Open Access

Timo Schroth, Jan SchuhmacherORCiD, Vera HummelORCiD, Konrad von LeipzigORCiD

• Small and medium-sized enterprises encounter challenges in improving and integrating digital solutions in intralogistics’ pick and stow operations due to their reliance on manual processes. Traditional methods based on historical data are becoming obsolete since they are not capable of coping with shortened product life cycles and decreased order quantities. Instead, AI reveals potential for data-driven analysis and enhancement of these operations. Currently, applied AI algorithms tend to optimize either pick or stow operations individually, missing opportunities for holistic improvement by reducing travel distances and times. This paper bridges these gaps by developing an unsupervised machine learning model to conduct AI-driven cluster analysis, bundling the operations conjointly into travel distance-optimized clusters while simultaneously targeting time efficiency. The approach is verified using synthetic data generated by a simulation model based on Werk150, the learning factory of Reutlingen University. The paper’s research addresses the validation as proof of concept within Werk150, focusing on incorporating real-world operations data. A smart glove driven by AI algorithms captures near-real-time process data on the logistician’s position and the components being handled, enriching the model with this information. To generate a database for the scenario-based validation, a learning module is developed at graduate level. After a theoretical introduction to provide an understanding of the AI clustering model for pick and stow operations, the module includes practical exercises with and without the use of the developed approach. This enables students to experience AI’s potential to reduce travel distances and times practically while generating the required database for validation.