Data-driven mixed-sensitivity structured control of SISO multi-model systems with application to a reconfigurable industrial oven

This paper addresses data-driven robust control design for Single-Input Single-Output (SISO) multi-model systems with mixed uncertainties. The proposed scheme tackles uncertainties arising from three sources: model estimation variance, measurements noise, and plant behavior variations across operating points. These uncertainties are combined into a single global output multiplicative uncertainty, encompassing the variability of all local models. The uncertainty quantification utilizes: (i) kernel methods for identifying the model of the plant, and (ii) a randomized approach for estimating the uncertainty region, guaranteeing a high probability of capturing the true plant behavior within the estimated uncertainty region. The designed robust controller for multi-model systems is evaluated in simulation on a model of a reconfigurable industrial oven employed in the packaging industry.