Identification for control: Optimal input intended to identify a minimum variance controller

It is well known that the quality of the parameters identified during an identification experiment depends on the applied excitation signal. Prediction error identification using full order parametric models delivers an ellipsoidal region in which the true parameters lie with some prescribed probability level. This ellipsoidal region is determined by the covariance matrix of the parameters. Input design strategies aim at the minimization of some measure of this covariance matrix. We show that it is possible to optimize the input in an identification experiment with respect to a performance cost function of a closed-loop system involving explicitly the dependence of the designed controller on the identified model. In the present contribution we focus on finding the optimal input for the estimation of the parameters of a minimum variance controller, without the intermediate step of first minimizing some measure of the model parameter accuracy. We do this in conjunction with using covariance formulas which are not asymptotic in the model order, which is rather new in the domain of optimal input design. The identification procedure is performed in closed-loop. Besides optimizing the input power spectrum for the identification experiment, we also address the question of optimality of the controller. It is a wide belief that the minimum variance controller should be the optimal choice, since we perform an experiment for designing a minimum variance controller. However, we show that this may not always be the case, but rather depends on the model structure.     

Controller performance benchmarking and tuning using generalised minimum variance control

The use of the generalised minimum variance control law for control loop performance assessment and benchmarking is considered. A novel derivation of the control law enables the link to minimum variance benchmarking to be explored and exploited. The main advantage lies in the generality of the weighted cost index and the simplicity of the results. The only price for this simplicity is the assumption on the choice of weightings. Simple expressions are provided for each of the cost terms that enable performance to be assessed, including the total performance index, variance of error and control signals and variance of weighted signals. These can be used to compare existing (classical) designs with optimal solutions using either models or real time normal operating records.     

Non-linear generalized minimum variance feedback, feedforward and tracking control

A generalized minimum variance control law is derived for the control of nonlinear, possibly time-varying, multi-variable systems. The solution for the tracking and feedback/feedforward control law was obtained in the time-domain using a nonlinear operator representation of the process. The cost index involves both error and control signal costing terms and is normally quadratic but may also involve nonlinear functions. The feedback controller obtained is simple to implement and includes an internal model of the process. The tracking controller can include future reference change information, providing a predictive control capability. In one form the compensator might be considered a nonlinear version of the Smith Predictor that has feedforward action.     

Constrained minimum variance control using hybrid genetic algorithm – An industrial experience

This paper develops a method for minimum variance control of proportional–integral (PI) controllers in the presence of input stochastic noise, the abatement of which is an important issue in many control systems. The underlying objective is to mitigate the effect of input noise in the process output, subject to process inequality constraints. For this purpose, a hybrid genetic algorithm is used. It combines the genetic operations of selection, crossover, and mutation with Newton search. The developed method is applied in an industrial setting to find the optimal controller parameters of different control loops at Falconbridge Smelter in Sudbury, Canada. The optimal parameters significantly improve the performance of the PI controllers.     

Least costly identification experiment for control

All approaches to optimal experiment design for control have so far focused on deriving an input signal (or input signal spectrum) that minimizes some control-oriented measure of plant/model mismatch between the nominal closed-loop system and the actual closed-loop system, typically under a constraint on the total input power. In practical terms, this amounts to finding the (constrained) input signal that minimizes a measure of a control-oriented model uncertainty set. Here we address the experiment design problem from a “dual” point of view and in a closed-loop setting: given a maximum allowable control-oriented model uncertainty measure compatible with our robust control specifications, what is the cheapest identification experiment that will give us an uncertainty set that is within the required bounds? The identification cost can be measured by either the experiment time, the performance degradation during experimentation due to the added excitation signal, or a combination of both. Our results are presented for the situation where the control objective is disturbance rejection only.     

