Discrete-Time Finite-Horizon Optimization of Singularly Perturbed Nonlinear Control Systems With State-Action Constraints

An algorithm is introduced for the computation of an approximate optimal control policy for discrete-time finite-horizon nonlinear singularly perturbed systems. This is achieved through timescale separation and by utilizing ideas from parametric optimization and dynamic programming. We demonstrate that our proposed method produces a control policy that is both theoretically robust and nearly optimal.