INEXACT RESTORATION AND ADAPTIVE MESH REFINEMENT FOR OPTIMAL CONTROL

A new adaptive mesh refinement algorithm is proposed for solving Euler discretization of state- and control-constrained optimal control problems. Our approach is designed to reduce the computational effort by applying the inexact restoration (IR) method, a numerical method for nonlinear programming problems, in an innovative way. The initial iterations of our algorithm start with a coarse mesh, which typically involves far fewer discretization points than the fine mesh over which we aim to obtain a solution. The coarse mesh is then refined adaptively, by using the sufficient conditions of convergence of the IR method. The resulting adaptive mesh refinement algorithm is convergent to a fine mesh solution, by virtue of convergence of the IR method. We illustrate the algorithm on a computationally challenging constrained optimal control problem involving a container crane. Numerical experiments demonstrate that significant computational savings can be achieved by the new adaptive mesh refinement algorithm over the fixed-mesh algorithm. Conceivably owing to the small number of variables at start, the adaptive mesh refinement algorithm appears to be more robust as well, i.e., it can find solutions with a much wider range of initial guesses, compared to the fixed-mesh algorithm.