A Group of Tsinghua Alumni and Four Decades of Embodied Intelligence
In June this year, the news of Galaxy Universal's 1.1 billion yuan in financing shocked the venture capital circle. This single largest annual financing in China's embodied intelligence track has enabled the company, which was founded just two years ago, to attract more than 2.4 billion yuan in cumulative investment, making it a unicorn valued at $1 billion.
When Wang He brought Galaxy Universal into the spotlight, people suddenly realized that a group of entrepreneurs with the imprint of Tsinghua University are making a collective breakthrough in the field of embodied intelligence.
Stellarium, founded by a post - 90s Tsinghua team, secured tens of millions of dollars in angel financing in just four months. Lu Ming Robotics, led by Yu Chao, completed three rounds of financing within half a year. Cheng Hao's Acceleration Evolution has already obtained commercial orders. On the business cards of the leaders of these companies, there is almost always a "footnote" of Tsinghua University.
As of 2025, nearly 20 embodied intelligence enterprises founded by Tsinghua alumni or with core teams from Tsinghua University have emerged. The cumulative financing scale has exceeded 5 billion yuan. These enterprises have formed differentiated competitiveness in multiple dimensions such as hardware R & D, algorithm breakthroughs, and scenario implementation, becoming an important force in China's embodied intelligence field.
This is not a coincidence. As early as 1988, the State Key Laboratory of Intelligent Technology and Systems at Tsinghua University launched the research on the THMR series of robots, planting the first seed for China's embodied intelligence. Today, the soil nurtured by years of technological accumulation is giving rise to the most vibrant entrepreneurial ecosystem for embodied intelligence in China.
From being the darlings of capital with tens of millions in financing to becoming industry pioneers with multi - scenario implementation, why have Tsinghua - affiliated embodied intelligence companies become the focus of both capital and the industry? How have these entrepreneurs, bearing the school motto of "Self - discipline and Social Commitment", led their enterprises to break through the encirclement?
01
Forty Years: Planting the First Seed of Embodied Intelligence
2025 is regarded as the first year of the explosion of the embodied intelligence industry.
The successive hosting of the World Artificial Intelligence Conference, the World Robot Conference, and the Humanoid Robot Games has pushed the popularity of this field to its peak. The industry is presenting a prosperous scene of explosive growth. Renowned investment institution Goldman Sachs predicts that by 2035, the global market size of humanoid robots alone will reach $38 billion.
In this global upsurge, few people notice that Tsinghua University's exploration in the field of embodied intelligence has already spanned forty years. This long journey from the laboratory to the forefront of the industry is also a microcosm of China's embodied intelligence research evolving from following to running side - by - side with the world.
The progress of this journey cannot be separated from the continuous leadership of Zhang Bo.
As an academician of the Chinese Academy of Sciences and the honorary dean of the Institute for Artificial Intelligence at Tsinghua University, Zhang Bo is known as the grandmaster in China's artificial intelligence field. He devoted himself to the field of artificial intelligence more than forty years ago and pioneered many firsts in China: publishing the first academic paper on artificial intelligence in China, winning the first important international award in the field of artificial intelligence in China, leading the establishment of the first intelligent robot laboratory in China, cultivating the first domestic - born doctor in artificial intelligence, and organizing the first national key laboratory for artificial intelligence in China.
This path of exploration was not easy. Looking back over the forty - year timeline, we can see that Zhang Bo's every step was taken at a crucial development node. His thinking, changes in direction, and perseverance deeply reflect "half" of the history of artificial intelligence in China.
In 1978, the Department of Automatic Control at Tsinghua University was renamed the "Department of Computer Technology and Applications". Teachers who were originally engaged in automatic control research in the department faced two choices: staying in the computer department or moving to the newly established automation department. Zhang Bo, who was 43 years old at that time, resolutely chose to stay in this nascent discipline and devote himself to the unknown field of artificial intelligence.
This was an uncharted path. Zhang Bo, together with other senior teachers at Tsinghua University such as Zhang Yukai, Lin Yaorui, Shi Chunyi, and Huang Changning, designated "Artificial Intelligence and Intelligent Control" as a new teaching and research direction, which also propelled Tsinghua to become one of the earliest institutions in China to systematically conduct teaching and research in artificial intelligence.
