Outperformed by a Computer? - Comparing Human Decisions to Reinforcement Learning Agents, Assigning Lot Sizes in a Learning Factory: 11th Conference on Learning Factories, CLF2021

Research output: Working paperWorking papers

Authors

Given the fact that some products in the production system require larger cycle times and a setup process during production, the application of different lot sizes might be beneficial regarding economic and logistic key performance indicators. Maximizing the profit considering the according work in process, inventory, set up and penalty costs is a complex challenge. In this contribution, we present the complete workflow from collecting data, deriving a simulation model, as well as training and deploying a machine learning model for assigning lot sizing in a dynamic production system. Furthermore, we compare the results of human participants and the reinforcement learning agent in a real-world learning factory workshop. The comparison of the results shows, that the reinforcement learning-agent is able to achieve the same level of results as the human experts.
Original languageEnglish
Place of PublicationRochester
PublisherElsevier Inc.
Number of pages6
DOIs
Publication statusPublished - 06.2021

Bibliographical note

11th Conference on Learning Factories; July 1st - July 2nd, 2021. Online Event. Posted: 4 Jun 2021

    Research areas

  • Engineering - reinforcement learning, Artifical intelligence, Lot sizing, Learning Factory

Recently viewed

Publications

  1. Closed-form Solution for the Direct Kinematics Problem of the Planar 3-RPR Parallel Mechanism
  2. Towards productive functions?
  3. Factored MDPs for detecting topics of user sessions
  4. Investigating the Effect of Noise Elimination on LSTM Models for Financial Markets Prediction Using Kalman Filter and Wavelet Transform
  5. Who can receive the pass? A computational model for quantifying availability in soccer
  6. An Outcome-Oriented, Social-Ecological Framework for Assessing Protected Area Effectiveness
  7. Combining Evaluative and Generative Diagnosis in ActiveMath
  8. “Ideation is Fine, but Execution is Key”
  9. Towards a spatial understanding of identity play
  10. Introduction
  11. Foreign bias in institutional portfolio allocation
  12. Active and semi-supervised data domain description
  13. Double-fading support - A training approach to complex software systems
  14. Trajectory tracking using MPC and a velocity observer for flat actuator systems in automotive applications
  15. ℓp-norm multiple kernel learning
  16. Individual Scans Fusion in Virtual Knowledge Base for Navigation of Mobile Robotic Group with 3D TVS
  17. Functional Richness and Relative Resilience of Bird Communities in Regions with Different Land Use Intensities
  18. Pressure fault recognition and compensation with an adaptive feedforward regulator in a controlled hybrid actuator within engine applications
  19. Efficacy of an internet and app-based gratitude intervention in reducing repetitive negative thinking and mechanisms of change in the intervention's effect on anxiety and depression