Brazilian programmers are using Chinese-made humanoid robots as a platform to test AI, motion-control software, and soccer strategies.

The feat is on display at the second World Humanoid Robot Games (WHRG) in Beijing, highlighting how international collaboration is helping accelerate the development of embodied AI.

Team Brazil, assembled from young researchers at five Brazilian universities, is competing in 5-on-5 humanoid robot soccer using robots supplied by Chinese robotics company Booster Robotics. 

Rather than designing and manufacturing the machines themselves, the Brazilian researchers are concentrating on the software layer — developing algorithms that allow the robots to perceive the field, locate the ball, coordinate movements, choose actions and execute kicks.

Coding robot soccer

Building a capable humanoid requires expertise in mechanical engineering, actuators, batteries, sensors, embedded computing and control systems. Using an established development platform lets researchers spend more time improving AI, perception, and decision-making. 

Booster Robotics’ T1 is designed specifically as a development platform. The humanoid has 23 degrees of freedom, including six in each leg, four in each arm, two in the head, and one at the waist. Its head incorporates a depth camera, microphone array, and speaker, while the control electronics and battery are housed inside the torso. The platform is designed to support walking, interaction, motion control, and research applications.

For robot soccer, however, hardware is only part of the challenge. A machine must continuously interpret its surroundings and translate that information into coordinated physical movements. Cameras and other sensors provide information about the ball, teammates, and opponents, while software determines where the robot should move and when it should attempt a kick, China.org reported.

The robot must also maintain balance while moving, recover from disturbances, and coordinate multiple joints in real time. Unlike conventional wheeled robots, humanoids must solve the additional problem of keeping a relatively tall body stable while walking, turning, and striking a moving object.

These requirements make soccer an unusually demanding test of embodied intelligence. A successful system needs perception, localization, motion planning, balance control, and decision-making to work together rather than as isolated functions.

Humanoids meet challenges

The Brazilian team’s work is happening amid a much larger expansion of humanoid robotics in China. The 2026 WHRG features 2,056 robots from 666 teams representing 16 countries and six continents. The five-day event has expanded to 51 events, including 30 competitive contests and 21 scenario-based challenges, compared with 26 events at the inaugural competition in 2025.

China accounts for the overwhelming majority of this year’s entries, with 1,975 robots from 641 teams involving 157 companies and 200 universities and research institutions. Organizers say the number of participating robots has roughly quadrupled from last year’s event, reports China.org.

The competition is also moving beyond demonstrations of basic movement. Athletic events test running, jumping, weightlifting, martial arts, gymnastics and soccer, while scenario-based competitions are designed around tasks that humanoids could eventually perform in workplaces and public environments.

This year’s scenario challenges include industrial production, logistics, emergency response, retail, office services, hospitality, household tasks and electric-vehicle charging. Robots are being tested on activities such as manipulating objects, sorting materials, performing maintenance and interacting with equipment. Some challenges require the machines to operate autonomously, making them a test of software reliability as much as mechanical capability.

The distinction is important because impressive athletic demonstrations do not necessarily translate into useful workplace robots. Recent competitions have shown humanoids capable of extraordinary running speeds, including machines that surpassed the human 100-meter world record under robot-specific competition conditions. At the same time, robots have struggled with stopping, maintaining balance and completing delicate physical tasks.

According to experts, the gap between movement and practical usefulness is where international software development could become particularly valuable. Researchers can train and test algorithms on the same robotic platforms while adapting them to different environments, rules and operational requirements.