In 1984, Zhang Bo and his younger brother Zhang Ling, who was teaching at Anhui University, jointly published a paper in the top international journal in the field of artificial intelligence, "IEEE Transactions on Pattern Analysis and Machine Intelligence", which attracted high attention from international peers. This was also the first academic paper on artificial intelligence by Chinese scientists.
The SA algorithm they proposed won the ICL European Artificial Intelligence Award at the European Conference on Artificial Intelligence in 1984, making them the first Chinese to receive an important international award in this field. In the same year, Zhang Bo returned to Tsinghua University to continue teaching after completing his overseas study with fruitful research results.
From 1982 to 1984, the teaching and research group of Artificial Intelligence and Intelligent Control at Tsinghua University, where Zhang Bo was located, had been conducting investigations and research, visiting a large number of research institutes and factories in the southwest, northeast, and other regions. Through extensive research, they realized that robots would play a significant role in the industrial sector and become a major demand. To seize this opportunity, Zhang Bo and his colleagues began to run around in various ways, raise funds, and contact domestic and foreign manufacturers to prepare for the establishment of the Intelligent Robot Laboratory at Tsinghua University.
As the saying goes, "Those who keep their wishes in mind will surely receive a response." In 1985, their wish came true. They purchased the first imported robotic arm, laying a hardware foundation for the research of embodied intelligence. In the same year, Tsinghua University established the first intelligent robot laboratory in China, and Zhang Bo became the head of the laboratory.
The laboratory purchased the first PUMA560 robot in China. Zhang Bo set the research direction as the application of robots in scenarios such as mines and industrial production. At the same time, research was also carried out in directions such as intelligent vehicles, Chinese character recognition, voice recognition, and natural language understanding.
They always caught up with the trends of the times. In 1986, the national "863 Program" was launched, and intelligent robots were listed as one of the key themes.
At that time, artificial intelligence was still in a trough, and there were few relevant studies at home and abroad. The Tsinghua team led by Zhang Bo once again keenly captured this opportunity. In 1987, he took the lead in preparing for the establishment of the State Key Laboratory of Intelligent Technology and Systems.
In 1988, the State Key Laboratory of Intelligent Technology and Systems launched the development of the THMR series of mobile robot systems. Although this series was mainly aimed at unmanned vehicle systems and did not have an "anthropomorphic" appearance, it accumulated underlying capabilities for mobile robot technology.
The laboratory officially opened for operation in 1990, and it was also the first national key laboratory for artificial intelligence in the country. The laboratory has successively undertaken a series of projects such as the National Natural Science Foundation, the National Climbing Program, the "863 Program", and the National Science and Technology Key Project Program, becoming one of the main undertakers of national scientific research tasks in artificial intelligence.
When talking about the State Key Laboratory of Intelligent Technology and Systems, Zhang Bo is filled with pride. The laboratory has won the title of "Excellent" three times in national evaluations. In those days when everything was in its infancy, the perseverance of a group of people was gradually recognized.
The real turning point occurred in 1998. With the special funds from the first phase of the "985 Project", Tsinghua University established a major research project on robots, and the systematic development of humanoid robots was officially put on the agenda.
Four years later, Tsinghua University successfully launched China's first autonomous robot system with a complete humanoid form - THBIP - I. This robot was jointly developed by the Department of Precision Instruments, the Department of Mechanical Engineering, and the Department of Automation. It is 172 cm tall, weighs 130 kg, and is equipped with 32 degrees of freedom throughout the body. The head, arms, torso, and legs are all modularly designed, marking a key step for China in the structural design and motion control of humanoid robots.
From the motion planning of robotic arms to the environmental perception of intelligent agents, Zhang Bo has led the Tsinghua team to delve deeply into the field of embodied intelligence for more than forty years. From theoretical construction, technological breakthroughs to platform building, they have laid a solid foundation for the development of embodied intelligence at Tsinghua University.
Today, the seeds sown by this pioneer have grown vigorously. A group of younger scholars and industry pioneers like Zhu Jun, Ma Shaoping, and Yuan Jinhui are taking over the baton and embarking on a new journey in embodied intelligence on the solid foundation laid by their predecessors.
02
Twenty Years: One Person, One Team
On another timeline, the inherited spiritual gene has quietly taken root, giving rise to a different scientific research community.
"We won!" The jubilant cheers echoed across the football field. In July this year, the Huoshen Team from Tsinghua University gloriously won the championship in the humanoid robot category of the RoboCup World Cup in Brazil. This was also the first time a Chinese team had won the championship in this category.
