On the night of August 23, at the Second World Humanoid Robot Games, robot competitors were standing by. The smallest robot at the event, the 65-centimeter-tall “Xiao Qing” from CUEB, drew everyone’s attention among robots averaging one meter in height. 

CUEB student team interviewed on site by China Media Group

Debut

From August 22 to 26, the Second World Humanoid Robot Games took place at the National Speed Skating Oval, known as the “Ice Ribbon”, in Beijing. It was the world’s first comprehensive competition whose main competitors were humanoid robots. This year, it attracted 666 teams and 2,056 robots from 16 countries across six continents to compete together. Leading companies, such as Unitree Robotics, Xiaomi, and the Beijing Innovation Center of Humanoid Robotics, as well as universities like Tsinghua University, Peking University, and the University of Hong Kong, were all present.

CUEB sent two student teams to the competition: the CUEB Zhizao (Smart Makers) Team, competing in the Tai Chi event, and the Chongde (Upholding Virtue) AI Team, competing in the DanceSport cheerleading event.

During the preparation stage, an open selection across the university was held and two cross-school teams were formed. All 11 members from 2 teams were 2025 undergraduates from several academic units, including Zhuoyu Academy, the School of Artificial Intelligence, andthe School of Statistics and Data Science. They were coached and led throughout by Associate Professor Zhou Zhenkun of the School of Artificial Intelligence.

“Baby” Robot Plays Tai Chi

At the stadium, making its debut, the CUEB Zhizao Team brought the smallest competitor at the event - the robot Xiao Qing.

Xiao Qing stands about 65 centimeters tall and weighs about 11 kilograms. Most competitors in this eventare medium-sized or large-sized models over 1 meter tall.The flagship models, like Unitree H1 and Embodied Tien Kung Ultra, are large-sized models. Standing amongcompetitors over one meter tall,Xiao Qing stood in striking contrast.Many people at the venue exclaimed, “So cute!”

According to Zhou Zhenkun, Xiao Qing is built on the Mini Pi commercial full-body robot from High Torque Robotics, which rolled off the production line four months earlier. The team spent two months completing secondary development and algorithm training on the platform. The team members independently built a complete technical pipeline from human motion to robotic reproduction, covering five stages: pose extraction, kinematic retargeting, reinforcement learning training, simulation validation, and real-robot deployment. This pipeline is not limited to a single type of motion; it can be flexibly transferred and adapted to various robot motion learning tasks, demonstrating strong generality and extensibility.

Xiaoqing’s outfit is also eye-catching.

Since there was no appropriate ready-made costume on the market, the team custom-made a Tai Chi suit for Xiao Qing. Using highly elastic and wear-resistant fabric, the team repeatedly revised the pattern according to the robot’s body proportions and range of motion, so that the suit fits its body without restricting its joint movement. Against the industrial-style setting of the competition venue, Xiao Qing in its little Tai Chi suit looked especially adorable.

The combination of a group of freshmen and the event’s smallest robotattracted the attention of a reporter from China Media Group. On August 24, a related special report aired on CCTV-5’s Sports Morning News.

Wang Zihao, captain of the CUEB Zhizao Team, was a 2025 undergraduate from Zhuoyu Academy. In the interview, he said that although the competition and preparation were intense, the process gave him a more intuitive understanding of the theory of algorithms studied in class and made the team more confidentand assured. Chen Junren, a team member and 2025 undergraduate from the School of Artificial Intelligence,noted that thanks to CUEB’s robust academic platform, participating in a world-class robotics competition as a freshman offered a rare opportunity to gain hands-on experience and foster personal growth.

Building from Scratch

Preparations for the competition began in this late June, just two months before the event. Starting from scratch, the team completed the entire process—from algorithm development and simulation training—to motion debugging and deployment on physical robots in just two months. During the preparation period, the School of Artificial Intelligence also organized special training in reinforcement learning for embodied intelligence at Suzhou Gaoqing Mechanical and Electrical Co., Ltd.., bridging the gap between theory and practice.

“The training moved us beyond the limitations of theoretical learning and put professional knowledge, including sensor data fusion and robot motion control, into practice. It also gave us a much clearer understanding of how to prepare for the competition and how to structure our training.” said Wang Weiyong, captain of the Chongde AI Team and a 2025 undergraduate from the School of Artificial Intelligence.

Zhou Zhenkun leads the team on a study visit to the training base.

For humanoid robot tasks that requireconstant trial and error, a two-month timeframe is an extremely tight timeframe. “These students had just completed their firstyear of university. They built the entire technical pipeline from scratch, so the fact that “Xiao Qing” made it to the competition arena was anachievement in its own right,” Zhou Zhenkun noted.

On the evening of August 23, the CUEB Intelligent Manufacturing Team brought “Xiao Qing” to the Tai Chi event. Accompanied by music, “Xiao Qing” performed the movements one after another, including Opening Form, Cloud Hands and Parting the Wild Horse’s Mane. However, the competition floor offeredless friction than the training environment. Once stepping onto the arena, “Xiao Qing”began to slip and its movements became unstable. Given the tight competition schedule, there was no time to readjust parameters, so the team opted to let the robot complete the full set of movements. On the evening of August 24, five “Xiao Qing” robots from the Chongde AI Team completed the cheerleading event without asingle error.

The experience was not without its challenges. Yet, as Zhou pointed out, simply making it onto the competition arena counts as an achievement. Over the two-month period, team members ran algorithms on physical robots for the first time and gained first-hand experience of the gaps between simulation environments and real-world competition arenas. At the event, they closely observed technical solutions from top-tier robots in the industry and developed a more intuitive and comprehensive understanding of the current status and technical trends in the humanoid robot sector.

Promoting Learning Through Competitions and IndustryEducation Collaboration

In recent years, CUEB has accelerated efforts to build an industry-education integration platform centered on artificialintelligence.The university is promoting the integration of artificial intelligence with applications in economics and finance, with a focus on cultivating innovative and interdisciplinary personnel equipped with both AI technical capabilities and financial or economic literacy. Participation in this competition represents the university’s practical effort to align with Beijing’s development strategy for future industries, as well as the concrete measure to deepen its talent‑training model of “learning through competitions and industry education collaboration”.

For CUEB, Xiao Qing’s appearance on the global competition stage was far more than a mere contest entry. It was also a practical test of the university’s efforts to innovate its talent-development model in response to the era of intelligent economy.

Starting with Xiao Qing, the university is setting its sights on broader competition arenas with a brighter future.