Optimal control of the multidimensional diffusion equation (Q1176844)

The authors consider the problem of finding the optimal boundary control for the \(n\)-dimensional diffusion equation. An existence result under suitable hypotheses is derived for this problem. In order to numerically determine the solution to this problem, the problem is first converted to one of finding a minimum of an objective over a convex and compact set of positive Radon measures which gives rise to an infinite dimensional linear programming problem. This is then approximated by a finite dimensional linear programming problem where now the set of measures consists only of atomic measures. The optimal solution to each finite dimension linear programming problem is used to construct a piecewise constant control function for the original problem. A sequence of the linear programming problems then gives rise to a sequence of piecewise constant control functions which, in the limit, satisfy the final state constraint and where the objective values converge to the optimal value. Several examples of this procedure are included.