abat receives award: Inventory with drones and AI is product of the year 2022

We received the award for the Product of the Year 2022 from the trade journal materialfluss. The readers of the trade journal voted the inventory solution, which was developed together with the University of Oldenburg and Getränke Essmann, onto the winner's podium in the WMS & Co. category. Dag Oeing, from the management team, and René Kessler, Research Assistant at Carl von Ossietzky University Oldenburg, accepted the award at a virtual awards ceremony on March 16, 2022.

"We are very pleased that we were able to convince the trade audience of materialfluss with the drone and AI-supported inventory solution," says Dag Oeing. "The vote shows that our proprietary development meets the special requirements in logistics." Warehouse inventory is usually associated with time-consuming manual activities, he said. "It ties up personnel resources for counting and sometimes even detracts from day-to-day business."

René Kessler from the University of Oldenburg has a similar view: "The collaboration with abat and Getränke Essmann was so successful that we were also able to present the project at the international symposium "AAAI 2021 Spring Symposium on Combining Machine Learning and Knowledge Engineering" (AAAI-MAKE 2021)."

The automation of inventory is based on a three-step approach: In the first step, images are captured by drone. The images from the warehouse (indoor or outdoor) are extracted and preprocessed by data services. To achieve maximum results with as few images as possible, the solution uses various methods for augmenting the image material that have been tested and adapted in research and practice. The database is selectively augmented with image transformations to optimize the subsequent counting results.

A major advantage of this development is that no expensive hardware needs to be purchased. The drone used is a commercially available version with a 4K camera. Additional hardware such as complementary robots or hallway vehicles is not necessary. The training of the AI models is comparatively low in terms of time and resource input. The models used can be adapted to include other products not previously considered or new products.

"I congratulate abat on this great development" commented Martin Schrüfer, Editor-in-Chief of materialfluss in conclusion. "And I am very excited to see how this solution will help improve warehouse inventory in many companies."