The members of the Huoshen Team are from the Robot Control Laboratory at Tsinghua University. Since they first participated in the World Cup in 2005, the students participating in the competition have changed generation after generation, but the leading teacher has always been Zhao Mingguo, a researcher at the Department of Automation of Tsinghua University and the director of the Robot Control Laboratory. He is also the founder of the Huoshen Team.
In Greek mythology, Hephaestus, the god of fire, not only rules over fire and forging skills but is also good at making tools and weapons. Zhao Mingguo thought this was highly consistent with their work of building robots, so the name "Huoshen Team" was born and has been used ever since.
When it comes to the Huoshen Team and Zhao Mingguo, the main thread running through is the twenty - year - long exploration and perseverance in the field of robots. Looking back at Zhao Mingguo's scientific research trajectory, it almost coincides with the growth curve of humanoid robots in China.
He began to pay attention to this field after Honda released the ASIMO in 1997. In 2000, he entered Tsinghua University for post - doctoral research, studying under Professor Chen Ken and focusing on the control system of humanoid robots. In an era when industrial robots still relied on imports and domestic research was scarce, he resolutely chose the difficult research direction of humanoid robot control systems.
At that time, there was a team from the Department of Computer Science at Tsinghua University participating in the simulation group of the "Robot World Cup" RoboCup. Since Zhao Mingguo's research direction was in line with theirs, he founded the Huoshen Team at Tsinghua University, specifically for humanoid robot competitions.
At that time, robotics was not yet a mainstream research direction at Tsinghua University. Due to high investment and low output, compared with information science competitions such as ACM, the positive feedback came much more slowly. So the establishment of the Huoshen Team faced many difficulties. Zhao Mingguo could only select a few students from the robot laboratory, and some other teachers from the Department of Automation also sent a few postgraduate students. Thus, a robot football team took shape.
After 2003, due to the adjustment of Tsinghua University's research direction, the research on humanoid robots was temporarily put on hold.
In 2004 and 2005, the Huoshen Team participated in domestic and international competitions for the first time respectively. In the first two years of competition, Zhao Mingguo focused on the research of the standard platform based on the AIBO robot dog. It was not until the international competition in 2006 that, seeing the team's mediocre performance, Zhao Mingguo reflected and believed that the humanoid form was more in line with the team's characteristics. So after returning to China, he decided to switch from quadruped to humanoid robots.
At that time, the international robot development also reached a climax. The year 2005 was a node with both technological explosions and theoretical breakthroughs. A team from Cornell University published a paper on passive walking in "Science", Honda's ASIMO achieved fast movement, and Boston Dynamics launched the Big Dog quadruped robot. A series of developments not only brought inspiration but also huge competitive pressure.
At the beginning of 2006, Zhao Mingguo began to study passive walking technology, and the Huoshen Team also started to engage in independent hardware development in addition to software. At that time, Zhao Mingguo realized that the determining factor for the strength of humanoid robots lies in mechanics and motion control.
The road to success was not smooth. In 2007, the Huoshen Team participated in the competition again, but the result was still not satisfactory, failing to make it into the top eight. Zhao Mingguo began to adjust the strategy. He found that software research could not be separated from hardware, so he started to pay attention to robot hardware and led the team to build several humanoid robots of different sizes.
In 2008, Zhao Mingguo led the team to develop a "virtual slope method" specifically for humanoid robot walking. This technology enabled the Huoshen Team to shine in the RoboCup competition of the same year, winning the first medal for the Chinese team in the humanoid category. The team members' morale was greatly boosted.
But around 2010, due to technological bottlenecks and unclear business prospects, many teams around the world withdrew from this field. In 2011, the theoretical innovation in the field of robotics slowed down, and both basic research and application development encountered bottlenecks. Looking back on those difficult years, Zhao Mingguo once admitted, "After twenty years, I'm the only one still insisting."
The turning point came in 2013 and 2015. The consecutive hosting of the DARPA Robotics Challenge reignited the enthusiasm of the academic and industrial circles for humanoid robots. The competition proved in practice that optimization methods such as overall behavior control and model predictive control can effectively meet the actual application needs in complex environments. These methods can not only effectively solve practical problems but also have a more complete theoretical framework.
With the development of technological trends, Zhao Mingguo also began to study optimization control methods in 2015 and planned to apply them to the development of a high - performance robot cluster force control