If you spotted a three-wheeled robot moving among people in central Prague with a researcher by its side, it may well have been Helhest. Over the past few months, researchers from the VRAS group have been taking it “for walks” in places such as the park in Charles Square, Kateřinská Garden and the Hvězda Game Reserve. They were training for RoboTour, a competition for autonomous robots.
The people they met along the way were delighted by the robot. “Children in particular love running after it. And when they realise it avoids them, they start standing in front of it too. Every day, several people came over to ask what Helhest could be used for,” says Associate Professor Karel Zimmermann from the Department of Cybernetics at FEE CTU. “One of its most straightforward applications is autonomous surveillance. For example, the robot could replace a night security guard patrolling large outdoor areas. For this purpose, it also has advanced sensors, such as thermal imaging cameras and detectors for toxic gas leaks,” he explains.
Helhest, named after a three-legged horse from Danish mythology, can reach speeds of up to 20 km/h and overcome obstacles up to half a metre high. It weighs 100 kg and can carry a load of the same weight. Its design allows it to right itself after overturning and continue its mission. In extreme conditions, it can even operate “upside down”. Last year, members of the VRAS group put it through a demanding exercise in the Alps organised by the Austrian Armed Forces and Graz University of Technology (TU Graz).
On Saturday 19 September 2026, the VRAS researchers took Helhest to the RoboTour competition. Held every autumn by Robotika.cz, the event brings together professional and amateur developers and enthusiasts interested in autonomous machines. Judges for the individual challenges are selected on site from competing teams, which include both university representatives and amateurs. Any disputes are settled by the competition’s president and founder, Martin Dlouhý, who has been organising the event for 20 years.
A leaf-covered path complicated the competition
The team was led by Associate Professor Karel Zimmermann and Bedřich Himmel. The lead systems integrators were Dr Martin Pecka, Adam Herold and Jan Vlk. Vlk received a scholarship from Valeo to support his research and is also co-supervised by Valeo researcher David Hurych. Students Zuzana Jindrová, Jakub Adamec, Matyáš Himmel and Martin Erben also contributed to the team’s work.
The task required precise autonomous navigation along park paths to repeatedly load, unload and deliver a five-litre keg of beer, while automatically avoiding moving obstacles such as people and competing robots.
According to the researchers, the greatest challenge was distinguishing between a footpath and ground that was no longer part of it. Leaving the path would result in disqualification. “The rules state that, in ambiguous cases, the decision on disqualification will be based on how a person would interpret the situation. But there are many well-trodden shortcuts that people regularly use. No grass grows on them, so their surface looks like a footpath. Yet people know they are not official paths. And that caught us out during the competition,” explains Associate Professor Zimmermann.
This was exactly what happened to Helhest in one of the competition rounds. A path forming part of the route was partially covered in leaves. Unfortunately, a well-trodden shortcut ran alongside the kerb, and the robot moved onto it, believing it was still on the course. It was therefore disqualified from that round.
Even so, the team still had a victory to celebrate. Helhest ultimately won an award for achieving the highest score in the competition’s history in the category for autonomous outdoor navigation using only OpenStreetMap map data. It competed against 11 other teams in this category this year.
