Fencing Robot: Practice Partners and ABB Arms

A fencing robot would be a practice partner that never gets tired and will repeat the same attack a hundred times at six in the morning. Nobody sells one yet, though people have been working toward it for years. This post starts with how fencers trained with a master and club mates, and how electric scoring turned the weapon into a sensor. It then follows the university robots that have tried to parry and attack with a blade, the AI video analysis Korean fencers used this autumn, and where ABB robots could fit. It ends with what still stands between today's prototypes and a robot in a fencing club.

Stergios Delis lunges against the parry of Nikolaos Theodoropoulos at the 2nd Leonidas Pirgos Fencing Tournament in Greece, 2019. Photo: George E. Koronaios / Wikimedia Commons (CC0)

From the fencing master to the electric touch

For most of the sport's history, a fencer learned from a master in a one to one lesson and then fenced club mates. The master fed attacks at a chosen speed and distance, and the pupil answered with the same parry or riposte over and over. Fencing was on the programme of the first modern Olympic Games in 1896, and clubs of that era used simple aids for solo work. A 1936 photograph of the Swords Club, taken by the Australian photographer Sam Hood, shows round targets fixed to the wall, which fencers could lunge at alone to train point control. Those targets stayed put and never answered back, so they trained accuracy and little else.

Men and women fence at the Swords Club in 1936, with round practice targets on the back wall. Photo: Sam Hood, State Library of New South Wales via Wikimedia Commons (public domain)

Machines first entered fencing to judge bouts. Before electric scoring, officials stood around the piste and watched for hits, and disputes were common because a blade moves faster than the eye can follow. In 1933 the Laurent-Pagan electric apparatus replaced side judges in épée: a touch closed a circuit and lit a lamp. Foil followed in 1956 and sabre in 1988, and each weapon became a sensor wired to a scoring box. For anyone building a robot sparring partner this is useful, because the machine can know exactly when it was hit.

A hand-kept épée scoreboard at the 1952 Helsinki Olympic Games. Photo: Unknown photographer / Wikimedia Commons (CC BY 4.0)

Robots that hold a blade came out of universities. In 2011 Stanford students in the Experimental Robotics course spent three weeks building JediBot, a robot arm that fenced people with foam lightsabers. It attacked with preloaded patterns and used two Microsoft Kinect sensors to see and block the incoming blade. In 2015 Ignacio de Erausquin and Humberto Gonzalez at Washington University in St. Louis described a robot that blocks an opponent's sword using a single camera on the arm. Their simulation modelled a FANUC LR Mate 200iC with an 880 mm sword and reached blade contact in 145 ms. A 2023 Princeton senior thesis, Moyuka, built a lightweight arm that defends a target for foil fencers.

A Microsoft Kinect sensor, the type Stanford's JediBot used in pairs to see a human's blade. Photo: James Pfaff / Wikimedia Commons (CC BY 2.0)

How AI and robots help fencers train today

The most advanced fencing technology in use this year analyses video. SK Telecom, lead sponsor of the Korean Fencing Federation since 2003, built an in-house system called Fencing AI Nexus. According to The Korea Times on 1 October 2026, it uses computer vision and sensorless motion capture on ordinary match footage to pick out the moments when points are won and lost and to track athletes' joint movements. The company says it can cut repetitive video analysis work by up to 90 percent. At the 2026 Aichi-Nagoya Asian Games, Korea's fencers won 10 medals, four of them gold.

Épée fencers training at the Athenaikos Fencing Club in Athens. Photo: George E. Koronaios / Wikimedia Commons (CC BY-SA 4.0)

A robot sparring partner would need that kind of perception and a fast, safe arm to go with it. It has to read the fencer's blade and body and then parry or attack within a fraction of a second, which is why the Washington University team measured its result in milliseconds. Its hits must be light enough to hurt nobody and accurate enough to train point control, and it has to survive being struck all session. Each research robot so far handles one slice of this, such as blocking or presenting a target, and none of them can fence a full bout.

Pistes, cameras and screens in the hall of the 2023 World Fencing Championships. Photo: Dispe / Wikimedia Commons (CC0)

ABB does not make a fencing robot and has announced no fencing project, so what follows is a possibility. Its cobots are the closest fit. The GoFa CRB 15000 senses contact and stops, and it can be taught by moving it by hand, so a coach could guide the arm through a parry drill and have it repeat. The faster SWIFTI CRB 1100 uses SafeMove to slow down when a person comes near. ABB's Externally Guided Motion lets outside software, such as a camera system tracking a blade, steer an arm in real time. Google took that route with ABB arms for its table tennis robot research.

ABB's GoFa and SWIFTI cobots at their launch in February 2021. Photo: ABB

Where robotic sparring partners could go next

For clubs and home salles, the first useful products will probably be simple. An arm in the style of Moyuka that moves a target, or a blade that offers a parry for the fencer to beat, would cover a lot of solo training, from footwork and distance to accurate hits on a moving goal. Because it does not attack, it would be cheaper and safer than a full sparring robot. A coach could load drills into it for pupils to work through between lessons, with the electric scoring box recording every touch.

Early morning training at the Athenaikos Fencing Club, with a diamond wall target on the left. Photo: George E. Koronaios / Wikimedia Commons (CC BY-SA 4.0)

Safety is the hardest problem. A robot that attacks has to stay within the forces a fencing jacket and mask can absorb, and an arm fast enough to trouble an elite fencer moves faster than normal collaborative speed limits allow. Industrial arms and their safety systems are also expensive for a club, and a robot that fences like a person is still far off. Factories have a similar question to answer about robots reacting to people who move fast and unpredictably near them, so fencing robots and factory safety tools such as SafeMove zones on ABB controllers deal with the same physics.

A class works with ABB's IRB 1010, which ABB pitches for education. Photo: ABB

Skills may decide how quickly this arrives. Coaches know fencing but rarely know robot programming, so a training arm would need drills a coach can set up without writing code. ABB takes that approach with Wizard and lead through teaching, covered in our guide on how to control robots with no code. Engineers building a prototype can model the piste, the arm and the blade in RobotStudio first and test reach and speed there, before a real fencer ever stands in front of it.

An engineer programs a robot cell in RobotStudio next to an ABB industrial robot. Photo: ABB

Innovation

AI solutions for effortless ABB robot control.

Automation

Robotics

ceojohntran@chatwithrobot.net

+84905311611

© 2025. All rights reserved.

qtran1215@gmail.com