Adaptive generalized minimum variance congestion controller for dynamic TCP/AQM networks

Generic generalized minimum variance-based (GMV) controllers have been adopted as efficient control mechanisms especially in presence of measurement noise. However, such controllers exhibit degraded performance with change in process dynamics. To overcome this problem, a novel congestion controller based on active queue management (AQM) strategy for dynamically varying TCP/AQM networks known as adaptive generalized minimum variance (AGMV) is proposed. AGMV is the combination of the real-time parameter estimation and GMV. The performance of the proposed scheme is evaluated and compared with its adaptive minimum variance (AMV) counterpart under two distinct scenarios: TCP network with unknown parameters and TCP network with time varying parameters. Simulation results indicate that, in either case, AGMV is able to keep the queue length around the desired point. In addition, the superior performance of the proposed controller has been shown with regard to the PI controller, which is well-known in the AQM domain.     

基于小波网络的一类非线性系统稳定最小方差控制

研究一类非线性系统的最小方差控制．将非线性系统等价表示为时变线性系统，利用小波网络的非线性逼近特性在线辨识时变系数，利用改进的投影算法在线调整小波网络的权值；在此基础上设计了非线性系统的最小方差控制器，并分析了闭环控制系统的稳定性．仿真结果表明了该算法的有效性．     

Self-tuning control based on generalized minimum variance criterion for auto-regressive models

Theoretical problems on self-tuning control include stability, performance and convergence of the recursive algorithm involved. In this paper, the problem of controlling minimum or non-minimum phase auto-regressive models with constant but unknown parameters is considered. The stability of an algorithm obtained by combining a recursive estimator for the controller parameters and a generalized minimum variance criterion is proved. The main result is the theorem which assures the overall stability for the closed-loop system in presence of white noise in the input-output relation, where the estimated parameters do not necessarily converge to the true values. The algorithm is proved by the Lyapunov theory.     

最小方差观测器逆设计实现方法

针对最小方差观测器正向设计存在的局限性,提出了一种允许部分极点配置的逆设计方法.该方法由观测器部分极点配置和观测器最小方差化两个环节组成.该文重点研究了逆设计方法提出的理论依据和最小方差化实现的技术细节,并讨论了对噪声协方差矩阵的设定估计问题.数值仿真结果印证了该方法的有效性.     

基于多模型的非线性系统自适应最小方差控制

对于一类典型的离散时间非线性系统, 提出了一种基于多模型的自适应最小方差控制方法. 通过在平衡点附近建立线性模型, 用径向基函数神经元网络来补偿建模误差和未建模动态, 形成了非线性系统的多模型表示. 采用了具有积分性质的切换指标函数作为切换法则和最小方差的控制方法构成了多模型自适应控制器. 仿真实验的结果表明了这种方法的有效性.     

Extension of minimum variance estimation for systems with unknown inputs

In this paper, we address the problem of minimum variance estimation for discrete-time time-varying stochastic systems with unknown inputs. The objective is to construct an optimal filter in the general case where the unknown inputs affect both the stochastic model and the outputs. It extends the results of Darouach and Zasadzinski (Automatica 33 (1997) 717) where the unknown inputs are only present in the model. The main difficulty in treating this problem lies in the fact that the estimation error is correlated with the systems noises, this fact leads generally to suboptimal filters. Necessary and sufficient conditions for the unbiasedness of this filter are established. Then conditions under which the estimation error and the system noises are uncorrelated are presented, and an optimal estimator and a predictor filters are derived. Sufficient conditions for the existence of these filters are given and sufficient conditions for their stability are obtained for the time-invariant case. A numerical example is given in order to illustrate the proposed method.     

非线性广义最小方差控制律综述

如何设计简单的控制策略对复杂非线性系统进行控制是控制界还未解决的难题．非线性广义最小方差控制律的提出使得非线性控制器的设计可以基于更为一般的非线性模型，并且控制器易于实现．整个系统包含时滞环节，稳定的非线性输入子系统和一个可以用多项式或者状态空间描述的子系统．通过最小化由误差加权项、状态加权项和输入加权项组成的信号的方差得到优化控制器．在系统为开环稳定的情况下，可用史密斯预估器进行控制．本文首先介绍了非线性广义最小方差控制的发展进程，然后综述了基于状态空间和多项式描述的系统的非线性广义最小方差控制器的设计以及其现状和今后的发展方向．     

Minimum variance performance map for constrained model predictive control

A minimum variance performance map is introduced for constrained linear model predictive control (MPC). The minimum variance performance map provides a demonstration of the effect of constraints in an MPC on the best achievable controller performance. The constrained minimum variance controller is formulated for the MPC system to be monitored. Using multi-parametric quadratic programming (mp-QP), the linear, piecewise control law is obtained for the constrained minimum variance controller. The linear, piecewise control law is used with a Kalman filter to obtain the minimum output variance in each region of the state space partition. The minimum variance performance map is demonstrated on a second order process with a constraint on the input amplitude.     

