Rolf Henriksen

Direct Adaptive Generalized Predictive Control

This paper is concerned with the direct approach of adaptive generalized predictive control. An implicit model With control law parameters is developed. A direct adaptive generalized predictive control algorithm and an improved variant are suggested. Global convergence of the algorithms is analyzed under some assumptions.

Convergence Analysis of Some Decentralized Parameter Estimators

Abstract By prefiltering the input/output data and employing certain decentralized estimation techniques, it is possible to improve the robustness of some estimators significantly. The form of the prefilters generally depends on the estimated model, and the estimator will therefore usually have to be implemented in a bootstrap fashion. The convergence properties of this bootstrap estimator is analyzed for three different estimators based upon, respectively, least squares and instrumental variable methods.

Convergence Aspects of Some Robust Estimators Based Upon Prefiltering of the Input/Output Data

Abstract By prefiltering the input/output data and employing certain decentralized estimation techniques, it is possible to improve the robustness of some estimators significantly. Earlier papers on these techniques have been focused on local convergence properties of certain bootstrap estimators based upon these techniques. This paper is devoted to (1) global convergence properties. and (2) convergence rates when the underlying system is stiff.

Comparison of different parameter estimation methods in flotation processes

Summary In this paper we investigate three different methods for parameter estimation in flotation processes. Using an augmented state vector we have designed two recursive filters for the process, a second-order filter and an extended Kalman filter. The third method which has been investigated is a maximum likelihood method. The system model being used is based on the concept of flotation rates and drainage rates.

Accuracy of Some Robust Estimators Based Upon Prefiltering of the Input/Output Data

Abstract By prefiltering the input/output data and employing certain decentralized estimation techniques, it is possible to improve the robustness of some estimators significantly. Earlier papers on these techniques have been focused on local and global convergence properties, and convergence rates. This paper is devoted to the accuracy properties of the estimators.

Adaptive Level Control: A Laboratory Teaching Experiment

Abstract A laboratory process for teaching in adaptive control is presented. The presentation includes a system description, a description of the software package and its options, and some results obtained from a test study carried out by one of our graduate students.

Application of Online Estimation to a Rougher Flotation Process

Abstract The problem of estimating states and flotation parameters in a bank of flotation cells is investigated. From extensive simulation experiments, an estimator based on an aggregated model of the process has been developed to estimate concentrate grades and aggregate net flotation rate constants. Specific results from the estimators performance in a real flotation plant are reported.

Estimation of Large Scale Implicit Models Using Two-Stage Methods

Abstract The problem of estimating large scale implicit (nonrecursive) models by twostage methods is considered. The first stage of the methods is used to construct or estimate an explicit (recursive) form of the total model, e.g., by constructing a minimal stochastic realization of the system. This model is then subsequently used in the second stage to generate instrumental variables for the purpose of estimating each submodel separately, primarily by ustilizing decentralized filtering algorithms and a prediction error formulation. A note about the connection between the original TSLS-method (two-stage least squares method) and stochastic realization algorithms is also made.

Parameter Estimation in Large Scale Econometric Models Using Decentralized Filtering Algorithms

Abstract The problem of estimating large scale econometric models using decentralized filtering algorithms is considered. The models are assumed to contain observation uncertainties, i.e., errors in the variables. Based on four different decentralized filtering algorithms, parameter estimators for each subsystem are derived using a prediction error formulation. The methods which are derived are in many ways similar to well-known econometric estimation methods, e.g., both OLS and TSLS can be considered to be special cases of the methods which are derived.