Stutensee, Germany, 27.01.2021. DR. THOMAS + PARTNER (TUP), together with the Institute for Materials Handling Technology and Logistics Systems of the Karlsruhe Institute of Technology (KIT) and the AI experts of AMAI GmbH, has qualified for funding in the “AI Innovation Competition Baden-Württemberg” of the Ministry of Economics, Labor and Housing of the State of Baden-Württemberg. The project aims to use artificial intelligence (AI) to make packing and shipping processes in e-commerce more economical and environmentally friendly without compromising delivery times. A first demonstration with real data is to take place in the fourth quarter of 2021.

Cover image ai-based reinforcement learning
The AI learns via reinforcement learning to create the perfect package throughout the entire process.

AI is considered a key technology for realizing the highest possible value creation potential in a globalized, complex and data-driven world now and in the future. The aim of the innovation competition is to create collaborative research projects between research and industry, through which the transfer of AI technologies and knowledge to small and medium-sized enterprises is to be accelerated. Funding is provided for projects that are characterized by a high degree of innovation and complexity.

With “The perfect package”, the project partners around TUP are addressing a problem that has attracted worldwide attention not only since the Corona crisis: The packaging and transport processes in B2C e-commerce. Today, parcels are not packed efficiently and shipping vehicles are not fully loaded in order to meet the needs of end consumers who expect prompt delivery.

This is where the joint project of the software manufactory TUP, the Karlsruhe Institute for Materials Handling Technology and Logistics Systems and AMAI GmbH comes in: AI-based reinforcement learning can be used to find better solutions for the ideal packing of individual packages and the utilization of transport vehicles more quickly. The focus is on making packing and shipping processes more economical and environmentally friendly without compromising delivery times.

A Tetris playing AI in the e-commerce packaging and shipping process, will help to conserve resources and CO2

Already in 2018, 3.52 billion parcels were shipped in Germany, primary driver of this increase being the growing e-commerce sector. This growth offers tremendous opportunities for small and medium-sized businesses as they gain access to international markets through global transportation services and increasingly accessible web technologies. But this opportunity is often clouded by less-than-economical shipping processes.

Eduard Wagner, CIO of the software manufactory TUP, elaborates on the problem: “The focus on speed leads to the easy route of oversized packaging, especially for the usual multiple orders in e-commerce. After the order is placed, the CEP service provider then drives out a cubic meter-sized box containing three books, four pairs of socks, an HDMI cable and a game console among masses of packaging material. Which annoys the end consumer and results in wasted loading capacity in the transport vehicle.”

“But the problem is not easy to solve, which is already evident in its description as a ‘Multi-Level 3-Dimensional Multi Bin-Size Bin Packing Problem’ in the English-language literature,” continues Eduard Wagner. That means the allocation of the items of an order into packing materials, the packing sequence of the individual elements inside a packing material as well as the assignment to packing stations and delivery pallets within the warehouse, and finally the selection of the ideal delivery vehicle. There is little time available to make these decisions, which, moreover, consist of several conflicting factors.

Solving such problems is one of AI’s core strengths. Therefore, the project partners want to work together to create a tetris-playing AI that not only solves partial aspects, but finds the best possible solution over the entire process, from receipt of the order to packaging and loading. Unlike the algorithms used so far, the AI can be continuously optimized via reinforcement learning and thus adapt to a variety of warehouse configurations and business models.

From the optimization of partial aspects of the delivery process to an overall model from goods receipt to goods issue

Dr.-Ing Meike Braun, Project Lead at TUP
Eduard Wagner - CIO

“In research, there are already approaches in which AI solutions with reinforcement learning beat purely heuristic programs and find better solutions in less time,” says TUP project manager Dr.-Ing. Meike Braun. In particular, the best possible sequence, orientation and position of items in a packaging material are the subject of current research. What has been missing so far is the consideration of all interactions and interrelationships. In most cases, optimization only takes place within the sub-aspects of picking, packing and transport, as the individual process steps are already so complex that the interrelationships can no longer be taken into account by the algorithms used to date. “Thanks to the higher performance of artificial intelligence, we have the goal of solving all these sub-aspects together and, above all, simultaneously,” says Dr.-Ing. Meike Braun.

The goal of the AI project "The perfect package"

Overall, “The Perfect Package” is dedicated to assisting companies and institutions in becoming more resilient, competitive and resource-efficient in the areas of packaging and shipping. To do this, it is imperative that as many of the aforementioned sub-problems as possible are solved simultaneously and together, which will be enabled by an AI solution with a reinforcement learning approach. Due to the shorter run times to the best possible solution compared to classical heuristics, “The Perfect Package” can be realized considering all process steps. A first demonstrator is to be tested in Q4 2021 with real data from a large distribution center.

The project partners

At the Institute for Materials Handling Technology and Logistics Systems (IFL) of the Karlsruhe Institute of Technology (KIT), one focus is on the analysis of picking and packing systems. Heuristics and AI methods are used for this purpose, with particular emphasis on the transferability and application of the results in practice. For example, AI methods were developed and used in the “Pack Assistant” project to guide, control, and log the packing process of individual items.

The AMAI GmbH specializes in advising and technically supporting companies in the digitization and integration of artificial intelligence in products and services. The focus is on data science and machine learning topics. In the context of various logistics projects, AMAI has already been able to gain domain knowledge and build up corresponding competencies in the area of reinforcement learning.

The Dr. Thomas + Partner GmbH & Co. KG (TUP) is a medium-sized family business in Stutensee, whose specialty is information and control systems for goods distribution centers. The company goal of the software manufactory is to design and develop customized and modular, but also pioneering IT solutions for intralogistics and warehouse logistics and to implement them directly at the customer. In addition, there is ongoing support for the systems at the customer’s site. In all projects, the continuous optimization of material flows is the primary focus.

Contact and further information

For interview requests, please reach your contact person Peter Klement via redaktion@tup.com or the following form.

Images of Eduard Wagner and Dr.-Ing. Meike Braun as well as the project image can be found at this link to Google Drive.

The article is available in German as Word document and PDF file.