On the equivalence of least costly and traditional experiment design for control

In this paper we establish the equivalence between least costly and traditional experiment design for control. We consider experiment design problems for both open and closed loop systems. In open loop, equivalence is established for three specific cases, relating to different parametrisations of the covariance expression (i.e. finite and high order approximations) and model structure (i.e. dependent and independently parameterised plant and noise models). In the closed loop setting, we consider only finite order covariance expressions. H_∞ performance specifications for control are used to determine the bounds on the covariance expression for both the open and closed loop cases.     

Generalized minimum variance control of MIMO systems—A stability result

This note describes a generalized minimum variance (GMV) control algorithm for square systems. The controller embeds an integral action on the error vector. It is shown how an appropriate yet simple choice of some design parameters guarantees overall asymptotic stability for open loop asymptotically stable and possibly nonminimum phase plants.     

Minimum variance control and performance assessment of time-variant processes

This paper is concerned with (1) an explicit solution of a minimum variance control law for linear time-variant (LTV) processes in the transfer function form, and (2) performance assessment of LTV processes using minimum variance control as the benchmark. It is shown that there exists a time-variant, absolute lower bound of process variance that is achievable under LTV minimum variance control and can be estimated from routine operating data. This lower bound can subsequently be used to assess the benefit of implementing LTV control such as adaptive control. The proposed methods are illustrated through simulated examples and an industrial case study.     

CUSUM control charts for monitoring optimal portfolio weights

A portfolio investor requires statistical tools for the timely detection of changes in the optimal portfolio composition. Several multivariate cumulative sum (CUSUM) control charts are proposed for the purpose of monitoring optimal portfolio weights. The ability of the CUSUM schemes to detect important types of changes in the optimal portfolio weights is analyzed in an extensive Monte Carlo simulation study. The empirical application of control charts shows that the proposed methodology can provide a significant reduction of the portfolio volatility.     

Non-linear generalised minimum variance control state space design for a second-order Volterra series model

This paper presents a non-linear generalised minimum variance (NGMV) controller for a second-order Volterra series model with a general linear additive disturbance. The Volterra series models provide a natural extension of a linear convolution model with the nonlinearity considered in an additive term. The design procedure is entirely carried out in the state space framework, which facilitates the application of other analysis and design methods in this framework. First, the non-linear minimum variance (NMV) controller is introduced and then by changing the cost function, NGMV controller is defined as an extended version of the linear cases. The cost function is used in the simplest form and can be easily extended to the general case. Simulation results show the effectiveness of the proposed non-linear method.     

A memory-based gravitational search algorithm for enhancing minimum variance distortionless response beamforming

This paper introduces a memory-based version of gravitational search algorithm (MBGSA) to improve the beamforming performance by preventing loss of optimal trajectory. The conventional gravitational search algorithm (GSA) is a memory-less heuristic optimization algorithm based on Newton’s laws of gravitation. Therefore, the positions of agents only depend on the optimal solutions of previous iteration. In GSA, there is always a chance to lose optimal trajectory because of not utilizing the best solution from previous iterations of the optimization process. This drawback reduces the performance of GSA when dealing with complicated optimization problems. However, the MBGSA uses the overall best solution of the agents from previous iterations in the calculation of agents’ positions. Consequently, the agents try to improve their positions by always searching around overall best solutions. The performance of the MBGSA is evaluated by solving fourteen standard benchmark optimization problems and the results are compared with GSA and modified GSA (MGSA). It is also applied to adaptive beamforming problems to improve the weight vectors computed by Minimum Variance Distortionless Response (MVDR) algorithm as a real world optimization problem. The proposed algorithm demonstrates high performance of convergence compared to GSA and Particle Swarm Optimization (PSO).