ROOKIE

 

Improving flexibility in assembly using an ontology and AI-based methodology with integrated expert knowledge

 

Motivation

Flow chart for a robot-based assembly © Oliver Petrovic Method for a flexible, robot-based assembly. A: Synthetic data generation. B: Ontology-based framework. C: Machine learning procedure

Due to a high variety of variants, short product life cycles and small batch sizes, small and medium-sized businesses often resort to manual processes in assembly. Compared to an automation solution, this leads to high labor costs and varying quality of the products. A technology combination that has the potential to counteract this problem is AI-based robotics. Using AI, robot-based applications can be implemented flexibly, autonomously and robustly. This means that automation can be justified even with smaller quantitites and the risk of bad investments can be reduced. Hereby, one challenge is to get the required data quality and quantity for training the AI models, which i soften difficult to achieve, expecially with small quantities.

 

Goal

As part oft he proposed research project „ROOKIE“, a comprehensive methodology with an ontology-based modeling approach and synthetic data generation is developed, which enables the use of AI-based robot operations in industrial assembly. This should increase the flexibility, autonomy and robustness of the planning, commissioning and execution of robot-based assembly operations, and result in accessability of autonomous assembly processes for small and medium-sized businesses.

 

Approaches

Possible solutions are the enrichment of smaller data sets with known engineering information in combination with the integration of expert knowledge. AI algorithms are used to enable flexible and automated execution of assembly operations.