A time-domain decomposition iterative method for the solution of distributed linear quadratic optimal control problems (Q704186)

The subject of this paper is the numerical solution of large-scale linear quadratic optimal control problems governed by parabolic partial differential equations. A time-domain decomposition is used and the problem is reformulated as a discrete-time optimal control (DTOC) one using a multiple shooting approach which is matrix free. It is based on the observation that the optimality conditions for the DTOC problem lead to a block tridiagonal linear system. Moreover the diagonal blocks are invertible and are related to the original linear quadratic optimal control problem restricted to smaller time-subintervals. A block Gauss-Seidel method is therefore used which is preconditioned by a Krylov-subspace method. Moreover some instantaneous control techniques can be interpreted as the application of one step of the preconditioned Gauss-Seidel method. Numerical experiments are presented such as Neumann control for the 1D heat equation and Dirichlet control for the 2D heat equation.