Generating the ideal pallet structure: Task-driven solution

A complex issue in the fresh produce industry is finding the optimal structure of mixed pallets, taking into account the properties of the goods that are stacked on the pallet. Reliance is placed on the knowledge of experienced employees, but this does make it difficult to transfer knowledge to othwer empoyees. To further shape this issue, a project group from TU Delft started research at the beginning of this year. Paul Tervoort, developer at jem-id and one of the students from the project group, explains what users can expect in this update.

Paul begins: “The assignment immediately appealed to us, because there was no fixed solution. We were free to choose which technology would best suit the issue.” Every year, jem-id invites a project group from TU Delft to conduct research into a technical issue. This time with a current topic for many fresh produce companies; creating a self-learning algorithm for building mixed pallets.

“We initially decided to build an algorithm using Artificial Intelligence (AI) or Machine learning techniques. These techniques make it possible to train a computer by feeding it with data. It becomes smarter from mistakes and eventually learns what the most favorable outcome is. Although the issue fits perfectly with an AI solution, it would take us too much time to create a workable situation. After all, we only had 10 weeks. That is why we opted for a different, algorithmic solution.” explains Paul.

This solution was powered by data made available by two GreenCommerce users. Paul continues: “We quickly encountered the obstacle of missing information, especially in the area of ​​pallet and product dimensions. We then suggested dimensions for the project ourselves so that the system could still generate a pallet structure.”

As a follow-up to his assignment, Paul started working on a new part in GreenCommerce whereby dimensions of products and pallets can also be recorded in the master data. GreenCommerce takes into account products that have the same hierarchy or dimensions. Paul: “Our solution worked well based on the data we had entered ourselves, but was therefore not a 100% match for the company or the industry. I am currently working on a new part in GreenCommerce where this information can be entered by the customer, so that the system works with the correct information. This will in the future make it possible to generate the most efficient pallet structure for the products.”

Because the TU students' algorithm is written in a different programming language, Paul is currently working on transferring the data to C#. This is the programming language with which GreenCommerce is largely built. “I like that jem-id does something with our assignment. You often see that internship assignments end up in the trash at companies. Because I can continue working on the issue, the project gains more depth and also becomes part of GreenCommerce.” concludes Paul.

Test sessions with customers will follow soon to validate the chosen direction. In the same period, two other colleagues are working on another part of the Task-driven solution. This part focuses on managing employees in the warehouse more specifically. New functionalities will be added to the WMS App. More information about this will follow later.

Do you have questions or would you like to receive more information?
Please contact us via:
info@greencommerce.nl
+31 174 - 642 622