Logistics

Tetris-playing AI: Optimizing the use of space and materials in logistics

The perfect package
Large pile of parcels in a warehouse

The "Perfect Package" research project focused on packing goods into large containers as space-efficiently as possible. The goal was to use AI to make the packaging process more efficient and optimize shipping, making both more economical and environmentally friendly without compromising delivery times.

Together with our project partners from the Institute for Materials Handling and Logistics at the Karlsruhe Institute of Technology (KIT) as well as the software company specializing in intralogistics TUP our AI experts at AMAI have developed heuristic and reinforcement learning (RL) approaches that master this problem.

Challenge

Packaging processes in the logistics industry have a problem: they are inefficient in terms of space utilization. This has negative economic and ecological consequences. Because air is being transported unnecessarily, more filler and packaging material as well as a higher number of packages are required. This, in turn, leads to higher costs, greenhouse gas emissions, and a larger ecological footprint.

This problem is known in scientific literature as the3D bin packing problem. The 3D-BPP is one of the most prominent optimization problems in logistics. Containers such as packages or pallets must be filled with goods packed as densely as possible to minimize the number of containers required.

The specific use case in this project required additional, project-specific constraints. For example, a split-order ban applied to individual orders, meaning goods could not be distributed across multiple containers. An overhang ban stipulated that items could not protrude beyond other items underneath them.

In this research project, funded by the state of Baden-Württemberg, we sought innovative approaches to reduce unused space and the number of containers in these packaging processes.

Approach

To solve the task, we first built asimulation environment. This virtual representation allowed us to run through many different packing scenarios with various container and item sizes.

Grafische Darstellung der Simulationsumgebung

Following this,two AI solution approacheswere developed for the bin packing problem: a heuristic approach and an approach using reinforcement learning (RL). Various model variants were designed for both options.

For comparison with other approaches from literature and industry, we used atarget metricthat takes into account both the number of containers used and the unused space within the containers (in liters).

To enable faster model training and thus shorter development cycles, we implemented aSimplificationwas implemented. The action space was reduced by discretizing the packing area into 3D grids with varying resolutions (from 0.1 cm to 16.0 cm).

Following a testing phase in the digital twin, the best models were selected. To validate the practical viability of the solutions, afield testwas conducted in a research partner's warehouse. A large order was packed according to computer-generated instructions.

Result

The project provided valuable insights into how AI can be used to solve packing problems in the logistics industry. Both the heuristic and the reinforcement learning approaches were able to place goods into containers fully automatically.

The various approaches were compared against an existing baseline using orders randomly selected from the test data.The heuristic consistently achieved the best target metric.The newly developed models were able to outperform the status quo, particularly with large orders.

Thefield test conducted in a warehouse was successful. A large order was successfully packed with the help of a model suggestion, demonstrating that the virtual solutions are implementable in practice.

Ausschnitt eines im Feldtest gepackten Modellvorschlags

Another outcome of this project was a simulation environment that can be used to train reinforcement learning agents for packing problems.

‍

Table of contents
Branche
Logistics
Thema
No items found.
Share now
Link kopiert!
Dr. Jürgen Stumpp

Ready for the next step?

Let's find out together how we can support your business with customized AI.

Dr. Jürgen Stumpp
Managing Partner | AI Strategy Consultant
+49 170 9374842
Contact us