Box, Run, Crash: Chinese Humanoid Robot Games Highlight Advances and Challenges

A swift left hook, a front kick to the chest, a series of cross jabs, and the crowd erupts in cheers. However, it isn’t traditional kickboxing skills that determine the outcome of the match; instead, an attempted roundhouse kick goes awry, leading to the kickboxer from a prestigious university team tumbling to the floor.

While conventional kickboxing involves risks like bloodshed, sweat, and severe head injuries, competitors in Friday’s match at Beijing’s inaugural world humanoid robot game encountered a unique set of challenges, including balance, battery life, and a deeper philosophical purpose.

The compact humanoid robot named Kickboxer, entered by a team from a leading Chinese University of Technology, is part of the Jamboree—a humanoid event showcased at China’s latest tech gatherings. This government-backed competition kicked off after an audience of 12,000 national speed skaters, training for the 2022 Winter Olympics, performed to the Chinese national anthem on Friday morning.

“I came here out of curiosity,” remarked Hong Yun, a 58-year-old retired engineer seated in the front row. He mentioned that watching a robot compete was “far more thrilling than seeing real humans doing the same.”

The robot is set to compete in five soccer matches on the event’s first day in Beijing. Photo: Tingshu Wang/Reuters

The event showcases China’s proficiency in humanoid robotics, a sector prominently featured within the country’s artificial intelligence landscape. The promotional efforts are in full swing.

Similar to kickboxing, humanoids engaged in various sports, including athletics, soccer, and dance. One robot stumbled during a 1500-meter event, losing its head mid-course. “Maintaining [the head] was our goal,” shared Wang Ziyi, a 19-year-old student from Beijing Union University who was part of the robotics team.

A troupe of humanoid dance robots took to the stage during the 2025 Spring Festival Gala, a televised celebration that captivated nearly 1.7 billion viewers online.

One robot got derailed midway through a 1500m event as its head detached. Photo: Kevin Frayer/Getty Images

These social media-friendly activities reflect more serious geopolitical dynamics, highlighting the intensifying technological rivalry between the US and China, which may reshape the AI landscape.

This technology has become a pivotal factor in relations between the two nations. Despite the US’s continued lead in frontier research, Beijing is heavily investing in practical applications like robotics, partly driven by restrictions on Washington’s advanced chip exports to China.

Several cities, including Beijing and Shanghai, have created 100 billion yuan (around 1 billion pounds) funds for the robotics industry. In January, state-owned banks revealed plans to offer 1 trillion yuan in financial support for the AI sector over the next five years.

“If there’s a sector where [Beijing] has heavily invested, it’s this one,” noted Kyle Chan, a researcher at Princeton University.

The robot is seen being transported after a kickboxing match during the competition’s opening day. Photo: China News Service/Getty Images

There’s something inherently unsettling about witnessing a jerky, human-like robot with two arms and legs, being dragged out of the ring by a human operator.

In the realm of humanoids, the Chinese industry possesses many strengths. While US firms like Tesla and Boston Dynamics remain dominant overall, several Chinese companies—including Ubtech and Unitree Robotics, who provided the boxing robots for Friday’s match—are swiftly catching up.

Tesla relies on China for numerous components needed to produce its physical humanoids. According to investment bank Morgan Stanley, the China-based supply chain is projected to manufacture robots using a third of its non-Chinese suppliers. “It appears remarkably challenging to disentangle this area from China completely,” Sheng Zhong, head of Chinese Industrial Research at the bank, noted in a recent report.

The robot, developed by the Chinese firm Unitree Robotics, is seen playing traditional drums. Photo: Tingshu Wang/Reuters

Beyond just generating positive attention on social media, China envisions humanoids as potential solutions to challenges posed by its aging population and shrinking workforce. A recent article from the People’s Daily, the mouthpiece of the Chinese Communist Party, suggested that robots could provide both practical and emotional support to the elderly. “The vision for robot-assisted elderly care is not far off,” it asserted. Humanoid robots could also replace factory workers as China seeks to retrain its workforce for more advanced technological roles.

