对偶自适应控制问题研究

项目名称： 对偶自适应控制问题研究

项目编号： No.60874033

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 轻工业、手工业

项目作者： 钱富才

作者单位： 西安理工大学

项目金额： 29万元

中文摘要： 世界充满不确定性，工程系统中的不确定性除了影响系统的性能外还会产生风险，本项目通过与香港中文大学系统工程与工程管理系的合作，研究了双重不确定性下LQG的自适应对偶控制。研究表明，绝大多数情况下系统会受到来自外界的随机干扰，这种不确定性无法控制、消除或减少，另外由于内部元件毁坏、参数漂移或者一些建模时无法确切知道的动态行为，这种不确定性通过学习可以减少甚至完全消除。本项目解决了具有双重不确定性系统控制器设计中的主动学习的最优度量问题，获得了两种学习策略- - 主动学习和被动学习，提出了具有主动学习特点并在统计意义下达到最优的控制器设计方法；对于取随机值的闭环性能指标，利用效用和概率技术，建立了求解不可分问题的优化控制理论；揭示了控制器的工作机理和基本性质；所设计的控制器能够成功追求两个互相冲突的目标:一方面控制要对系统进行优化,另一方面还可以学习以减少部分不确定性，尽管这两者互相冲突，但设计出的自适应对偶控制能够在优化和学习之间进行折衷；创立了标称控制理论，以最优方式打开了当前控制与未来后验概率的耦合环，实现了闭环控制；用对偶自适应理论克服了传统控制理论在处理双重不确定性时的缺陷与不足。

中文关键词： 不确定性理论;最优控制；随机系统;对偶控制.

英文摘要： The world is full of uncertainties.The existence of uncertainty in engineering systems produces risk besides affecting the performance of system.This project studies the adaptive dual control on LQG together with the department of systems engineering and engineering management of the chinese university of Hong Kong. Results show that the systems are subject to random disturbances from the external in most situations. Such uncertainty can not be controlled,cancelled or reduced. Otherwise, due to the internal component destroyed,parameters drifted or some unknown dynamic behavior, this kind of uncertainties can be reduced even eliminated by learning. The project solves the optimal measure problem in ontroller design with active learning feature for systems with dual uncertainties, obtains two types of learn strategies, i.e. active learning and passive learning, proposes a controller design method that is of active learning feature and achieves the optimality in the sense of statistics. For the closed-loop system performance index taking random value, using the utility and probability techniques establishes the related optimization theory to solve the nonseparable problem. This project proclaims the work mechanism and basic properties of controller. Such controller designed can pursue two conflicting objectives: to drive the system toward a desired state,and to perform active learning to reduce the systems reducible uncertainty. Although two functions are conflicting each other, the adaptive dual controller can perform the best balance between the control and the learning, realize the closed-loop control and overcome the defects and insufficiency when using the traditional control theory deals with the dual uncertainties in systems.

英文关键词： Uncertain Theory;Optimal Control; Stochastic Systems; Dual Control.