| Research project: | eCargo |  |
| Research area: | Loading optimization, synthetic training data generation, data processing, AI | |
| Funded by: | Federal Ministry for Economics Affairs and Climate Action | |
| Collaboration with: | 3D.aero GmbH, Boeing | |
| Start of the project: | April 2024 | |
| End of the project: | March 2027 | |
| Contact person at the institute: | M.Sc. Felix Geiger |
Description:
In the field of aviation logistics, manual processes are still used to assemble pallets for cargo aircraft. Employees have to find the best possible loading configuration under time pressure and in a changing cargo situation. Parameters to be taken into account here are the densest packing, weight distribution, not causing damage, ensuring that the cargo can be loaded later and having processed all cargo items on time. In particular, time components are often preferred in favor of a higher packing density, which is why optimization potential cannot be fully exploited by the individual. For this reason, the IFPT is researching ways of using the latest software applications for loading optimization in the eCargo project. On the one hand, machine learning is to be used to identify a denser packing structure under given boundary conditions such as weight distribution of the entire load, load capacity of the individual cargo items, consideration of the cargo hold geometry of the aircraft used and the entire cargo to be loaded. On the other hand, the IFPT is addressing a prominent problem in AI research: data availability. For this purpose, training data for state-of-the-art AI applications will be generated using simulative environments, thus creating the possibility of generating data that would otherwise be difficult to create. This will be used to further improve the evaluation using stochastic methods in areas where analytical methods reach their limits in the processing of sensor values. In its sub-project, the IFPT is concerned with the collection of measurement data for individual packages, the post-processing of sensor data, loading optimization, and the generation of synthetic training data. This results in the following work objectives:
1. Development of a simulation environment for loading freight pallets
- Development of a software tool for manipulating freight pallets in a 3D environment
- Investigation of AI applications in loading optimization
- Development of a loading optimization strategy
2. Generation of synthetic training data
- Construction of a software component to generate synthetic image material
- Identification of necessary parameters for annotations
- Development of an automated pipeline for training data generation
3. Improvement of sensor resolution
- Removal of sources of interference in the measuring field
- Determination of packaging material in the image
- Distinguishing between compressible and rigid areas of a load
- Determination of the external geometry of a packaged pallet without protective film or transport nets.
Contact person at the institute: M.Sc. Felix Geiger