A Toolkit for Efficient Computation of Sensitivities in Approximate Robust Optimal Control Problems

Efficient solution approaches for optimal control problems where the dynamics are described by uncertain differential equations are discussed in the present paper. Problems with uncertainties can be addressed by the robust worst-case formulation. In order to numerically solve the robust counterpart for the optimal control problem several approximation techniques can be employed. In this paper we use an approach based on linearization and solution of Lyapunov differential equations. We exploit the structure of the Lyapunov equation in the optimal control context providing an efficient numerical implementation. The capabilities and computational times of the new approach are demonstrated on two (bio)chemical examples.