Germany-based FMC³ Robotics has introduced Doceo, a four-part system designed to teach industrial robots manufacturing tasks by capturing how factory workers perform them. The technology gathers information during normal production work, aiming to reduce reliance on dedicated robot training facilities and simulated environments.
The company says access to detailed physical-world data is a major obstacle to developing robots that can manage complex movements, handle objects and adapt to production changes, according to Robotics & Automation News. Originally reported by Robotics & Automation News.
From worker demonstrations to robot controls
Doceo RealCap records visual, depth, movement and tactile data using camera headsets, external cameras and optional touch-sensitive gloves. Doceo Hub aligns those streams, brings together different viewpoints and supports recording, playback, annotation and data management.
Doceo Cloud handles ingestion, automated labeling, curation, model training, simulation and testing, drawing on Vector Informatik's cloud and data infrastructure technologies. Doceo Driver links the trained capabilities to various robot platforms and control systems. The components connect factory demonstrations with model development, validation and deployment through FMC³'s Industrial Embodied AI Agent Platform.
Factory trials target bins and electrode sheets
At a consumer electronics plant, FMC³ is testing a wheeled robot that moves bins to designated production-line locations. The trial involves different container sizes and multiple workstations, with the aim of avoiding extensive changes to the factory layout.
A separate project at a battery cell production facility is developing a robot to lift electrode sheets and place them precisely on trays. Both projects are proofs of concept, assessed on cycle time, placement precision and duration of continuous operation.
Joint venture seeks reusable training data
Separately, FMC³, Vector Informatik and PEM Motion are working toward a German joint venture to standardize data infrastructure for industrial embodied AI. The collaboration would combine robotics, production engineering and data technologies to make training information more consistent, traceable and reusable across factories and robot applications. The company's broader aim is to let robots improve through production feedback, reducing engineering work and retraining when tasks change.
Originally reported by Robotics & Automation News.