Direct data-driven design of LPV controllers with soft performance specifications. Issue 2 (January 2022)

• Record Type: Journal Article
• Authors: van Meer, Max; Breschi, Valentina; Oomen, Tom; Formentin, Simone
• Is Part Of: Journal of the Franklin Institute. Volume 359:Issue 2(2022)
• Page Start: 816
• Page End: 836
• Publication Date: 2022-01
• Subjects: Science -- Periodicals; Technology -- Periodicals; Patents -- United States -- Periodicals; 505
• DOI: 10.1016/j.jfranklin.2021.04.052
• Languages: English
• ISSNs: 0016-0032

Abstract

Abstract: If only experimental measurements are available, direct data-driven control design becomes an appealing approach, as control performance is directly optimized based on the collected samples. The direct synthesis of a feedback controller from input-output data typically requires the blind choice of a reference model, that dictates the desired closed-loop behavior. In this paper, we propose a data-driven design scheme for linear parameter-varying (LPV) systems to account for soft performance specifications. Within this framework, the reference model is treated as an additional hyper-parameter to be learned from data, while the user is asked to provide only indicative performance constraints. The effectiveness of the proposed approach is demonstrated on a benchmark simulation case study, showing the improvement achieved by allowing for a flexible reference model.