Design of Control Systems based on Data-Matching Using Input-Output Measurements

本文考虑了基于输入/输出测量的未知对象闭环系统设计的几个重要问题。主要研究结果如下：我们提出了一种直接的方法，通过从输入/输出测量和期望响应中识别控制器来获得期望控制器的参数，而不是对对象的模型识别。本文提出的方法是对现有VRFT和FRIT方法的扩展。该方法的一个显著特点是能够利用辨识误差检测出非预期的闭环不稳定性。我们解决了FRIT获得的参数的最优性问题，并提供了理论结果，即应用预滤波器可以使我们获得实际的最优参数，以实现最优的跟踪性能。在此理论结果的基础上，我们还将FRIT扩展为更实用的方法，即可以在线性计算中进行优化，这意味着这里得到的结果可以通过更简单的计算来检验。在基于控制器整定的系统辨识中，提出了“通过控制器辨识辨识被控对象”的新概念。也就是说，通过调整包含对象结构的控制器，我们可以得到反映数据信息和给定规格的对象的最优参数。我们已经解决了现有规范控制器和提议的FRIT之间的连接。我们在框架中重新制定了规范控制器。因此，我们从理论上改进了FRIT，即在FRIT中所做的调谐方法从理论上产生了与数据和给定规范相拟合的参数。我们将所提出的方法进行了扩展，得到了在植物动态变化情况下能统一实现控制性能的单一控制器。

We consider several important issues concerned with designing a closed loop system for an unknown plant based on input/output measurements. The main results are as follows.1. We have proposed a direct method to obtain parameters of the desired controller via controller identification from the input/output measurements and the desired response instead of model identification of the plant. The proposed method is an extension of the existing method, i.e., VRFT and FRIT. A salient feature of the proposed method is to be able to detect unexpected closed loop instability by using identification error.2. We addressed the issues on the optimality of the parameter obtained FRIT, and have provided the theoretical result that applying a prefilter enables us to obtain the actual optimal parameter for the achievement of the optimal tracking performance. Based on this theoretical result, we have also expanded FRIT into more practical method in the sense that the optimization can be performed in the linear computation, which implies that the result obtained here can be examined by more simple computation.3. We have proposed a new concept "identification of a plant via controller identification" of system identification based on controller tuning. That is, by tuning a controller including the structure of a plant, we have an optimal parameter of the plant reflecting the information of the data and the given specification.4. We have addressed the connection between an existing canonical controller and proposed FRIT. We have reformulated canonical controllers in our framework. As a result, we have refined FRIT in the sense that the tuning method done in FRIT yields the parameter fitting the data and the given specification from the theoretical points of view.5. We have extended the proposed method to obtain a single controller which uniformly achieves control performance under plant dynamics variation.