A constrained optimal control problem with application to production planning

In this paper, an aggregate production planning problem is formulated as a stochastic sequential optimization model with constraints on the main decision variables. The difficulty to find a true optimal policy leads to investigate the use of near-optimal techniques. Due to the easier computational implementation, the Naive Feedback Controller (NFC) approach is used. It provides a revised sub-optimal production policy to the stochastic problem. In order to improve the NFC solution, a equivalent mean value problem that preserves some statistics features of the original problem is proposed. An example of two products, multi-period, stochastic production planning problem is used to illustrate the applicability of this approach. Copyright (C) 1998 IFAC.