Research Initiation Award: System Identification for Control Design Applications

本提案涉及动力系统最坏情况识别领域的主题。本研究计划的目标是开发系统识别方法，以与鲁棒控制理论相匹配的方式减少给定动力系统周围的不确定性。将有三个重点领域。第一个研究领域将是使用先验和实验信息相结合的线性时不变系统的算法开发。辨识误差准则的选择将与鲁棒控制设计技术相兼容。本文将利用基于信息复杂性理论的技术，研究识别先验和实验信息的每一个选择的内在困难。第二个研究领域是研究和发展用于确定非线性系统线性化模型的最坏情况识别方法。本研究将发展光滑非线性系统在实验数据为闭环实验时，局部不稳定线性化动力学辨识的内在极限的研究方法。第三个研究领域是离线自适应控制技术的研究。该研究将研究最坏情况系统识别与鲁棒控制理论的合并，目的是开发有效和非保守的方法，利用工具来减少不确定性，并随后设计反馈控制器，以满足给定的性能目标。

This proposal addresses topics in the area of the worst case identification of dynamical systems. The goals of this research program are to develop methods for system identification that reduce the uncertainty surrounding a given dynamical system in a way that is matched to robust control theory. There will be three areas of concentration. The first area of research will be algorithm development for linear time invariant systems using a combination of a priori and experimental information. The error criterion of identification will be chosen to be compatible with robust control design techniques. A study will be made of the intrinsic difficulty of identification of each choice of a priori and experimental information using the techniques of Information Based Complexity theory. The second research area is the study and the development of worst case identification methods for the determination of linearized models of nonlinear systems. The research will develop methods for study intrinsic limits of identification of locally unstable linearized dynamics of smooth nonlinear systems when experimental data is obtained from closed loop experiments. The third research area is the study of offline adaptive control techniques. The research will study the merging of worst case system identification with robust control theory with the goal of developing efficient and nonconservative methods for utilizing tools for the reduction of uncertainty and the consequent design of feedback controllers in order to meet a given performance goal.