Projektakademie Medizintechnik: Biosignal based prediction of joint movements for mechatronic support systems with adaptive multi domain models

拟议的项目旨在通过扩展神经仿生和生物力学模型的拓扑结构来优化关节肌肉模型的精度，以减少身体支撑设备(如矫形器)操作过程中生物系统和技术系统运动之间的相移。特别是在非常慢(直到完全停止)和非常快的运动中，当前的模型会失败。这可以通过整合摩擦学模型(刷毛、粘滞、摩擦、滞后效应等)来改进。此外，黑盒方法将被应用于识别未知的生物学方面，并将这些方面表示为部分模型。所开发的模型主要由肌电信号驱动，并将与附加信号源(设备与生物系统之间的机械压力等)进行融合。使用学习和优化方法对模型进行在线调整。

The proposed project is supposed to optimize the precision of joint-muscle-models by expanding the neurobionic and biomechanic model topology to reduce the phase-shift between the movement of the biological and technical system during operation of body support devices (e.g. orthosis). Especially during very slow (up to complete stop) and very fast movements, current models fail. This can be improved by the integration of tribological models (bristle, stiction, friction, hysteresis-effects etc.). Additionally, black-box approaches will be applied to identify yet unknown biological aspects and represent those as partial models. The developed models are mainly driven by emg-signals and will be fused with additional signal sources (mechanical pressure between device and biological system etc.). Learning and optimization methods are used to adapt the models online.