Fault diagnosis and tolerance for elastic actuation systems in robotics: physical human-robot interaction

近几十年来，弹性驱动在机器人领域受到越来越多的关注。由于在传动系统中引入了柔性元件，这种执行器显示出安全人机交互和节能运行的有利能力。除了这些优点之外，这种执行器的实际实现比非弹性执行器具有更高的复杂性，并且可能在更关键的操作条件下运行。由于这些问题，故障敏感性可能会增加，但目前还缺乏对故障如何影响人机交互的结构化分析。“弹性驱动系统的故障诊断和容错：物理人机交互”项目旨在了解这些影响并开发能够实现容错交互的控制方法。实验研究了用户的刚度体验，并开发了可靠的人机交互容错控制方法。为此，先前项目阶段的故障诊断方法和心理学研究指导了容错控制算法的发展。通过心理测量和心理物理实验研究了可穿戴机器人的刚度缺陷对用户体验的影响。由此产生的对人类感知和人机交互的见解支持用户和交互模型的改进，并指导控制和自适应算法的设计，从而促进安全可靠的人机交互。这些容错控制器使用先前开发的诊断算法检测刚度故障并主动补偿故障后果。为了检验容错物理人机交互概念的实际可行性，实现并考虑了变刚度致动器和弹性致动膝关节矫形器。利用所开发的控制和自适应算法，向容错人机交互迈出了重要的一步。此外，该项目有助于弹性驱动可穿戴机器人的设计和控制，以及对用户刚度体验的理解。因此，它支持实现直观和可靠的运动援助健康和身体有挑战的个人的长期目标。

In the last decades, elastic actuation receives increasing attention in robotics. Due to introducing flexible elements in the drive train, such actuators show beneficial capabilities for safe human-robot interaction and energy-efficient operation. Besides these advantages, practical implementations of such actuators have higher complexity than non-elastic ones and might be operated in more critical operating conditions. Due to these issues, fault sensitivity can increase but a structured analysis of how faults influence human-robot interaction is missing up to now. The project "Fault diagnosis and tolerance for elastic actuation systems: physical human-robot-interaction" aims at understanding these influences and develops control methods that enable fault tolerant interaction. It experimentally investigates the users’ stiffness experience and develops fault-tolerant control methods for reliable human-robot interaction. To this end, fault diagnosis methods from the previous project phase and psychological studies guide the development of fault-tolerant control algorithms.The influence of stiffness faults in wearable robots on user experience is examined in psychometric and psychophysical experiments. The resulting insights in human perception and human-robot interaction support the improvement of user and interaction models and guide the design of control and adaptation algorithms that facilitate safe and reliable human-robot-interaction. These fault-tolerant controllers detect stiffness faults with the previously developed diagnosis algorithms and actively compensate fault consequences. To examine the practical feasibility of the fault-tolerant physical human-robot interaction concepts, a variable stiffness actuator and an elastically actuated knee orthosis are implemented and considered.With the developed control and adaptation algorithms, an important step towards fault-tolerant human-robot interaction is performed. Moreover, the project contributes to the design and control of elastically actuated wearable robots and the understanding of users’ stiffness experience. Thereby, it supports the long-term goal of achieving intuitive and reliable motion assistance for healthy and physically challenged individuals.