Because of the existence of mixed boundary conditions, the solution of nonlinear dynamic optimal control problems via the maximum principle often requires an algorithm which updates a trial solution from iteration to iteration. We show that two-dimensional system theory, expressed in the form of an unit memory repetitive process, can be employed to analyse the convergence and stability properties of these algorithms. Particular emphasis is given to novel algorithms for the solution of continuous and discrete optimal control problems subject to model reality differences. Unit memory linear repetitive process theory can be employed to analyse the local stability of the algorithms and to show that the technique converges to the correct optimal control solution in spite of model-reality differences. (5 pages)