From UFC to World Combat Ring! How Does CHINGMU's MoCap Help Robots "Learn" Combat?

Source：

2025 Aug. 28

The year 2025 can be called the breakthrough year for the robotics industry. From robots performing the Yangko dance at the CCTV Spring Festival Gala to the grand hosting of the world's first World Humanoid Robot Olympics, robots have undoubtedly become the brightest stars in the tech field. Behind this wave of excitement is the industry's urgent pursuit of making robots more agile, stable, and closer to human movement logic. As a leading motion capture service provider in China, CHINGMU provided crucial technical support for Unitree Robotics'humanoid robots to efficiently replicate human fighting movements.

CMG World Robot Fighting Championship & UFC Performance Institute:  

From Arena Competitions to Human Interaction, Equipping Robots with "Real Skills"  

In May this year, Unitree Robotics'humanoid robots made an appearance at the CMG World Robot Fighting Championship—the world's first fighting competition featuring humanoid robots as participants—showcasing precise and professional fighting combinations. Shortly after, they stepped into the UFC Performance Institute, the world's largest and most influential professional mixed martial arts training center, for a real-life "Real Steel" interaction with professional fighters. Although their strength still falls short compared to human fighters, the fluidity and speed control of the robots'punches have reached an astonishing level. To enable these natural and fluid fighting movements, Unitree Robotics utilized CHINGMU's optical motion capture system for pre-competition training data collection in both of these significant events, rapidly enhancing the robots'professional fighting performance.

Unitree's Humanoid Robot G1 Engages in Fighting Interaction with Athletes. Image Source: UFC

In the dynamic interactive environment of combat, the control algorithms of robots must achieve extremely high precision and stability. During combat, the robot must not only execute its own attacking and defensive actions but also respond to external force disturbances caused by the opponent's attacks, maintain balance at all times, and perform complex movements such as moving, punching, dodging, and side kicking. This requires a vast amount of combat data to support the algorithms and extensive training to enable the robot to "understand" human fighting techniques.

To accurately capture the "essence of movement" in human combat sports, Unitree Robotics utilized CHINGMU's optical motion capture system to collect data on multiple core metrics of fighters, including posture and speed. The system not only recorded the angular changes of limb joints but also captured detailed information such as punching speed and the continuity of motion trajectories. Yang Ya, Sales Director of Unitree Robotics, stated, "I believe CHINGMU's response speed is exceptionally fast, and the precision of the data collection, including its seamless integration with our robot models, is outstanding."

A martial artist performs combat movements while their motion data is captured via CHINGMU's optical motion capture system.  

The combat data collected from top-tier athletes not only accurately还原 the fundamental principles of human movement, providing a reliable blueprint for robot training, but also serves as a critical reference for Unitree's humanoid robots to break through the limitations of mechanical motion and replicate the fluidity and naturalness of human combat techniques.  

Continuous Learning and Evolution: Building Ever-Improving Combat Robots  

Training combat robots is not a one-time effort. With technological advancements, the emergence of new fighting techniques, and increasing performance demands, continuous learning and evolution are essential. Yang Ya noted, “This time, we collected combat data from UFC professional athletes using CHINGMU's motion capture system, which includes boxing, Muay Thai, and some jiu-jitsu movements. In the future, we will use this data to train our G1 humanoid robot to perform a wider variety of actions.”

Yang Ya, Sales Director of Unitree Technology, stated that using motion data captured by CHINGMU's motion capture system, the company trains robots to perform combat movements.  

To further validate and evaluate the locomotor performance of the combat robots, Unitree Technology utilized real-time data from CHINGMU's motion capture system to assess the robot's action execution during simulated training and carried out related control algorithm optimizations. It is through this closed-loop process of "data collection → training validation → feedback optimization" that the robot can undergo rigorous refinement of its capabilities before entering real combat arenas.  

From the professional showdowns between robots at the CMG Fighting Championship to the real-world interactions between robots and human fighters at the UFC Performance Institute, humanoid robots have undoubtedly become top-tier attractions in the tech field. CHINGMU's optical motion capture system not only assists research teams in humanoid robot training but also offers customized solutions such as locomotor performance evaluation, dexterous hand manipulation, and teleoperation. This extends the application value of robots to diverse scenarios including industrial, medical, and rescue operations, opening broader horizons for the intelligent evolution of robots.  

In the future, as motion capture technology becomes more deeply integrated with robot development, CHINGMU will continue to advance innovation in motion capture technology, enabling robots to break through mechanical limitations across more fields. Together with industry partners, CHINGMU aims to pioneer a new chapter in human-robot collaboration and coexistence.