However, there remains a significant gap between humanoids that can stumble through a sports match and those capable of managing everyday tasks. Ensuring safe interactions with vulnerable populations represents another considerable challenge. “The home is likely one of the last environments where humanoid robots will be welcomed for safety reasons,” Chan stated. “Overall, I maintain a somewhat skeptical view regarding the humanoid explosion.”

A technician works on humanoid robots in the vicinity of the game. Photo: Kevin Frayer/Getty Images

Two significant obstacles to deploying technology that is useful beyond PR stunts are the complexity of human environments and the dexterity required to navigate them.

While other forms of AI, like large language models, can be trained using vast amounts of digital data, there are far fewer datasets available to train algorithms for walking through crowded restaurants or maneuvering stairs. China’s initiatives to integrate robots into everyday settings might assist businesses in gathering more data, yet that remains a major bottleneck, according to Chang.

Dr. Jonathan Aitken, a robotics lecturer at the University of Sheffield, echoed this sentiment. “The current AI state is not yet prepared for humanoids operating in uncontrolled environments,” he asserted.

While impressive displays, such as a robot jumping or kicking, showcase remarkable capabilities, executing mundane tasks—like using a knife or folding laundry—demands a level of finesse. Human hands possess approximately 27 “degrees of freedom,” enabling independent movements. In contrast, one of the most advanced models available, Tesla’s Optimus Humanoid, has only 22.

Nevertheless, China has defied the odds before with rapid advancements. Just a decade ago, the nation exported fewer than 375,000 cars annually. Today, China stands as the world’s largest automotive supplier, shipping nearly 6 million vehicles each year. In response, the European Union has raised tariffs on electric vehicles produced in China to curb this trend.

In China, both the government and the populace are firmly behind the push for humanoids. Zhan Guangtao attended the Humanoid Games alongside her two daughters on Friday. “It’s essential to expose my kids to advanced robotics from around the world,” Zhan remarked. “Such exposure broadens their perspectives.”

Additional research by Lilian Yang

Source: www.theguardian.com

Amazon Tests Humanoid Robots for Package Delivery

Amazon is said to be working on software for humanoid robots intended to serve as delivery personnel and operate “bomber” vans.

The £1.47 trillion technology giant is establishing a “humanoid park” in the US to trial these robots, with insights being shared with tech news outlets, according to individuals involved in the initiative.

Reportedly, the robot could ultimately replace delivery workers. It is designed with artificial intelligence software that powers the robots but utilizes hardware sourced from other companies.

According to reports, the indoor obstacle course at Amazon’s San Francisco offices is about the size of a coffee shop, with hopes that the robots can navigate Amazon’s Libyan vans for deliveries.

Even if a human is operating the vehicle, the robot could potentially expedite drop-off times by managing deliveries to one address while human staff service another. Amazon is also exploring self-driving vehicles through its Zoox division.

As reported, Amazon has over 20,000 Libyan vans in the US, including one stationed in the Humanoid Test Zone. Following successful tests at Humanoid Park, the robots are slated for a real-world “field trip” to attempt package deliveries to customers’ residences.

Amazon has already initiated trials with humanoid robots and is utilizing devices developed by US-based Agility Robotics within its warehouses. Agility CEO Peggy Johnson mentioned to the Guardian last year that their humanoid robots are enabling employees to transition into “robot manager” roles.

Last year, Amazon obtained permission to test drone flights beyond the line of sight of British human controllers, furthering its innovations in home delivery technology.

Professor Subramanian Ramamoorthy, head of robot learning and autonomy at the University of Edinburgh, acknowledged that Amazon has a respected robotics team, emphasizing that their focus on “last-mile” delivery is expected. He noted that humanoid robot hardware is advanced enough for these tasks, but highlighted the challenge of achieving reliable performance outside controlled environments like the proposed “humanoid park.”

