Mixed pallet loading optimisation with stability constraints

Boxes of mixed sizes can be loaded on a pallet automatically by robotized palletizing systems. Determining the pattern in which the boxes should be placed is a complicated optimisation problem, which can be solved reasonably well by heuristic algorithms. However, known algorithms have only very limited consideration of pallet stability, i.e. how well the pallet holds together when moved without external support. The goal of our project was to develop a practical pallet loading optimisation algorithm that finds pallet loading patterns with both good stability and fill rate.

We developed novel computational formulations of pallet stability and the level of binding between successive layers, and incorporated them into state-of-the-art pallet loading algorithms from research literature. We combined the earlier algorithms into an optimisation method that splits the available boxes between pallets so as to minimise the total number of pallets needed, and constructs individual pallet loading patterns maximising fill rate under stability constraints. In particular, we developed novel push operations that rearrange boxes on a pallet under complex stability constraints, while maintaining a loading order that is implementable by a robot. A separate physics-based simulation tool was connected to the optimisation software to further validate the results: if a pallet would fall over when moved in a detailed physics simulation, the optimisation would have to replace the pallet.
The commissioned research & development project was carried out in close collaboration with the customer Orfer Oy, an engineering company that provides turnkey robot automation solutions for material handling systems. The project involved requirements specification, mathematical modelling, algorithm development and testing. The customer provided visualization and physics-based simulation tools to aid in the development of the optimisation software, and performed extensive testing in both computational simulations and in practical experiments with a palletizing robot.