News
Polish Team Brave Mind Fighters Reaches Top 4 of Global Robot Combat Tournament
A seven-person Polish team of AI engineers and researchers advanced as the only European team to the TOP 4 of the international URKL humanoid robot combat tournament in Shenzhen, training their machine partly on motion capture footage of a stunt performer.
Contents
The Polish team Brave Mind Fighters has advanced to the top four of the international URKL 2026 humanoid robot combat tournament held in Shenzhen, China. It is the only European team to make it this far among more than 200 squads entered in the competition, and it taught its robot to fight partly through motion capture recordings of a real martial arts specialist.
The Road to the Final
The URKL tournament, short for Ultimate Robot Knock-out Legend, is a humanoid robot combat league organized by the Chinese company EngineAI. More than 200 teams from over a dozen countries, including China, the United States and Europe, entered the competition. After an online qualification stage, several dozen teams advanced to the final phase in Shenzhen, and Brave Mind Fighters placed among the world's top four.
The key difference from earlier robot combat shows in China is that all teams in URKL work with identical hardware - the humanoid robot EngineAI T800, standing 173 centimeters tall and weighing between 75 and 85 kilograms depending on the version. Since the machine is the same for everyone, the outcome is determined solely by the quality of the software, training data and motion control algorithms, not by any hardware advantage.
Who Are Brave Mind Fighters
The team consists of seven Polish engineers and AI researchers: Arkadiusz Cybulski, Piotr Gerke, Konrad Suchodolski, Nikodem Kuliś, Walery Kusznirski, Jacek Caban and Piotr Zarzycki. The core of the group comes from the Humanoid scientific club at the Faculty of Mechatronics of Warsaw University of Technology (Politechnika Warszawska), joined by students from the Faculty of Electronics and Information Technology.
The project is also supported by the HumanTech Center at SWPS University and the community around initiatives such as Śniadania Robotyczne (Robotics Breakfasts) and the community around the Polish language model Bielik. The team stresses that it operates independently, without support from major corporations, and is actively seeking sponsors and strategic partners to help fund the trip to China and access to lab infrastructure.
Martial Arts as a Movement Template
Instead of manually programming every punch and dodge, the team relied on reinforcement learning - in simulation, the robot repeats sequences of movements millions of times until it learns to keep its balance, land strikes and react to contact with an opponent. The starting point, however, was real human movement: the team recorded the motion of a friend who works as a stunt performer and martial arts specialist, whose body served as the template for the machine.
The recordings were fed into a custom motion capture system built by the students on a low budget, using Xbox Kinect cameras. The captured movement then went through a retargeting process, transferring motion sequences from the human body model to the robot model while preserving correct posture, tempo and continuity.
We built our own low-cost motion capture system, and then transferred the recorded movements onto the robot model - Konrad Suchodolski, Brave Mind Fighters
What's Next for the Team
The next stage is the trip to Shenzhen, where the team will for the first time test its algorithms on a physical unit of the EngineAI T800 robot provided by the organizer, rather than only in simulation. The tournament's grand prize is a golden championship belt worth 10 million yuan, or roughly $1.44 million, though for the team the experience of working with physical artificial intelligence matters most.
For Polish robotics and the AI community, this is a rare chance to make a mark in an arena dominated by Chinese companies, which for months have been showing off humanoid robots dancing, running marathons and fighting in a cage. The success of a student team without corporate backing stands in contrast to the model followed by Chinese robotics giants, which are pouring billions of dollars into humanoid development.
The team says that regardless of the result in Shenzhen, it intends to use the experience gained and the tools it developed - its own training engine and a low-cost motion capture system - in further research projects related to Physical AI, meaning artificial intelligence that operates in the real physical world rather than only in simulations or in chat.
