Analysis of Model Set and Time-Varying Control based on Complexity

1时变系统的实时辨识我们成功地获得了4-SID方法的序列版本，使我们能够跟踪系统的参数变化。本文从实现论的角度阐明了序贯4-SID方法的一些基本结构。这是我们在本研究计划中获得的最显着的结果。2模型复杂性的分析和量化及其与学习的关系我们将学习理论应用于模型集的识别，并获得了满足指定模型精度所需样本的上下界。我们通过这个样本数界定义了模型集的复杂度。我们指出了目前流行的与控制有关的模型集辨识的随机分析方法的一些缺点。3线性变参数系统的控制算法我们在这个问题上得到了一些基本的结果。H ∞控制结构和拓扑方面的模型集已被认为是从LPV的角度来看。此外，我们还考虑了所谓的即时（JIT）建模的应用。然而，我们目前还不能根据（1）和（2）中的结果建立LPV系统的固体控制结构和算法。

1 Real-Time Identification of Time-Varying SystemsWe are successful in obtaining sequential version of the 4-SID method which enables us to track the parameter variation of the system. We have clarified some fundamental structure of sequential 4-SID method from the viewpoint of realization theory. This is the most remarkable result we have obtained in this research program.2 Analysis and Quantification of Model Complexity and Its Relation to LearningWe have applied learning theory to identification of model set and obtained the lower and upper bound of the samples required for guaranteering a prescribed model accuracy. We have defined complexity of model set through this sample number bounds. We have pointed out that the prevailing stochastic analysis of model set identification connected with control has some drawback.3 Control Algorithm for Linear Parameter Varying (LPV) SystemsWe have obtained some fundamental results on this issue. H-infinity control structure and topological aspects of model set have been considered from the LPV point of view. Also, we have considered the application of the so-called just-in-time (JIT) modeling. However, we are unable to establish so far the solid control architecture and algorithm for LPV systems based on the results obtained in (1) and (2).