A Benchmark for Selecting Real-Time Scheduling Approaches for Fleets of Automated Guided Vehicles in Intralogistics

Abstract In intralogistics, fleets of automated guided vehicles are becoming a key technology offering competitive advantages in performance. Scheduling automated guided vehicle fleets is a complex problem involving the spatial and timely coordination of vehicles and transported cargo. So far, scheduling has been addressed with different traditional approaches as well as machine learning approaches, for example deep (reinforcement) learning models. However, despite the increasing number of approaches, there is no guidance for their selection to optimize fleet efficiency. Hence, decision-makers risk selecting a subpar approach sacrificing efficiency and potentials for competitive advantages. Drawing on the “No Free Lunch Theorem” and using design science, this research contributes with five design requirements, three design principles, and seven design features for a benchmark that ranks approaches for the scheduling of fleets of automated guided vehicles. For the purpose of demonstration and evaluation, the authors instantiated the design knowledge in a benchmark applied in a case study. In terms of implications for practice and theory, the normative knowledge on approach selection can be appropriated to related problems against the background of proliferating AI-based models.