Problems also arise for SMEs throughout industry, for example when planning production sequences, reducing costs, or organizing supply chains. The “OptimizationChat” project at Bielefeld University, which is supported by the state of North Rhine-Westphalia with around 280,000 euros, aims to develop a chat system that helps companies solve optimization problems in an uncomplicated way.

Practical problem as a mathematical calculation model

Optimization problems can be solved by weighing up numerous possibilities. Mathematics helps to find the best solution. According to the project developers, this requires formal modeling, i.e., the precise translation of a practical problem into a mathematical calculation model. Until now, this task has usually been carried out by appropriately trained experts on the basis of lengthy discussions with the companies. However, this takes time and requires in-depth and expensive expertise, which is often a major obstacle to the use of mathematical optimization, especially for small and medium-sized enterprises.

“The system recognizes the planning task”

OptimizationChat is designed to greatly simplify this process, explains junior professor Dr. Michael Römer from the Faculty of Economics at Bielefeld University: “The system accepts descriptions in normal language and automatically recognizes the underlying planning task. It also asks for clarification if assumptions are unclear and points out missing information. In addition, it supports the automated provision and preparation of operational data for optimization.”

From user dialogue to optimization model

According to Dr. Römer, OptimizationChat translates the results of the user dialogue into an optimization model, a mathematical framework that computer programs, known as solvers, can solve. Solvers are programs that select the best solution from many possible solutions, for example using the integer linear programming (ILP) method, which calculates optimal decisions under fixed conditions. “When companies make better use of materials, energy, and working time, it strengthens competitiveness and makes Europe less dependent on global supply chains. At the same time, the project promotes digital transformation, i.e., the shift to data-based and automated ways of working, thereby supporting innovation in North Rhine-Westphalia,” says Römer.

Explanatory component mediates between modeling and industrial practice

The OptimizationChat project, which contributes to research in the FAITH focus area at Bielefeld University, where the fundamentals and implications of human-AI teamwork are being investigated, aims to take problem knowledge from a database into account for problem solving. In addition, it will have an explanatory component that clearly describes the connection between mathematical modeling and industrial practice. The focus on problems from everyday industrial life and the associated processing of natural language inputs in industry-oriented language is the particular challenge that the project faces.