Identification and model analysis for dynamic systems control design with Gaussian process priors

Principal investigator:

Prof. dr. Juš Kocijan

Duration:

Maj 1, 2009 - April 30, 2012

Funding:

ARRS - Slovenian Research Agency (J2-2099)

Abstract:

The purpose of this project is research of experimental modelling methods suited for engineering use in automatic control design. Various methods that can be principally divided into two modelling approaches for nonlinear dynamic systems based on Gaussian process prior are foreseen: black-box and grey-box modelling approaches. Gaussian process prior models are probabilistic nonparametric models and as such only applicable for model predictive control methods. Grey-box Gaussian process prior models would on the other hand make use of control design methods based on parametric models possible. Further a problem of computational burden of learning Gaussian process prior models will be dealt with changes in computational hardware and modifications of computational algorithms. Research on dynamic systems analysis, e.g. system stability, for use in control systems design will be also pursued. All research results will be assessed with case studies on real-life complex systems. A software library of general purposed functions based on programme package Matlab will be also developed for modelling with Gaussian process prior models and control design based on these models.















Publications:

KOCIJAN, Juš, GRANCHAROVA, Alexandra. Gaussian process modelling case study with multiple outputs. Comptes Rendus de l Academie Bulgare des Sciences, 2010, vol. 63, no. 4, 601-607.

AŽMAN, Kristjan, KOCIJAN, Juš. Dynamical systems identification using Gaussian process models with incorporated local models. Eng. appl. artif. intell, 2011, vol. 24, no. 2, 398-408.

GRANCHAROVA, Alexandra, KOCIJAN, Juš, JOHANSEN, Tor Arne. Explicit output-feedback nonlinear predictive control based on black-box models. Eng. appl. artif. intell, 2011, vol. 24, no. 2, 388-397.

PETELIN, Dejan, KOCIJAN, Juš, GRANCHAROVA, Alexandra. On-line Gaussian process model for the prediction of the ozone concentration in the air. Comptes Rendus de l Academie Bulgare des Sciences, 2011, vol. 64, no. 1.

AŽMAN, Kristjan, KOCIJAN, Juš. Fixed-structure Gaussian process model. International Journal of Systems Science, 2009, vol. 40, no. 12, 1253-1262.