CSI UniCA : Projet COEPLEX

Crédits Scientifiques Incitatifs d’Université Côte d’Azur, volet recherche
Incentive Scientific Credits of Université Côte d’Azur, research component

COEPLEX : Contrôle Optimal d’Ensembles pour la PLanification d’EXpériences
Ensemble optimal control for experimental design

Project leader. Ludovic Sacchelli

Partners. Alessandor Scagliotti (TUM), Radomyra Shevchenko (LJAD, UniCA)

Summary. Experimental design involves the systematic planning and execution of experiments to maximize information gain while minimizing resource use. It focuses on selecting inputs and conditions to optimize data collection, which is crucial for accurate parameter estimation and model validation. Traditionally, this is achieved by considering the Fisher Information Matrix and optimizing its content to enhance the efficiency and robustness of estimation algorithms. When dealing with parameter estimation for dynamic input-output systems, this objective can be reframed as an optimal control problem. This approach allows the use of control theory to guide input selection, providing a structured method for exploring and optimizing experimental conditions. However, closed-loop strategies for experimental design often rely on initial guesses of the parameter values, which can introduce uncertainties. Ensemble control, a method designed to address parameter uncertainty, offers a novel solution to these challenges. By considering a range of possible parameter values simultaneously, ensemble control can enhance the robustness of experimental designs, ensuring reliable outcomes even in the presence of initial estimation errors.