Automotive user interface (AUI) evaluation becomes increasingly complex due to novel interaction modalities, driving automation, heterogeneous data, and dynamic environmental contexts. Immersive analytics may enable efficient explorations of the resulting multilayered interplay between humans, vehicles, and the environment. However, no such tool exists for the automotive domain. With AutoVis, we address this gap by combining a non-immersive desktop with a virtual reality view enabling mixed-immersive analysis of AUIs. We identify design requirements based on an analysis of AUI research and domain expert interviews (N=5). AutoVis supports analyzing passenger behavior, physiology, spatial interaction, and events in a replicated study environment using avatars, trajectories, and heatmaps. We apply context portals and driving-path events as automotive-specific visualizations. To validate AutoVis against real-world analysis tasks, we implemented a prototype, conducted heuristic walkthroughs using authentic data from a case study and public datasets, and leveraged a real vehicle in the analysis process.

翻译：汽车用户界面评估因新型交互模态、驾驶自动化、异构数据以及动态环境背景而日益复杂。沉浸式分析技术可高效探索由此产生的人、车与环境之间的多层次交互关系。然而，目前尚无面向汽车领域的此类工具。为此，我们提出AutoVis，通过结合非沉浸式桌面与虚拟现实视图，实现对汽车用户界面的混合沉浸式分析。基于对汽车用户界面研究及领域专家访谈（N=5）的分析，我们确定了设计需求。AutoVis支持在复现的研究环境中，利用虚拟化身、运动轨迹和热力图分析乘客行为、生理状态、空间交互及事件。我们采用上下文入口与驾驶路径事件作为汽车专用可视化方法。为验证AutoVis在真实分析任务中的有效性，我们开发了原型系统，使用案例研究与公开数据集中的真实数据进行启发式走查，并在分析过程中引入了真实车辆。