Reinforcement Learning for Decision Support
Research output: Thesis › Master's Thesis
Standard
2022.
Research output: Thesis › Master's Thesis
Harvard
APA
Vancouver
Author
Bibtex - Download
}
RIS (suitable for import to EndNote) - Download
TY - THES
T1 - Reinforcement Learning for Decision Support
AU - Roth, Martin
N1 - no embargo
PY - 2022
Y1 - 2022
N2 - Personnel costs are an important performance indicator in warehouse operations as they directly influence the cost and profit. Therefore, to maximize the profit, good personnel planning is essential. In this Master's Thesis, we investigated the use of reinforcement learning using neural networks to support decision making for assigning work operators. At first, an introduction to reinforcement learning and neural networks is given. Then, based on a system analysis, a Markov Decision Process (MDP) is designed. Based on the MDP a simulation is proposed to provide the data necessary for reinforcement learning. Finally, the reinforcement learning approach to predict the expected profit for different decisions is described in the final step. The evaluation on different validation scenarios shows, that the proposed reinforcement learning approach achieves higher profits than a naive algorithm and should therefore be considered as valuable support in future warehouse operations.
AB - Personnel costs are an important performance indicator in warehouse operations as they directly influence the cost and profit. Therefore, to maximize the profit, good personnel planning is essential. In this Master's Thesis, we investigated the use of reinforcement learning using neural networks to support decision making for assigning work operators. At first, an introduction to reinforcement learning and neural networks is given. Then, based on a system analysis, a Markov Decision Process (MDP) is designed. Based on the MDP a simulation is proposed to provide the data necessary for reinforcement learning. Finally, the reinforcement learning approach to predict the expected profit for different decisions is described in the final step. The evaluation on different validation scenarios shows, that the proposed reinforcement learning approach achieves higher profits than a naive algorithm and should therefore be considered as valuable support in future warehouse operations.
KW - Reinforcement Learning
KW - Deep Learning
KW - Maschinelles Lernen
KW - Mitarbeiterzuweisung
KW - Reinforcement Learning
KW - Machine Learning
KW - Operator Assignment
KW - Deep Learning
M3 - Master's Thesis
ER -
