On 8th of November, 2021, Martijn Beeks received the Dow Inc. Best OML Master Thesis Award 2021. This year, the award was handed over by Nancy Gutierrez (Regional Customer Service Director), representative of Dow Inc. at the Diploma Ceremony 2021.
The jury consisted of Aram de Ruiter, Bao Lin, and Christian Hubbs from Dow Inc., and Laura Genga and Tugce Martagan from Eindhoven University of Technology. Marco Slikker served as non-voting jury chair. The jury judged the theses on their academic contribution and industrial impact. The students that graduated in the past academic year and received at least a grade of 9 for their theses have been nominated for the award. 14 students satisfied this criterium. The jury concluded that all these 14 theses were of outstanding quality. They were happy to see that the high quality of Master Theses was maintained throughout the pandemic conditions, as well as the responsiveness of the university to include topics like supply chain disruption (for example because of a pandemic) and sustainability.
In the thesis, titled “Deep reinforcement learning for solving a multi-objective online order batching problem”, Martijn adopts a deep reinforcement learning and Bayesian optimization approach to study a complex order batching and sequencing problem at Vanderlande. The model and the analysis are thorough and rigorous, and the work provides new insights for multi-objective online order batching problems. The quality of the experimental setup and the report are impressive.
Two main differentiators are the use of deep reinforcement learning (DRL) and the accommodation for an objective function that is not the same all the time. To start with that last concept, the thesis evaluates service, being on time, and cost in one objective function and recognizes that this is not a stable aim over time. This matches with the reality in industry and in for example Dow’s Supply Chain at multiple levels. It also takes on a timely approach as it illustrates that Deep reinforcement learning (DRL) can rightfully be acknowledged by companies like Dow Inc. as a potential solution to large-scale decision problems with significant uncertainty. Further research and efforts in this space are encouraged.