interviews (N = 10), and organizes and extracts a set of human-vehicle interaction scenarios and interaction content (a total of 7 scenarios and 13 interaction contents) serving the development and evaluation of external interfaces of self-driving vehicles, based on the importance and commonness of the interaction information.

In Study 2, several sets of eHMI icon prototypes with consistent styles are developed based on the interaction scenarios and contents of Study 1 output. Study 2a conducts literature analysis, organizes and screens eHMI icons, and aggregates them into an icon library. Study 2b Based on focus groups (N = 7), collect icon solutions from users' perspectives and encourage drawing expression. Study 2c Users, experts, and designers formed groups for participatory icon design (N = 10). Combining the icon library summarized from the literature analysis with the icon concept sketches generated during the interview process, the final draft was completed by iterating several times and optimizing the ideographic approach. Each scene content outputs 3 static icons and 1 dynamic icon, totaling 52 icons.

In Study 3, the 52 icons produced in Study 2 were systematically evaluated for comprehensibility, and a set of eHMI icons with the highest comprehensibility was produced, and the icons were evaluated for memorability. Study 3a established a combined subjective and objective, multi-dimensional comprehensibility assessment system based on literature analysis. First, a pre-experiment was conducted in the form of an online questionnaire (N = 34), and common traffic signs were selected from the National Standard for Road Traffic Signs GB5768-2009 as the experimental materials. Among them, 47 were selected from the warning category, 48 from the prohibition category, and 36 from the instruction category, totaling 131 traffic signs. According to the average scores of familiarity and importance of traffic signs, the traffic signs were divided into three groups: high familiarity and importance group, medium familiarity and importance group, and low familiarity and importance group. Four icons totaling 12 were then selected from each group as baseline icons. The formal trial (N = 132) was an online experiment in which subjects were invited to fill in the meanings of the 52 design icons using free text, followed by a 7-point scale ranging from accurate (Cao et al., 2021), vivid (Bartneck et al., 2008), easy to understand (Laugwitz et al., 2008), interesting (Laugwitz et al., 2008), and useful (Laugwitz et al., 2008). ., 2008), useful (Laugwitz et al., 2008), enjoyable (Laugwitz et al., 2008), intelligent (Bartneck et al., 2008), natural (Bartneck et al., 2008), and novel (Laugwitz et al., 2008) dimensions. were scored for subjective evaluation. After completing the comprehensibility test for the design icons, the evaluation process was repeated for the 12 transportation icons. Finally, the icon with the highest intelligibility in each scenario was selected and the reasons for the lack of icon intelligibility were analyzed. Study 3b To determine whether the icons were easily remembered by users, the same subjects (N = 132) were invited to retest the icons' comprehensibility at 20 minutes, 1 day, and 7 days after the comprehensibility study. The design icons were found to consistently maintain a correctness rate higher than 90%, indicating that the design icons performed well in terms of memorability.

This study combines user and expert perspectives to systematically analyze the scenarios applicable to eHMI interactions and the information content to be expressed. Based on this, we systematically designed eHMI icon prototypes with consistent styles and systematically evaluated their comprehensibility. The results of this study enhance our understanding of icon comprehensibility and provide a reference for vehicle manufacturers and policy makers when selecting icons or design standards for eHMI interaction.