Minimax Control and Dynamic Games
This project is aimed at developing a time-domain based theory for derivation
of minimax (worst-case) identifiers and controllers for nonlinear systems
with norm-bounded and partially stochastic uncertainties, and analyzing
their robustness with respect to modeling inaccuracies and model reduction.
Robustness is imposed here on the top of minimaxity, and introduces a further
classification of minimax estimators and controllers according to their
admissibility, or sensitivity to inaccuracies in plant modeling. Of particular
emphasis here is the inaccuracy that results from model simplifications
due to time-scale separation (such as presence of unmodeled fast dynamics)
or weak coupling of several subsystems. A further topic of study is the
characterization of robust controllers for parametrized families of plants
and under multiple criteria. The general approach adopted is that of dynamic
or differential game theory.