Collaborative Research: Multivariable Fixed Order Controller Synthesis and Design from Data

该奖项是根据2009年美国复苏和再投资法案（公法111-5）资助的。本研究旨在发展新的基础理论和有效的设计方法，以解决现有控制理论无法有效处理的实际重要问题：具体而言，该提案将开发用于多变量系统的固定和低阶控制器的设计方法，这些控制器满足多种设计规范，并且基于测量数据而不是基于模型。这是一个非常现实的工程问题，尽管计算机辅助设计取得了重大进展，但仍未解决。反馈控制系统设计的目标是精确调节动态系统中的物理量，如位置、速度、温度、压力、流量和水平，尽管存在潜在过程行为的重大不确定性和知识。本文提出的控制设计问题是为了克服几个突出的挑战。这些是a)需要开发直接使用测量数据的设计方法，因为大多数系统无法使用数学模型；b)低阶或低复杂度控制器的设计，因为现有方法产生的高阶控制器通常是脆弱的或非常敏感的，因此无法实现；c)满足多个用户定义规格的设计。积分控制器和一阶控制器的初步研究结果是有希望的。这项研究将对控制应用产生重大影响，包括化学过程、制造系统、磁盘驱动器、导弹和飞机控制、互联网拥塞控制和包括遗传网络在内的生物控制系统。在这些应用中，几乎没有可用的模型和有效的设计技术来有效地设计简单的控制器。

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).This research aims to develop new fundamental theory and effective design methodologies to address practically important issues that existing control theories cannot effectively handle: specifically the proposal will develop methods for the design of fixed and low order controllers for multivariable systems which satisfy multiple design specifications and which are based on the measured data and not on models. This is a very realistic engineering problem that remains open despite significant progress in computer-aided design. The objective of feedback control system design is to precisely regulate physical quantities such as position, velocity, temperature, pressure, flow-rate and level in dynamic systems despite significant uncertainty and knowledge of the behavior of the underlying processes. The present proposal approaches the control design problem with a view to overcoming several outstanding challenges. These are a) the need to develop design methods which use measured data directly since mathematical models are unavailable for most systems b) the design of low order or low complexity controllers since the existing methods yield high order controllers which are often fragile or acutely sensitive and therefore unimplementable and c) designs which satisfy multiple user defined specifications. Preliminary results on integral and first order controllers look promising. The research should have a significant impact on control applications ranging from chemical processes, manufacturing systems, disk drives, missile and aircraft control, internet congestion control and biological control systems including genetic networks. In each of these applications models are scarcely available and few effective design techniques exist for the effective design of simple controllers.