Ph.D. received on:  8/7/1997

E-mail: donato@dsi.ing.unifi.it

Tutor: Prof. Edoardo Mosca, Dipartimento di Sistemi e Informatica, Università di Firenze 

 ___________________________________________________________________________________________________________

Supervised Switching Control   ___________________________________________________________________________________________________________

Advisor:

Prof.  Edoardo Mosca, Dipartimento di Sistemi e Informatica, Università di Firenze 

Referee:

Prof. Edoardo Mosca, Dipartimento di Sistemi e Informatica, Università di Firenze
Prof. A. Stephen Morse, Yale University, New Haven (CT), USA.

Summary:

In many practical control problems the plant model  is uncertain. This arises either from real time changes in the operating environment or from components aging or failures. As a consequence fixed controllers based on such models, often are not able  to give good performance, even when highly developed  control strategies  are employed. In order to preserve  a good behaviour in the face of real time changes in the plant, a ‘smart’ controller needs to incorporate an intelligent device, the ‘supervisor’, whose job is to switch into feedback with the plant, in response to such changes the most appropriate controller among a set of candidates.  During the last decades many research efforts have been devoted along this lines by different groups of researchers and many  new schemes and approaches have been proposed.
In my Ph.D. thesis the method proposed by Morse (1991)-(1998) has been considered. In this approach, known as supervisory control the switching is operated by means of an estimator-based supervisor which makes use of a set of known nominal models able to characterize all the plant configurations. The switching is performed at time instants separated by an interval named the dwell-time. The main purpose of the thesis has been the extension to discrete time systems of the analysis previously done for the continuous time case. During this work some differences between the continuous and discrete time cases have been shown, namely regarding the sufficiency condition for convergence of the algorithm. Furthermore, extensions to architectures more optimized for performance, and a robustness analysis have been discussed and analyzed. A final section of the thesis is devoted to illustrate the results of a practical application of  switching control to a non linear solar plant, and some simulations from a case study.
 

MINORS
 
Advisor: Prof. E. Mosca, Università di Firenze

Summary:

In the field of parameter adaptive control, a crucial issue is how to guarantee global convergence of the algorithm in spite of disturbance and unmodeled dynamics. The first part of the Ph.D. has been devoted to the study of techniques such as self-excitation combined with constant-trace identification, aimed at successfully dealing with the singularity problem, arising in these context when the estimated model of the plant loses its stabilizability.

 
 _______________________________________