Research Initiation Award: Development of a Systematic Approach to Uncertainty Modeling for Robust Control Design

本建议的目的是制定一个 不确定性建模的系统方法， 鲁棒控制的实验数据。 过去 十年来，鲁棒控制理论的发展 导致更严格和适用的设计技术 它允许明确地将植物 不确定性问题公式<$1，2，3，4！ 这些发展提供了一种新的方法， 可用于控制复杂系统， 性能规格是无法想象的 几年前 目前缺失的关键环节 这些理论的发展和他们的 在工业中的应用是一种系统的方法， 性能和不确定性的产生 描述。 准确的重要性 不确定性描述的特征是生动的 在实验的控制器的设计中示出 柔性结构<$5！ 控制器综合为 对电厂不确定性的“准确”描述 系统性能的显著改善， 与其他不太准确的不确定性相比 描述。 拟议的研究将制定一种方法， 理论系统模型的误差归因 和实验数据的不确定性和噪声模型 在线性分式框架中， 用结构价值论的方法论。 这些相同 技术也可以用来开发 性能描述。 的 MU框架将用于分析和 将这些新技术与复杂系统控制器相结合 性能和不确定性描述。 的 改进的物理系统建模， 控制设计的目的将提供一个双重的 优点：改进的鲁棒性和闭环 控制器的系统性能。

The objective of this proposal is to develop a systematic approach to uncertainty modeling from experimental data for robust control. In the past decade, theoretical developments in robust control have lead to more rigorous and applicable design techniques which allow the explicit incorporation of plant uncertainty into the problem formulation ı1,2,3,4!. These developments provide a fresh approach that can be exploited to control complex systems and achieve performance specifications which were unthinkable several years ago. The key link currently missing between these theoretical developments and their application in industry is a systematic approach to the generation of performance and uncertainty descriptions. The importance of accurate characterizations of uncertainty descriptions was vividly shown in the design of controllers for an experimental flexible structure ı5!. Controllers synthesized for "accurate" descriptions of the plant uncertainty yielded significant improvements in system performance when compared with other less accurate uncertainty descriptions. The proposed research will develop a methodology for attributing the error between theoretical system model and experimental data to uncertainty and noise model in a linear fractional framework which is compatible with the structure value mu-methodology. These same techniques can also be employed to develop performance descriptions in for this framework. The mu-framework will then be used to analyze and synthesize complex system controllers with these new performance and uncertainty descriptions. The improved modeling of a physical system for the purpose of control design will provide a two fold benefit: improved robustness properties and closed-loop system performance of the controller.