University of Bremen: Artificial intelligence helps people with assembly

The assistance system analyzes the process data collected at the assembly station and camera-based information. Image processing and machine learning methods are used with regard to the ergonomic and production-related work situation. The system checks the assembly process as well as the quality of the finished product. It integrates employee-centered assistance functions. Until now, assistance systems only focused on the products to be manufactured and their quality.

“The focus is on the working person with his or her individual abilities”
“We don’t want to replace human work with the assistance system, we want to support it in the best possible way. It should become easier and also be possible for those who are disadvantaged in the labor market, ”says Christoph Petzoldt, BIBA scientist and project manager. “This assistant provides individual support as required, and it motivates and creates a constructive working environment. We put the working people with their individual abilities in the foreground. “

Thats how it works
Optical sensors (depth cameras) record the progress of the process at the workplace from several perspectives, for example handles such as removing the individual components from the storage containers and the assembly activity itself. The cameras deliver their image and depth data to the system, which in a first step recognizes and evaluates them in real time using image processing methods.

In the analysis to assess the posture (ergonomics) and for tracking the hand movements of the worker (English hand tracking ) sets the system to methods of “Deep Learning”. This stands for “in-depth learning”, a key technology in Artificial Intelligence (AI). The system gets better with each of its calculations, because it builds its analyzes and forecasts independently on what has already been learned. Based on this analyzed information and general process data such as process times and errors, the assistance is then individualized.

The information generated with the help of the AI ​​is processed by the system for a wide range of uses – initially for manual work directly at the specific assembly station: projections on the work surface are used to display the current work. If required, the workers can also receive additional information and help – on the one hand, on technical assembly work with options to learn on the side, and, on the other hand, on health-friendly, individual optimization of their posture at work, ergonomics.

By monitoring the progress of assembly, providing targeted information and taking into account the needs of the assembler, the system increases process efficiency and assembly quality and improves the work situation through specific support using motivation and training strategies and techniques.

Motivation and learning through “gamification”
Another goal is the better and more targeted qualification of employees as well as increasing self-motivation. “Gamification” stands for systemic motivation through incentives and describes the process of learning through play using techniques that originally come from the world of computer games and have been further developed for use in industry. By means of gamified representations of the information, the work process is made more ergonomic and stimulating for the workers. In addition, they learn “playfully” during their practical work. Using the data supplied by the AI ​​system, the gamified elements are controlled by the game design concept.

Efficient, effective and with a high level of user acceptance
With the help of the new assembly assistance functionalities through the combination of informational process management with projection, automatic monitoring of assembly process and component progress, ergonomic posture recognition and incentive-based gamification, a significant reduction in assembly errors and process times was achieved. High increases in efficiency were found, particularly in the case of confirmation steps. In addition, the user studies accompanying the project showed that the measures to support and create incentives lead to a high level of acceptance among workers.

“Ensuring that small and medium-sized companies also participate in Industry 4.0 development”
“With this assistance system, a solution has been created that takes social and economic aspects into account and takes into account the current labor market situation. It enables – while at the same time guaranteeing high product quality – the integration of people who today still often fall through the selection grid of HR departments due to age, handicap or low qualification, ”says BIBA head Professor Michael Freitag. “The system can be used for manual assembly processes in companies of all sizes and industries. It also enables small and medium-sized companies to participate in the rapidly advancing Industry 4.0 developments. “

Key data for the “AxIoM” project
The cooperation project “Gamified AI assistance system to support the manual assembly process” (AxIoM) under the direction of BIBA ran for 22 months. Armbruster Engineering, a Bremen-based specialist for assembly assistance systems, was involved as a development partner. The project was supported and supported by the Bremer Aufbaubank (BAB) with funds from the European Regional Development Fund (ERDF). The result is the prototype of an AI-supported assistance system for an assembly station in manual production such as the manufacture of small series.