Unfalsified adaptive control with MPC candidates

In this paper, we consider unfalsified adaptive control with candidate controllers that are in the structure of model predictive control (MPC) and are associated with parametrized models. Thus, each candidate controller corresponds to a parameter of the models and a switching corresponds to an update of the parameter selection. We introduce a new switching algorithm in order to pursue a performance criterion imposed on an output signal of an underlying system as well as identification of the system. This switching algorithm achieves an asymptotic boundedness of the performance criterion while the parameter update may be performed endlessly. At the beginning of an experiment, the switchings can be set to be performed mainly focused on identification of the system. Thus, we implement an input design scheme in the MPC structure for a better transient behavior of the closed-loop system via fast system identification.