Art Conservation Robot: From Scaffolds to ABB Arms

Conservators still do most of their work with small brushes, swabs and a lot of patience. An art conservation robot takes on the parts of the job that suit a machine: recording a painting or statue in fine detail, cleaning stone, guiding the making of exact replicas and, in one EU project, fitting fresco fragments back together. This post follows the work from restorers in museum workshops to laser scanners, Factum Arte's facsimiles and the RePAIR robot built for Pompeii. Then it looks at where ABB's precise arms could help and what still holds them back.

A restorer at work on stone at the Loggia dei Lanzi in Florence, 2008. Photo: Cyberuly / Wikimedia Commons (CC BY-SA 3.0)

Brushes, scaffolds and the first laser scans

For most of history, conservation meant a skilled person working close to the object. In 1904 the US Capitol had its own restorer of paintings, William H. Duckstein, photographed in an office crowded with portraits. The best known project of the 20th century, the restoration of the Sistine Chapel frescoes, including Michelangelo's ceiling, ran from 1980 to 1994. It brought back colours and details that nobody had seen for centuries. It also drew critics, among them the art historian James Beck, and the argument over it has never quite ended.

William H. Duckstein, restorer of paintings at the US Capitol, in his office among portraits, 1904. Photo: Library of Congress via Wikimedia Commons (public domain)

Easel paintings still get that kind of close attention. A conservator sits at a bench under a magnifying lamp, takes off old varnish with swabs and retouches losses with a fine brush, a small area at a time. Every step is written down, because the work should be reversible and the next conservator needs to know what was done. This habit of documentation is why scanners caught on: a 3D record of a surface shows cracks, chisel marks and old repairs, and nobody has to touch the object to get it.

A conservator retouches a small portrait under a magnifying lamp at the Skagens Kunstmuseer, 2016. Photo: Skagens Kunstmuseer / Wikimedia Commons (CC BY-SA 2.0)

The Digital Michelangelo Project at Stanford University and the University of Washington scanned Michelangelo's sculptures during the 1998 to 1999 academic year. A team of 30 people used a laser triangulation scanner built by Cyberware and mounted on a motorized gantry, and the resolution was fine enough to pick up the chisel marks. In 2001 Adam Lowe, Manuel Franquelo and Fernando Garcia-Guereta founded Factum Arte in Madrid, at first to record the Tomb of Seti I. Its Lucida laser scanner records surface relief without contact at 100 micron resolution, and the studio turns such scans into facsimiles.

Visitors stand before a large facsimile painting in the Factum Arte workshop in Madrid, 2019. Photo: Diario de Madrid / Wikimedia Commons (CC BY 4.0)

Scanners, lasers and robot hands in conservation now

Some cleaning has moved to machines as well. Lasers can lift dirt and crusts off stone without abrasives, and in 2012 laser cleaning was used on the marble facade of the US Supreme Court building. In 2015 Marco Ceccarelli and colleagues described a service robot that could survey frescoes and fill in missing areas on curved walls with an inkjet printer. Digitisation is moving fastest. On 30 September 2026 the BBC reported that more than 70 million specimens from 100 UK collections will go into an online digital museum, with robots and AI helping to scan them.

Laser cleaning of the marble facade of the US Supreme Court building, 2012. Photo: Architect of the Capitol via Wikimedia Commons (public domain)

The most hands-on robot project is RePAIR, an EU Horizon 2020 project coordinated by Ca' Foscari University of Venice. It targets two frescoes from Pompeii that sit in thousands of broken pieces in storerooms. Cameras digitise the fragments, machine learning looks for matching colours and patterns, and two robotic arms with soft hands lift the pieces and join them. Archaeologists confirm each match. "It's like you buy four or five jigsaw puzzles," said the coordinator, Marcello Pelillo. "You mix everything together, then you throw away the boxes and try to solve four or five puzzles at the same time."

The 3D scanner head mounted on an ABB robot in the 3DQI inspection cell. Photo: ABB

ABB has announced no conservation project, so anything it could do here is a possibility. Its 3D Quality Inspection cell, 3DQI, puts a 3D scanner on a robot and compares each scan with a CAD model, which is close to recording a statue or checking a replica against its scan. In September 2024 ABB gave the GoFa cobot an Ultra Accuracy option with 0.03 mm path accuracy. The dual arm YuMi IRB 14000 was built to handle small, light parts next to people, which is what a fragment table needs. Our posts on robotic stone carving and museum guide robots cover related work.

A GoFa cobot with ABB's Ultra Accuracy option, here fitted with a measuring probe. Photo: ABB

Where conservation robots go from here

In museums, the next few years will probably be about recording. A robot that moves a camera or scanner around an object at a fixed distance can cover thousands of objects with the same settings, while a person handles only the object itself. The same scans can serve small museums and private owners, who could order a copy for display and keep the original in storage. Workshops like Factum Arte's, which has worked with the British Museum, the Louvre and the Prado, already turn scans into facsimiles, and carving robots and printers can make more of them.

A painting conservator at work under a fume extractor at Kanut in Estonia, 2022. Photo: K4tr1nas44r / Wikimedia Commons (CC BY 4.0)

Repair is much harder than scanning. Every object is different, so a robot cannot follow one fixed program the way it does on a car line. The path has to come from the object's own scan and be checked for reach and collisions before the tool moves, which is the kind of offline work engineers do in RobotStudio. Force control matters just as much. Old plaster, paint and stone can break under a small push, so any arm working on them needs firm limits on force and speed.

An ABB YuMi dual arm robot handling small parts beside a lab worker. Photo: ABB

Conservators decide what to clean and what to leave alone, and RePAIR keeps archaeologists in the loop for exactly that reason. Money is another limit. Museum budgets are small next to factory budgets, so a robot cell would probably have to be shared or rented. Conservators would need training in scanning and robot programming, and engineers would need to learn why a reversible repair matters more than a fast one. For now the robots scan, lift and hold, and a person makes the decisions.

A full size sculpture replica in the Factum Arte workshop in Madrid, shown to visitors in 2019. Photo: Diario de Madrid / Wikimedia Commons (CC BY 4.0)

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com