Torque-Controlled Humanoid Robot

与人类相当的机器人操作和运动能力、视觉和多感官感知、对环境的认知感知和理解、物理人机交互、从运动动作到认知的不同层次的学习，以及它们与类人机器人的发展和评估，是机器人技术和人工智能研究的主要挑战。尽管取得了很大的进展，但这些令人兴奋的研究领域无论是单独还是相互作用都没有得到充分的探索。所要求的人形机器人旨在成为达姆施塔特工业大学（TU）新成立的人形机器人跨部门实验室的中心研究平台。在那里，它将形成一个新的和特别富有成效的界面，用于机器人和达姆施塔特工业大学全校范围内的主题“人工智能”和“认知科学”的跨学科研究，并有在上述领域特别合格的研究人员参与。为了研究这些相关的研究问题，需要一个装备齐全的仿人机器人，特别是在关节中安装扭矩传感器。例如，操作和运动所需的低级控制与解决由许多中间步骤组成的繁琐任务所需的抽象和符号状态和动作表示之间的相互作用可以使用应用机器人进行研究。这些不同的抽象层次是相互依赖的。一方面，运动技能必须根据抽象目标来学习或完善，另一方面，表征必须被学习，以有效地总结所学习的原语的主要效果。由于在达姆施塔特工业大学提供的任何机器人都没有提供必要的运动和传感器能力以及高度操纵和运动多样性，因此所要求的类人机器人对于此类相关研究课题的调查是必不可少的。达姆施塔特工业大学的“模拟、系统优化和机器人”、“智能自主系统”、“人工智能和机器学习”、“信息处理心理学”和“运动生物力学/运动实验室”等教授职位都很好地涵盖了所需的跨学科专业知识。“智能机器人辅助系统”初级教授职位和独立初级研究小组“人机交互AI算法和认知模型”，均由新兴研究人员领导，以及认知科学中心。

Human-comparable robotic capabilities of manipulation and locomotion, visual and multisensory perception, cognitive perception and understanding of the environment, physical human-robot interaction, learning at different levels from motor action to cognition, as well as their development and evaluation with humanoid robots represent major challenges for research in robotics and artificial intelligence. Despite great progress, these exciting areas of research have not been sufficiently explored either individually or in their interactions. The requested humanoid robot is intended to be the central research platform of the newly created interdepartmental laboratory for humanoid robotics at the Technical University (TU) Darmstadt. There, it will form a new and particularly fruitful interface for interdisciplinary research in robotics and the university-wide profile topics "Artificial Intelligence" and "Cognitive Science" of the TU Darmstadt with the participation of researchers who are particularly qualified in the aforementioned fields. A fully equipped humanoid robot, in particular featuring torque sensors in the joints, is required for the investigation of these interrelated research questions. For example, the interaction between the low-level control required for manipulation and locomotion and the abstract and symbolic state and action representations required for solving tedious tasks consisting of many intermediate steps can be investigated using the applied robot. These different levels of abstraction are mutually dependent. On the one hand, motor skills must be learned or refined with respect to abstract goals, and on the other hand, representations must be learned that efficiently summarize the main effects of learned primitives. Since the necessary motor and sensor abilities in combination with a high degree of manipulation and locomotion diversity are not provided by any of the robots available at the TU Darmstadt, the requested humanoid is indispensable for the investigation of such interrelated research topics. The required interdisciplinary expertise is excellently covered at the TU Darmstadt by the professorships "Simulation, Systems Optimization and Robotics", "Intelligent Autonomous Systems", "Artificial Intelligence and Machine Learning", "Psychology of Information Processing" and "Sport Biomechanics / Locomotion Laboratory", which are each firmly established, the junior professorship "Intelligent Robotic Systems for Assistance" and the independent junior research group "Interactive AI Algorithms & Cognitive Models for Human-AI Interaction", both led by emerging researchers, as well as the Centre for Cognitive Science.