Tutorial introduction

The toolbox estimates a range of model structures, using a range of estimation methods and can accommodate several different data types

Data types

The following data types are supported

• Multivariable time domain data that is regularly or irregularly sampled;

• Multivariable frequency domain data

Model types

The following model types are supported

• Linear transfer function structures

• Hammerstein-Wiener non-linear structures

• Linear state space structures

• Bilinear state space structures

• Linear grey box structures

These may be expressed with regard to the shift operator in all cases, with respect to the operator in the transfer function case, and with respect to the continuous time derivative operator in the linear state space case.

Estimation Methods and Algorithms

The following estimation methods are supported

• Prediction Error Estimation (quadratic criterion)

• Maximum Likelihood Estimation

• Subspace-based estimation

These are implemented via the following algorithms

• A range of gradient based search methods including robust Gauss-Newton, Levenberg-Marquardt, and trust region techniques

• The expectation-maximisation (EM) algorithm

Tutorial overviews and examples

The following links provide tutorial examples illustrating the use of the toolbox across the above range of data types, model types and estimation methods.

• Estimation of Transfer Function Structures

• Estimation of State-Space Structures

• Estimation from Frequency Domain Data

• Estimation of Hammerstein-Wiener Nonlinear Structures

• Estimation from Non regularly sampled Data]