a) A significant progress has been made on the automatic conversion of given first-principles models of engineering systems (which are often highly complex and nonlinear) to low complexity LPV models that can be readily used for analysis of the system with computationally cheap convex methods and also for controller design, which significantly lowers the required expertise form the side of control engineers to apply the LPV framework in industrial applications. This has been achieved both in terms of direct model conversion tools using various techniques (loss-less multipath-feedback linearization, function factorization, etc.) and model reduction methods (moment matching, PCA and ANN based methods, etc.) The resulting methods automatically minimise complexity and conservativeness of the resulting LPV models. Significant progress on data-driven modeling (plug & paly modelling) in terms of system identification has been made by working out a unified framework for subspace and max. likelihood identification, connecting nonlinear identification with LPV modelling and also developing machine learning-based methods (ANN, SVM, etc.) that can directly learn low complexity LPV models with their scheduling maps from data. Also methodologies for preservation of important structural properties in the resulting LPV models (like stability and controllability has been established to ensure feasibility of the follow up LPV analysis or controller design. In overall, a modelling toolchain for system oriented design together with software implementation and many case studies have been developed.
b) We have established a novel frequency domain understanding of LPV systems, paving the way for using industrial experience in controller shaping for linear time invariant systems to be used through the LPV framework for nonlinear systems, ensuring wider application of current industrial methods in controller design. These results were further exploited for frequency-domain tuning of LPV controllers, characterization of performance shaping for controller synthesis and incorporation of shaping objectives into modelling.
c) As a major accomplishment of the research, a novel way has been found in terms of incremental and shifted dissipativity concepts based analysis/synthesis results established in this project to give solid guarantees of stability and performance of synthesized LPV controllers for nonlinear plants and understand the boundaries of their application. As a consequence, it became possible to use "linear" methods and the connected vast industrial experience to design directly controllers for nonlinear plants with clearly understood stability and performance guarantees for the first time. This also allows to have a unified framework of performance shaping for highly complex mechatronic/aerospace/etc. systems with non-linear behavior and also backtracking the possible performance loss of the controllers in terms of the applied design choices. A significant number of case studies with experimental results have been established with many publications and software tools.
d) Development of breakthrough results in terms of connecting the established control design toolchain with LPV data-driven modelling has been also accomplished, resulting in a direct system oriented design approach. By using the LPV behavioural framework, extension of the Fundamental Lemma for the LPV case was developed together with a complete data-driven stability and performance analysis framework and a direct controlled system synthesis approach for NL systems based on time-domain data. In parallel, direct control design tools for frequency domain data has been established with the same generality of the results, but giving a frequency-domain understanding of the design which is highly important for the mechatronic industry.
e) An open-source toolbox (www.LPVcore.net) has been developed to give off-the-shelf tooling for the results of the research. Resulting methods in the WPs have been tested on benchmark examples and in experimental studies on laboratory setups. Successful testing of the developed methods on industrial use cases has been accomplished in co-op with the European Space Agency, ASM Pacific Technologies, and also Mathworks, developer of Matlab has established a co-op to push the main results into their software products.
f) Wide-range scientific dissemination of the results have been established in terms of many conference presentations, 70+ scientific publications, research co-ops, 2PhD theses (+2 in progress), but also industrial dissemination in terms of spin-off development projects, co-ops and knowledge transfer projects.
