Indirect methods for optimal control of parabolic hybrid PDE-dynamical / switching systems using relaxation
- We propose a novel algorithmic approach to computationally solve optimal control problems governed by linear parabolic partial differential equations (PDEs) including a state-dependent control-regime switching mechanism. We state an equivalent mixed-integer formulation featuring vanishing constraints (VCs) arising from methods of disjunctive programming. We embed the problem into the class of equilibrium constraints (ECs) by introduction of an additional slack variable. Based on theoretical results associated with Sum-Up-Rounding (SUR) strategies, we proceed with the solution of the related relaxed formulation by an indirect approach. In order to obtain a computationally tractable optimality system, we apply a Moreau-Yosida type penalty approach for the VCs. After a theoretical discussion, we introduce and exert the algorithmic framework founded on a semismooth Newton method. Finally, we communicate computational experiments based on the proposed approach.