RoboGrind: Intuitive and Interactive Surface Treatment with Industrial Robots

TL;DR

RoboGrind tackles robust automation of surface finishing in remanufacturing by integrating high-fidelity 3D perception, AI-assisted program synthesis, task-driven path planning, and force-controlled execution. The three-step workflow—3D surface scanning, AI-driven robot programming, and force-controlled execution—enables rapid, interactive generation and validation of robot programs for sanding, polishing, and deburring. A KnowRob-based cognitive loop (meta-wizard) with a natural-language interface grounds tasks in a domain-specific ontology, while a hybrid force-position controller ensures safe manipulation in contact. Experimental results in laboratory and real-world wind-turbine blade remanufacturing demonstrate competitive perception accuracy, planning precision, and force control performance, marking a significant step toward automated, data-driven surface finishing. The work highlights practical impact for remanufacturing and provides a foundation for extending to broader industrial finishing tasks, with future directions including task learning from demonstrations and broader domain applications.

Abstract

Surface treatment tasks such as grinding, sanding or polishing are a vital step of the value chain in many industries, but are notoriously challenging to automate. We present RoboGrind, an integrated system for the intuitive, interactive automation of surface treatment tasks with industrial robots. It combines a sophisticated 3D perception pipeline for surface scanning and automatic defect identification, an interactive voice-controlled wizard system for the AI-assisted bootstrapping and parameterization of robot programs, and an automatic planning and execution pipeline for force-controlled robotic surface treatment. RoboGrind is evaluated both under laboratory and real-world conditions in the context of refabricating fiberglass wind turbine blades.

Figures (6)

• Figure 1: RoboGrind is an intuitive, interactive system for robotic surface treatment comprising perception, program generation, planning and control.
• Figure 2: RoboGrind integrates perception (1), AI-assisted programming (2), planning and force control (3) into a comprehensive assistance system for robotic surface treatment.
• Figure 3: Illustration of the meta-wizard's interactive symbol grounding mechanism. The meta-wizard's main control module retrieves symbolic knowledge about tasks, workpieces, etc., via Prolog queries to a domain-specific reasoner connected to the KnowRob krr engine. Missing information about the concrete task, workpiece, etc., is obtained via dialog with the user.
• Figure 4: Path planning on point cloud data.
• Figure 5: Test setup under real-world conditions.