Genetic learning through simulation: An investigation in shop floor scheduling

This paper considers the automated learning of strategies for real-time scheduling in dynamic factory floor environments. A simulation model of the shop floor provides continuous inputs to a genetic algorithm based learning system. Learning is used to update the knowledge bases of "intelligent" dispatchers in the floor shop setup. The performance of the learning system is compared with that of commonly used dispatching rules, and experimental results are presented for a two-stage flowline and for a more general jobshop environment.