Assembly Task Modelling Method for Automatic Robot Program Generation

The assembly of parts is a common task in the manufacturing industry. Robotic systems are used to automate repetitive processes, but the long time required for writing and testing programs, limit the application to cases of high-volume production. Modern production systems are expected to quickly adapt to product variants and process changes, and robots should be programmed in a flexible manner to allow for quick reconfiguration. In this paper, we present a method to model assembly tasks in a Digital Twin platform that allows the simulation and control of robots using a Visual Programming editor. We enrich geometric data of parts with 3D frame annotations that are used to design assembly steps. Furthermore, we introduce an interface where each step of the assembly sequence is matched with a ServiceNetwork, a reusable program that can solve a specific task. This produces a sequence of blocks in the Visual Programming editor, where robot poses are generated automatically based on the assembly data. The presented method allows for the flexible reconfiguration of robotic assembly tasks based on digital product data without the need of teaching robot poses. The process is demonstrated by assembling a product in simulation and comparing it with other methods for task reconfiguration.