An active set solver for min-max robust control

An efficient optimization procedure is proposed for computing robust control laws for linear systems with linear state and input constraints and bounded additive disturbances. We describe an active set method for solving the dynamic programming problem associated with the min-max optimization of a predicted cost. The computational complexity per iteration is shown to depend linearly on the length of the prediction horizon. We consider the continuity of solutions, derive bounds on the closed loop disturbance l²-gain and provide numerical comparisons with a disturbance-affine feedback law.