Chancen und Herausforderungen bei der Anwendung von KI-Methoden für die Entscheidungsfindung in der operativen Produktionsplanung

Abstract

This bachelor thesis analyses the potentials and challenges of using artificial intelligence (AI) methods for decision support in operational production planning. In view of growing demands for efficiency and flexibility in production planning, the extent to which AI methods such as reinforcement learning (RL), deep reinforcement learning (DRL) and multi-agent systems (MAS) can contribute to the optimisation of operational decision-making processes is analysed. The focus is on specific application areas such as dynamic scheduling, machine utilisation and real-time production planning. The results show that multi-agent deep reinforcement learning (MADRL) models in particular have great potential to make operational decision-making more efficient, robust and adaptable. By setting up decentralised decision-making processes, continuous learning in simulated environments and coordinated agent interaction, planning efficiency, adherence to schedules and responsiveness can be significantly increased. At the same time, disruptions such as machine breakdowns or fluctuations in demand can be flexibly compensated for. However, these results are based on theoretical analyses and simulation models, not on practical applications. The main challenges lie in the high data requirements, the lack of transparency of AI decisions due to the ‘black box’, integration into existing information technology (IT) infrastructures and acceptance by employees. Ethical and legal issues, such as the responsibility of automated decisions, must also be critically scrutinised. Based on these findings, the Bachelor's thesis concludes that AIbased planning methods can make a significant contribution to increasing efficiency and flexibility in production. However, the introduction of such systems requires well thought-out integration into existing IT infrastructures, a suitable data strategy and explainable AI models in order to ensure long-term trust and benefits.

Date of Award	2025
Original language	German (Austria)
Supervisor	Roland Braune (Supervisor)