Inhaltszusammenfassung

In an industry project, the authors had successfully modelled and simulated the inbound and outbound traffic of a warehouse with the aid of high-level Petri nets. But instead of taking this simulation tool to improve the future planning, the practitioners confronted the authors with another problem: the reasons for the incorrect planning is less the planning process but the inability to foresee which of the scheduled trucks will be late and sabotage the time plan. As a solution to this proble...In an industry project, the authors had successfully modelled and simulated the inbound and outbound traffic of a warehouse with the aid of high-level Petri nets. But instead of taking this simulation tool to improve the future planning, the practitioners confronted the authors with another problem: the reasons for the incorrect planning is less the planning process but the inability to foresee which of the scheduled trucks will be late and sabotage the time plan. As a solution to this problem, the authors considered methods of predictive artificial intelligence and especially machine learning. The idea is to take past schedules to train a neural network in order to forecast deviations of upcoming schedules. The paper answers the question on how to extend the previous simulation model by a suitable forecast component which is now ready to be tested with real-world data. The paper explains the scenario of the real-world warehouse, the simulation of its traffic and the information needed for this. Afterwards, the necessary extensions of the data set are explained and how to set up a machine learning component to predict future deviations of schedules. The adapted schedules can then be simulated to create alternative schedules. This is a next step of the authors' research to conduct their simulation on a set of future scenarios in order to chose the schedule that performs not worse than the initial schedule on the original data, but performs better under the alternative scenarios.» weiterlesen» einklappen

Autoren

Klassifikation

DFG Fachgebiet:
Informatik

DDC Sachgruppe:
Informatik