In DD control design, once the reference model has been chosen the design follows a well specified algorithm, and thus is completely automated. The choice of reference model must express mathematically either one or some combination of the various control objectives (noise rejection, reference tracking performance, robustness, etc). In order to make a fully automated controller tuning, we would like to automate also the choice of the reference model. The development of such an algorithmic method for the choice of the reference model, given only data that the controller itself can collect without any human intervention, is the purpose of this dissertation.
The student will follow courses on Linear Systems, Control Theory, Optimization, Stochastic Processes and System Identification. His/her work will include the analytical treatment of different design criteria, coding of the method in MatLab and assessment of the performance of the resulting controllers through the analysis of case studies simulated in Simulink, and possibly the development of a prototype coding of the method in a microprocessor.
