Control design of uncertain systems with few prior assumptions has been recognized as a challenging problem in robust control of widespread importance for a long time. Alongside its advantages, several open questions have surfaced: as current analysis and synthesis methods in robust control are mainly based on prior assumptions given in the form of models, when those models are replaced or enhanced by data, the problem of describing the uncertainty, analyzing robustness, and designing controllers becomes critical. Since the mid‐1990s, a control paradigm termed data‐driven control has been gaining growing attention for both academic research and its practical applicability reasons. Instead of relying on prior assumptions about the plant and uncertainty, fully data‐driven control tries to utilize the information extracted from raw measured plant input–output data to guide the selection of controllers in order to improve control system performance. Prior knowledge of the plant, such as assumed plant and disturbance models, are avoided except when used in the design of candidate controllers. This control design paradigm, based on the principle of unfalsification, represents an increasingly promising method for achieving robustness against model mismatch problems that have been known to plague numerous model‐based techniques. At the heart of it is the notion of cost detectability—the design of data‐driven cost functions, which, in conjunction with the chosen candidate controller set, appropriately represent control performance and stability goals. Dedicated to Professor Michael Safonov on the occasion of his 70th birthday, this Special Issue aims to report current state‐of‐the‐art in data‐driven robust control and draw further research attention into different aspects of this methodology. We hope that this sample of relevant papers will form a catalyst that will lead to a more unified understanding of data‐driven approach to improving robustness, where models are replaced or supplemented by data.

Data-driven robust control: Special issue dedicated to the 70th birthday of Michael G. Safonov / Jin, Huiyu; Stefanovic, Margareta; Tesi, Pietro. - In: INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL. - ISSN 1099-1239. - STAMPA. - (2018), pp. 3665-3666. [10.1002/rnc.4239]

Data-driven robust control: Special issue dedicated to the 70th birthday of Michael G. Safonov

Tesi, Pietro
2018

Abstract

Control design of uncertain systems with few prior assumptions has been recognized as a challenging problem in robust control of widespread importance for a long time. Alongside its advantages, several open questions have surfaced: as current analysis and synthesis methods in robust control are mainly based on prior assumptions given in the form of models, when those models are replaced or enhanced by data, the problem of describing the uncertainty, analyzing robustness, and designing controllers becomes critical. Since the mid‐1990s, a control paradigm termed data‐driven control has been gaining growing attention for both academic research and its practical applicability reasons. Instead of relying on prior assumptions about the plant and uncertainty, fully data‐driven control tries to utilize the information extracted from raw measured plant input–output data to guide the selection of controllers in order to improve control system performance. Prior knowledge of the plant, such as assumed plant and disturbance models, are avoided except when used in the design of candidate controllers. This control design paradigm, based on the principle of unfalsification, represents an increasingly promising method for achieving robustness against model mismatch problems that have been known to plague numerous model‐based techniques. At the heart of it is the notion of cost detectability—the design of data‐driven cost functions, which, in conjunction with the chosen candidate controller set, appropriately represent control performance and stability goals. Dedicated to Professor Michael Safonov on the occasion of his 70th birthday, this Special Issue aims to report current state‐of‐the‐art in data‐driven robust control and draw further research attention into different aspects of this methodology. We hope that this sample of relevant papers will form a catalyst that will lead to a more unified understanding of data‐driven approach to improving robustness, where models are replaced or supplemented by data.
2018
3665
3666
Jin, Huiyu; Stefanovic, Margareta; Tesi, Pietro
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1139345
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? 1
social impact