An HRI Framework for Designing AV-Pedestrian Interaction

随着自动驾驶汽车（AVs）的出现，我们每天都会遇到自动驾驶汽车与行人的互动，在过马路、使用人行横道或穿过停车场时。通过将自动驾驶汽车视为机器人，自动驾驶汽车与行人的互动成为人机交互（HRI）理论与实践的大规模案例研究，人们在使用城市道路时每天都会与机器人互动，可能会产生严重而严重的后果。该项目以我们集团丰富的HRI经验为基础，重点关注自动驾驶汽车和行人之间互动的挑战和机遇。司机与行人互动时，会使用一套复杂的提示，反映出他们的意图和意识。随着自动驾驶汽车的出现，以及司机的消失，这些提示将不得不被技术取代或改进。我们的项目通过研究自动驾驶汽车和行人之间互动的设计空间来解决这一挑战。我们将自动驾驶汽车与行人的互动视为人与机器人之间社会互动的一个实例，并计划受益于计算HRI中现有的知识体系，以及该领域在社会互动、非语言行为、安全性和机器人界面接受方面的实践和见解。我们计划的方法是建立在允许自动驾驶汽车向行人传达意识和意图的接口设计上，以下三个主题：(1)原型化自动驾驶汽车-行人接口和交互技术。我们将原型化，设计和实现一套自动驾驶汽车和行人之间的接口和交互技术。这些接口将把自动驾驶汽车的感知和意图传达给行人，并将安装在自动驾驶汽车上，由行人携带，嵌入到道路基础设施中，或将这些实现混合在一起。(2)开发沉浸式自动驾驶-行人测试平台。由于安全问题和自动驾驶汽车的不可用性，研究自动驾驶汽车与行人的相互作用是困难的。为了解决这个问题，我们将开发虚拟现实（VR）和增强现实（AR）沉浸式自动驾驶-行人模拟器，并将研究行人在各种环境下与自动驾驶汽车的互动。(3) AV接口评估与框架。我们将在模拟和实际的自动驾驶汽车设置中整合和评估我们的自动驾驶汽车-行人接口和交互技术，开发示范应用系统，工具包和自动驾驶汽车与行人之间交互的理论框架。该计划的三个子项目将提供新颖的界面原型、交互技术、试验台、研究和发现。我们将把这些与计算HRI理论结合起来，形成新的工具包、模型和框架，以理解自动驾驶汽车和行人之间的相互作用。开发的工具、原型和理论将为研究人员、行业和政策制定者提供实用技术，支持自动驾驶汽车与行人之间的互动。长期影响将是安全有效的自动驾驶汽车与行人的互动，以及提高我们社会对自动驾驶汽车的接受程度。

With the emergence of autonomous vehicles (AVs), we will all encounter AV-pedestrian interaction daily, when crossing streets, using crosswalks, or walking through parking lots. By viewing AVs as robots, AV-pedestrian interaction emerges as a massive-scale case study of human-robot interaction (HRI) theory and practice, where people will interact with robots everyday as they use their city roads, with possibly serious and critical consequences. This program focuses on the challenges and opportunities of interaction between AVs and pedestrians, building on our group's extensive HRI experience. Drivers interact with pedestrians using an elaborate set of cues reflecting their intentions and awareness. With AVs, and the absence of drivers, these cues will have to be replaced, or improved, by technology. Our program addresses this challenge by examining the design space of interaction between AVs and pedestrians. We consider AV-pedestrian interaction as an instance of social interaction between people and robots, and plan to benefit from the body of knowledge existing in computational HRI and the domain's practice and insight on social interaction, non-verbal behavior, safety, and acceptance of robotic interfaces. Our program's methodology is founded on the design of interfaces allowing AVs to communicate awareness and intent to pedestrians, following three themes: (1) Prototyping AV-Pedestrian Interfaces and Interaction Techniques. We will prototype, design and implement a set of interfaces and interaction techniques between AVs and pedestrians. The interfaces will communicate AV's awareness and intent to pedestrians, and will be installed on-board AVs, carried by pedestrians, embedded within roads infrastructure, or a mixture of these implementations. (2) Developing Immersive AV-Pedestrian Testbeds. Studying AV-pedestrian interaction is difficult due to safety concerns and AVs unavailability. To address this we will develop virtual reality (VR) and augmented reality (AR) immersive AV-pedestrian simulators, and will examine pedestrian' interaction with AVs in various settings. (3) AV Interfaces Evaluation and Framework. We will integrate and evaluate our AV-pedestrian interfaces and interaction techniques in mock and actual AV setups, developing demonstration application systems, toolkits, and theoretical frameworks for interaction between AVs and pedestrians.  The program's three sub-projects will provide novel interface prototypes, interaction techniques, test-beds, studies, and findings. We will use these in conjunction with computational HRI theories to form new toolkits, models and frameworks for understanding interaction between AVs and pedestrians. The tools, prototypes and theories developed will help researchers, industry, and policymakers with practical technologies supporting interaction between AVs and pedestrians. Long term impact will be safe and effective AV-pedestrian interaction, and improved acceptance of AVs in our society.