After successful completion of the course, students are able to achieve two targets of data-driven modelling:

1. For a given application problem with existing measurement data both the selection of a suitable model structure as well as the optimal estimation of the optimal model parameters can be performed. Alternative models can be assessed in a quantitative way and can be validated by state-of-the-art methods.
2. Advanced literature on methods for system identification can be acquired and implemented idependently.

System parameter estimation using the least squares method, statistical properties of the estimated parameters, estimation of weighting sequences, the extended least squares method, the generalizes least squares method, parameter estimation in closed loop, design of identification experiments, artificial neural networks, nonlinear optimization

Presentation using white-board and beamer, presentation of computational examples using MATLAB/Simulink, links to current research projects at the department, discussion of results and alternatives.

The lecture is actually held in combination with the respective exercise comprising presentations of computational examples and homeworks.

Oral exam of the complete learning matter, or

Home project completing two identification tasks + basic oral presentation.

Lecture notes for this course are available (institute 325A5). Further literature references in the script.

Opening hours of the secretary's office: Tuesday, Wednesday and Thursday from 09:00 a.m. till 11:00 a.m.

Excellent knowledge in the fields of mathematics and stochastics is expected.