This research work presents a hierarchical nonlinear optimization-based framework for planning and scheduling of supply networks modeled as multi-class stochastic queueing networks. More precisely, the framework has two decision layers. At the top level a tactical processing plan is designed. This is accomplished by means of a nonlinear optimal control formulation, along with suitable solution algorithms, to compute on a rolling-horizon basis a tactical processing plan which yields the lowest cost expected-value inventory trajectory. In doing so, raw materials, inventory and demand mismatching costs are considered. As the main outcome, a target inventory trajectory is obtained for each inventory buffer in the network at all times within a rolling planning time window. The bottom layer in turn deals with a distributed scheduling framework to best track the inventory targets generated by the tactical processing plan. In this regard, a processing schedule is generated for each server in the system so that sequence-dependent and inventory holding costs are minimized for each server within the current planning time window. Uncertainty in the inventory accumulation and depletion processes as well as network interrelations is accounted for by means of a robust optimization formulation.

Several numerical examples are presented in order to illustrate the framework mechanics as well as the algorithmic issues inherent to the different optimization steps performed on a hierarchical basis. The planning framework can also be adapted to standard Enterprise Resource Systems (ERPs) to best support the planning and scheduling functions on a regular basis.