不确定性慢时变系统最优自适应鲁棒控制及其应用研究

Due to the non-consistency of the identification criteria with the optimal control goal, and the lack of a definite description of the overall performance of adaptive system, it is very difficult for the classical adaptive control methodology to analyze and design the adaptive control system based on the robust performance optimization criteria. This project is focused on the development of optimal adaptive robust control theory for uncertain slowly time-varying systems. We will concentrate on the following challenging issues: to construct the adaptive control-oriented model set and nominal model persistent identification algorithm for uncertain slowly time-varying systems, to present the optimal closed-loop persistent identification algorithms that are consistent with the optimization of robust performance of the adaptive system and the optimal probing inputs, to study the convergency of the new identification algorithm; to develop the optimal adaptive laws, to analysis the interactions between identification and control in the adaptive systems, to investigate the substantial problems that possibly arise in the new adaptive scheme,such as the condition of convergency and robust stability, the reachable optimal degree of control performance, the influence of performance with higher convergence rate, the computation complexity and the method to reduce complexity.Then, the presented control schemes will be applied to open linkage mechanisms and winding system to overcome the problems of complexity of dynamics and variation of parameters and payload changes.The research will contribute new theory and methodology for synthesis of high performance adaptive control systems for uncertain time-varying systems.

通常的自适应控制由于辨识准则与控制指标不一致和缺乏对鲁棒性能指标的明确描述，往往难以按照鲁棒性能优化的原则进行分析和设计。本项目针对不确定慢时变系统，研究自适应控制的鲁棒性能优化问题。研究与此相关的若干关键问题：面向控制的不确定慢时变系统模型集描述及其持续估计算法，与自适应控制鲁棒性能优化相一致的最优闭环持续辨识方法及最优探测输入设计；辨识的收敛性和加速收敛的方法；适应系统中辨识与控制的相互作用；最优自适应鲁棒控制算法；自适应系统收敛性和稳定性、鲁棒性能指标的描述及可达最优性能、辨识收敛性对鲁棒性能的影响、计算复杂性和如何减少复杂性等根本性问题。建立不确定慢时变系统最优自适应鲁棒控制理论及应用基础。将理论结果应用于开链机构的跟踪控制和卷绕系统的张力控制，解决工况及负载变化、环境干扰情况下控制系统的鲁棒稳定性分析和最优性能综合问题。旨在为不确定时变系统的自适应控制设计提供新的方法和理论依据。

本项目针对不确定慢时变系统，较为深入地研究了最优自适应鲁棒控制及其在复杂工业系统中的应用。一、探明了冻结时间(Frozen-time)系统局部性能与全局性能之间的关系，时变速率与冻结设计的时间窗之间的关系，为构造最优适应控制设计解决了若干关键性问题。二、提出了未知信号的最优稳态跟踪自适应控制新方法。与以往的方法不同，新方法由于辨识准则和控制目标是一致的，构成的自适应律与参数已知的最优控制具有相同的最优性能。解决了通常自适应系统性能难以优化及辨识与控制的相互配合等问题。三、针对非线性相关的信号-响应系统的优化鲁棒设计，提出了一种基于 Kriging 模型和遗传算法的实现方法。该方法可以更合理地反映各类因子与响应之间的复杂作用关系，得到全局性的稳健优化解。四、提出了一种 LED 环形阵列的结构光源设计和分析方法，降低了该类光源设计过程的复杂性。为设计高性能基于图像反馈的控制系统解决了光源设计关键问题。五、将自适应控制理论的最新观点和方法应用于工业系统研发,很好地解决动力学性质复杂、工况变换快速频繁、参数及负载大范围变化等关键技术难题。在此基础上，成功地主持研发了高速高精柔性化全自动LED 固晶机并已实现产业化。