Convergence of action dependent dual heuristic dynamic programming algorithms in LQ control tasks

The purpose of the paper is to present an algorithm to solve the optimization problems concerning with robust LQ control, as well as a method how the exposed problems can be reduced to a standard formulation using steepest-descent gradient function principles, where dual heuristic dynamic programming is used. Robust LQ control design is specific by more general one-step cost function, than ones usually used for critic-based training, and with greedy minimization for updating not only the action but also the critic.