Adaptive control via finite modelling and robust control

This paper examines an adaptive control scheme which involves the selection of a fixed controller out of a finite collection, where each controller is robustly designed to account for the true plant being in a not necessarily small set of uncertainty. There are no adaptive 'parameters' in the conventional sense where adjustments are made to either model or controller parameters. The parameters here involve pseudo-probabilities or weights assigned to each of the finite controllers. These quantities are computed recursively from the measured data.