Introduction to Reinforcement Learning

This project was part of the "Introduction to Reinforcement Learning" course at the Catholic University of Eichstätt-Ingolstadt.

Project description

Inventory management deals with the questions on when to order, in which quantity, potentially even from which source (supplier) to balance availability and costs. The main challenge in inventory management consists in dealing with changing demands, varying lead times, and / or disruptions. There are many different specific problem settings in inventory management (e.g., single vs. multiple supplier, fixed reorder vs. variable reorder quantity, etc.). Therefore, the objective was to work on an implementation of the Q learning reinforcement learning algorithm in an inventory management scenario, which is a typical application area in supply chain management. The file RL_inventory_management_groupwork.pdf contains the report, which describes the problem setting, the data generation process, an outlook on extending the model to DQN, as well as the results and a discussion. The implementaion is visible in the inventory_beer_game_implementation.ipynb file.

Team Members

• Dennis Götz
• Vincent Bläske
• Paul Posselt
• Serkan Akdemir