The project developed an end‑to‑end digital dispatch system for freight operators. The workflow starts with the import of customer orders and forecast results. In the first step, orders are loaded from CSV files; missing coordinates are derived from postal codes and invalid records are shown in a separate panel. The second step imports forecast results, allowing the dispatcher to enable or disable individual orders manually. After the data are assembled, a solver generates an optimal tour plan for the fleet. The result is displayed in a tree view and on a map, and key metrics such as total kilometres, load kilometres and the percentage of empty runs are shown. Orders that cannot be scheduled within the planning horizon are listed separately. The dispatcher then reviews the suggested changes. If a proposed change cannot be implemented, the system triggers a re‑optimization that respects the already accepted changes. This loop continues until all changes are confirmed or the user stops the process. The final plan is exported to a file.

Technical performance was evaluated during a pilot with two industry partners. The optimisation reduced the empty‑kilometre ratio by up to 31 % in a scenario with high load surplus, and by 21 % when there was no surplus but a large volume. In a third scenario with small surplus, the reduction was 27 %. The system also achieved a 1.87 % improvement in the empty‑kilometre ratio when the fleet was heavily loaded. These figures demonstrate that the solver can significantly improve utilisation and lower emissions. An Excel tool was built to calculate CO₂ emissions from the generated routes, allowing the partners to quantify the environmental benefit of the optimisation.

A key scientific contribution is a heuristic that identifies potentially acquirable orders. For each transport relation the heuristic evaluates whether calling a customer would improve the empty‑kilometre ratio, using historical data and domain knowledge. The heuristic classifies relations as “call‑worthy” or “black‑listed”. The partners also built a MySQL database that stores external time‑series data such as economic indicators, holidays, industry indices and weather. The database can be updated automatically and extended with new sources, providing a robust foundation for forecasting.

Strategic network optimisation was addressed with a Python‑Dash visualisation tool. The map colours counties according to net freight flow; deep red indicates a net inflow, deep blue a net outflow. Clicking a county reveals the transport relations and the number of shipments. Dashed lines show potential orders that could balance the network. The tool also displays historical time series and machine‑learning forecasts for each relation, enabling planners to spot bottlenecks and opportunities for back‑hauling.

The project was carried out in close collaboration between the logistics companies Schmahl & Stöpel and BLG, the Fraunhofer Institute for Intelligent Systems and Image Analysis, and a software partner that provided the optimisation engine. The partners jointly defined the requirements, validated the prototype, and conducted the pilot. The work was funded by German research agencies and supported by the participating companies. The iterative development cycle, involving user feedback and technical refinement, ensured that the final demonstrator met the operational needs of the partners and delivered measurable improvements in route efficiency and environmental performance.