Dynamic control of a class of discrete event systems using a state reconstruction algorithm

The problem of dynamic control of Discrete Event Dynamic Systems (DEDS) is addressed in this paper as a dynamic optimization problem : some resources must be allocated to the system in order to optimize a performance function which is assumed time-varying. The control scheme exploits a state reconstruction algorithm to compute an estimate of the performance for perturbed sample paths. The algorithm is based on the use of data extracted from the observation of the system and allows to accurately reconstruct its state behavior, for resource allocations different from the nominal one. The proposed control scheme is then used for dynamic allocation of buffer capacities in mamlfaeturing systems, such as Kanban systems. A parallel implementation of the whole algorithm is also mentioned.