System Identification via machine learning

A big issue in the identification of dynamic systems is to obtain low-complexity models of some form. Using concepts and techniques popular in the machine learning community, suitable criteria like the Elastic Net cost have been derived that allow to achieve the desired purposes with high numerical efficiency and satisfactory simulation accuracy in the case of SISO LTI systems. The research activity is to consider also the case of MIMO and/or nonlinear systems.

 ERC Sector:

  • PE7_3 Simulation engineering and modelling
  • PE7_7 Signal processing
  • PE1_19 Control theory and optimisation


  • System identification
  • Machine learning

Research groups