4D4L - Data- and Target-Driven Sequential Decision Making for Time-Dynamic Logistics Systems


In the operational control of logistics systems, decisions have to be made repetitively over time. Typical framework conditions are a constantly changing data situation and simultaneous uncertainty about future developments. An example of this is the creation of a batch from incoming customer orders.

To ensure effective and efficient operation of a logistics system, it is not sufficient to execute methods from the areas of data analysis and decision-making several times in succession due to the complexity and dynamics of occurring decision-making situations. Instead, an overlapping and mutually complementary coordination and integration of data analysis and optimisation processes into method pipelines is required in order to be able to make the decisions required over time in a target- and data-driven way in the sense of the respective logistics application.

There is a lack of a consistent approach that appropriately takes into account the interactions between the data situation and decision-making in a dynamic context, as is often found in operational logistics. Instead, the respective questions have so far been addressed in different communities: The analysis of (raw) data is carried out in the fields of statistics, machine learning and data engineering; time-dynamic optimisation problems are dealt with in the mathematical disciplines of online optimisation or multi-stage robust optimisation or stochastic programming; logistical issues are mostly investigated with the help of a fixed methodology for decision-making but with insufficient consideration of associated dynamic data processes.

The aim of the research project is


  • the development of a Dynamic Data-Driven Decisions for Logistics (4D4L) metamodel
  • the integrated consideration of the entire chain of data situation, target system, methods of data analysis and decision making
  • the consideration of interactions in the context of time-dynamic logistical decision-making situations.

The metamodel should provide an adaptive and feedback-coupled connection of methods for data analysis and decision-making as method pipelines, which enable a data- and target-oriented control of logistics systems. This results in the possibility of providing a structured method repository for time-dynamic logistics systems and supporting decision-making processes not in isolation, but in a target-oriented manner and in conjunction with suitable methods for data analysis.

To validate the metamodel, the project focuses on the area of warehouse operations, which serves as a meaningful example of time-dynamic logistics problems. In the long term, the project contributes to the realisation of an automated data- and goal-driven composition of method pipelines for decision support.