Patient-cooperative control of variable impedance actuators (PatRiA)

本计画的目的是研究病人合作控制可变阻抗致动器（VIA）的新方法。重点在于一个合成框架的发展，计算反馈控制器，在同一时间，能够提供一个稳定和温和的病人和VIA之间的互动，并保证机器人和病人之间的互动稳定性。开发的合成框架，使其可以用于任何类型的VIA。对于力/转矩控制回路，所研究的合成过程必须考虑致动器输出阻抗的时变行为，以及多变量VIA工厂中的任何耦合。一个新的综合过程是开发，这是基于Hinf-最优控制通过线性矩阵不等式，相互作用的稳定性作为一个额外的约束。在算法收敛问题或过于保守的优化结果的情况下，采用凯莱变换和非光滑Hinf优化的替代方法进行了研究。由于该项目的目标是开发一个病人合作控制，一个可变程度的支持是通过结构固定的级联阻抗控制器的参数化来实现。合成过程必须提供一个空间的控制器参数，保证稳定的机器人和病人之间的相互作用。一种新的方法，通过伯恩斯坦扩展算法使用不确定的传递函数和多项式的积极性将进行研究。此外，应研究能够根据患者的运动状态调整阻抗参数（通过控制和物理交互）的非启发式程序。开发的病人合作的控制策略进行了验证，在两个VIA致动器，在椅子上开发的信息技术。为了估计患者的运动状态，身体传感器网络是可用的。最后，所开发的方法进行了验证，在一个测试台，是专门设计的运动支持的膝关节的主题。

The aim of this project is the investigation of novel methods for patient-cooperative control of variable impedance actuators (VIA). The focus lies on the development of a synthesis framework to compute a feedback controller which is, at the same time, able to provide a stable and gentle interaction between patient and VIA and guarantees interaction stability between robot and patient. The synthesis framework is developed such that it can be used for any kind of VIA. For the force/torque control-loop, the investigated synthesis procedure has to take into account the time-varying behaviour of actuator output impedance, as well as any coupling in the multivariable VIA plant. A novel synthesis procedure is to be developed, that is based on Hinf-optimal control via linear matrix inequalities, with interaction stability as an additional constraint. In case of algorithm convergence problems or too conservative optimisation results, an alternative approach employing a Cayley-transformation and nonsmooth Hinf-optimisation is investigated. Since the goal of this project is to develop a patient-cooperative control, a variable degree of support is to be achieved via parameterisation of a structurally fixed cascaded impedance controller. The synthesis procedure has to provide a space of controller parameters that guarantees stable interaction between robot and patient. A novel approach using uncertain transfer functions and polynomial positivity via the Bernstein expansion algorithm will be investigated. Moreover, non-heuristic procedures should be investigated that are able to adjust the impedance parameters (via control and physical interaction) based on the patient s motor state. The developed patient-cooperative control strategies are validated in two VIA actuators that were developed at the Chair for Information Technology. For the estimation of the patient s motor state, a body sensor network is available. Finally, the developed methods are validated with subjects in a test-bench that was specifically designed for movement support of the knee joint.