Learning from Humans – Building for Humans

网络物理系统在其技术潜力和对不同使用场景的适用性方面正在增加。然而，人与技术系统之间的互动概念是安全、高效和可接受地使用这些系统的关键成功因素。为了能够在这个星座中进行合作，该项目将调查人类感知能力、脑力劳动和人类对CPS的信任之间的相互关系。尤其是用于交通应用的人类网络物理系统(HCPS)在其发展过程中具有高度的社会相关性和动态化。在遵循一个基本的合作原则的基础上，HCPS应该同时利用人和机器的力量，并弥补它们各自的局限性。视觉受限的用户对自动驾驶系统的功能寄予了很高的期望。自动驾驶系统仍然试图在复杂的交通情况下模仿人类的策略。在这个项目中，向人类学习意味着，参与者在视觉障碍下成功驾驶的实验应该提供如何补偿不完整的技术感知的知识。此外，我们还将学习认知负荷与人类决策的关系，以及驾驶员如何补偿不断增加的负荷。第三个方面关注这些因素如何影响对自动驾驶辅助系统的信任。为人而建意味着在此知识的基础上实施适当的人机界面概念，从技术系统向用户提供信息，减少工作量，增加信任，并导致有效的决策行为。目标是一种适应人类在感知、工作量和信任方面的需求的HCPS。为此，该项目的研究伙伴将进行高度同步的驾驶模拟实验，并分享他们特定的测量专业知识。这包括模拟练习、眼球跟踪、功能磁共振测量、自动驾驶实施和交互原型制作。不同复杂性的驾驶场景将在共同努力下定义，由一个合作伙伴实施，并分发给其他合作伙伴。这项工作计划导致了人类因素、认知科学、神经科学和计算机科学之间的跨学科合作。在这些分布式实验中收集的数据将用于指定专门用于感知、认知负荷和信任的模块。这些模块将被集成到一个联合模型中，使用动态贝叶斯网络描述相互关系及其对决策过程的影响。

Cyber-physical systems are increasing in their technical potentials and their applicability for different usage scenarios. Still the interaction concept between human and technical system is a crucial success factor for safe, efficient and acceptable usage of these systems. To enable cooperation in this constellation this project will investigate the interrelation between human perceptive capabilities, mental workload and human trust into the CPS. Especially Human Cyber-Physical Systems (HCPS) for traffic applications are of high societial relevance and dynamic in their development. Following a basic cooperation principle an HCPS should use the strength of both human and machine, and compensates their individual limitations. Users with visual limitations put high expectations into the capabilities of automated driving systems. Automated driving systems still try to mimic human strategies in complex traffic situations.Learning from Humans in this project means that experiments with participants successfully driving under visual impairments should provide knowledge how to compensate incomplete technical sensing. Additionally we will learn how cognitive load is related to human decision making and how increasing load is compensated by drivers. The third aspect focuses how these factors influence trust into an automated driving assistance system. Building for Humans means that based on this knowledge a suitable HMI concept shall be implemented that provides information from the technical system towards the user decreasing workload, increasing trust and leading to efficient decision behaviour. The goal is an HCPS adapted to human requirements regarding perception, workload and trust. For this reason the research partners in this project will conduct highly synchronized driving simulation experiments and share their specific measurement expertise. This includes simulation practice, eye-tracking, fMRI measurement, automated driving implementation and interaction prototyping. The driving scenarios of different complexity will be defined in a joint effort, implemented by one partner and distributed to the other partners. This work plan leads to an interdisciplinary cooperation between human factors, cognitive science, neuroscience, and computer science.The data gathered in these distributed experiments will be used to specify modules dedicated to perception, cognitive load and trust. These modules will be integrated into a joint model using Dynamic Bayesian Networks describing the interrelations and their influences on decision processes.