A planning and execution framework for robotic assembly with deformable objects using a centralized database based on the Rαβγ categorization

Abstract: Automation of process planning and execution of robotic assembly can lead to process optimization and shorter setup times. Several such automation frameworks have been developed for the assembly of rigid objects. However, many products require assembly with deformable objects. Robotic assembly with deformable objects typically incurs more complex dynamics and requires more collaboration during execution than rigid object assembly. In addition, process documentation includes more documents that are less structured. The current research suggests a data-driven planning and execution automation framework suitable for robotic assembly with deformable objects. The framework includes the three required modules: data extraction, process planning, and process execution. The modules interact with a central database constructed according to the Rαβγ ontology. Data extraction is based on commonly used manufacturing documents. Process planning is based on parametrized hybrid automata models, which encompass process and collaboration complexity using two layers: assembly operations and robotic skills. Process execution integrates a digital twin for sequence validation, process adaptation, and monitoring. The framework was successfully demonstrated in a small factory environment with three case studies for products with deformable objects: two smart light boards which include parts with plastic deformations (electric wires) and a medical infusion kit with parts with elastic deformations (tube, connectors). The framework facilitated optimized planning with significant reuse of assembly operations for all products. Both light boards had a high rate of assembly operation reuse (78%, 86%). The medical infusion kit had a somewhat lower rate (62%) due to the need for dedicated operations.