Humanoid Robots Shatter Sprint Records at Beijing’s World Humanoid Robot Games

The second World Humanoid Robot Games concluded in Beijing on August 26, marking a significant milestone in the integration of advanced robotics and artificial intelligence. Held at the National Speed Skating Oval, famously known as the "Ice Ribbon," the five-day event featured over 2,000 robots competing in a series of athletic and practical challenges. The competition served as a high-profile demonstration of rapid advancements in hardware engineering, motion-control algorithms, and embodied-intelligence foundation models, with several machines outperforming human world records in track events.

In the marquee 100-meter sprint, the Tiangong Ultra, developed by the Beijing-based firm X-Humanoid, secured victory with a record-breaking time of 8.64 seconds. This performance, alongside other track results, highlighted a dramatic leap in capability compared to the previous year’s competition, where winning times were significantly slower. While the speed of these machines has reached new heights, the event also underscored the ongoing challenges in mechanical design, particularly regarding braking systems, as some robots were observed colliding with safety mats after crossing the finish line.

The rapid improvement in robot athleticism is largely attributed to an across-the-board upgrade in core components. Han Gang, an expert in motion-control algorithms at the Beijing Innovation Center of Humanoid Robotics, emphasized that the shift toward higher-torque joint modules has been a primary driver of performance gains. "Last year, we were using joint modules with a torque of 300 newton-meters. This year, higher-torque joint modules have become widely adopted across the industry, driving robots to run faster and faster," Han noted.

Beyond torque, the integration of batteries with higher discharge rates and the use of lighter, stronger materials have allowed for more aggressive movement profiles. These hardware improvements are coupled with sophisticated control technologies that enable robots to maintain stability during high-speed maneuvers. The result is a generation of humanoid machines that are not only faster but also more capable of executing complex physical tasks that were previously considered beyond the reach of autonomous systems.

While hardware provides the physical foundation, the most significant developments were observed in the robots' "brains." During the opening ceremony, a humanoid robot named Xingzai, developed by Galbot, engaged in a high-speed tennis match against former professional player Zheng Jie. The robot demonstrated the ability to make autonomous decisions in real-time, even recovering quickly from an unexpected fall to continue the rally. This performance was powered by AstraBrain, an embodied-intelligence foundation model that integrates decision-making, neural control, and physical movement.

Zhao Yuli, chief strategy officer of Galbot, explained the architecture behind this capability: "The 'brain' is responsible for understanding and decision-making, while the 'cerebellum' translates those decisions into physical movements." By training in virtual environments for millions of rounds, the system has effectively simulated decades of human experience, allowing the robot to adapt to dynamic scenarios, including doubles matches where it must coordinate with human teammates.

The Games extended beyond speed, introducing 21 scenario-based events to test robots in industrial, logistics, and service environments. A central focus was the dexterous-hand competition, which evaluated fine-motor skills such as weighing powder, opening containers, and manipulating small objects with tweezers. The OmniHand, developed by AgiBot, dominated this category by winning seven of the eight available gold medals. The device features 16 degrees of freedom and has already seen significant real-world deployment, with over 20,000 units shipped.

Qiao Tianjie, CEO of AgiLink, a subsidiary of AgiBot, highlighted the importance of these interfaces: "Dexterous hands are a critical interface through which humanoid robots interact with the physical world. They set the upper limit for robots' ability to perform precision tasks, adapt to different scenarios, and operate in real-world environments." This focus on precision is essential for the industry's goal of creating robots that can function autonomously in hazardous or complex manufacturing settings, such as engine assembly lines.

The practical application of these technologies remains the ultimate benchmark for the industry. In the power tool assembly event, the LinkerBot team achieved a gold medal by completing 18 screw installations in five minutes without human intervention. Zuo Jiaping, co-founder of LinkerBot, noted that the industry is moving toward an "industrial closed loop" where robots are increasingly used to manufacture other robots. This shift is driven by the urgent need to automate precision processes and replace human labor in dangerous environments.

To foster continued innovation, the Beijing Municipal Bureau of Economy and Information Technology released a comprehensive dataset at the closing ceremony. Containing over 2,500 hours of real-world operational data from the Games, the initiative aims to provide researchers with the resources needed to validate new ideas. As Jiang Guangzhi, head of the bureau, stated, the goal is to "give industry the real-world scenarios it needs to turn experiments into applications." Despite the impressive demonstrations, industry observers like Hua Rong of Zeroth caution that the true test lies in long-term reliability: "The competition will be whether, after the product is sold, it can really solve users’ problems."