Skip past newsletter promotions

He remarked: “If Amazon restricts the environment, utilizing relatively clear driveways and standard doorway layouts, the task becomes fairly straightforward. Challenges arise as the environment becomes more diverse, particularly with variables like pets and small children entering the equation.”

Amazon has been approached for a comment.

Source: www.theguardian.com

AI-Powered Humanoid Workers and Surveillance Buggy: The Impact on Daily Life in China

oOn Saturday afternoon in Central Park in Shenzhen, a teenage girl gag evacuates from a drizzle under a concrete canopy. Putting stacked bags of potato chips in front of them, they swarm around some smartphones and sing towards the Mandoppo ballad. Their laughter rang across the grass around them, until they drilled holes in a mechanical, lively sound. Someone ordered dinner.

A few meters away from the improvised karaoke session is the “Airdrop Cabinet.” This is one of over 40 things in Deep Shenzhen, run by Meituan, China’s largest food delivery platform. The Hungry Park offers everything you can order, from rice noodles to subway sandwiches and bubble tea.

Loaded with items from a shopping mall less than 3km, drones watch, listen, listen, hover over the delivery station before lowering and depositing the items in a sealed box that can only be unlocked by entering the customer’s phone number. Dinner is served with non-humans. Meituan aims to beat human delivery times by about 10%, perhaps for a journey through the clouds in a thin polystyrene box.

The drone will take off from the rooftop of a shopping mall in Shenzhen, China on April 3, 2025. Photo: Anthony Kwan/The Guardian

Drones are just a part of the broader robotics and artificial intelligence industry that China intends to expand this year.

With the trade war furious, demographic challenges are dragged into the economy, and the prospects for productive relationships with the world’s largest economy seem farther than ever. Chinese leaders see artificial intelligence as key to solving problems created by the shrinking workforce, upgrading its military power, and solving the source of public pride, especially if Chinese companies avoid US-led sanctions on core technologies. And as technology companies have tried to crack down on excessive wealth and influence outside of the control of the Xi Jinping state, which has been shunned by Chinese leaders for many years, Xi’s neighbors are welcomed by folds as they seek to restore confidence in the private sector and encourage domestic innovation.

In March, Prime Minister Li Qiang promised to “unleash the creativity of the digital economy” with a special focus on “embodied AI.” Guangzhou, including the deep Shenzhen high-tech hub, is at the forefront of this movement. The state government recently announced 60m yuan (£6.4m) in new funding for the innovation centre. In particular, Demi Shenzhen is known as China’s drone capital due to its progressive approach to drone regulation, allowing the “low-altitude economy” to develop faster than the rest of the country. China’s Civil Aviation Authority predicts the sector’s value will increase by 3.5TN Yuan by five times over the next decade.

Drones aren’t the only ones who promise or threaten the tempo of Chinese urban life. Humanoid robots are particularly lively. The highlights of this spring festival gala have been seen almost 1.7 billion times, A dance performed by a group of humanoid robots It was created by a company called Unitree. On Saturday, the world’s first humanoid vs humanity – half marathon – took place in the suburbs of Beijing.

The robots will be participating in Saturday’s race. Photo: Ng Han Guan/AP

Rui Ma, a Chinese technology analyst and investor based in San Francisco, said: This shift will enable industry growth In 2025, it’s much faster than in the past few years. Reinforcement learning means training robots to learn from experience rather than relying on hard models, training humanoid robots in months rather than years, speeding up the pace of innovation. Toy robot dogs are already part of everyday life in China. At Yiwu’s wholesale market, a trade hub in Zhijiang province in eastern China, mothers stay with exporters beyond the price of eyelashes while children play with robotic dogs. In Streets in Shanghaiwoman walking robot dog. This carries a shopping basket on its back.

The drone run by Meituan, which has been loaded with products, will take off from the rooftop shopping at Shenzhen, China on April 3, 2025. Photo: Anthony Kwan/The Guardian

The development of China’s robotics industry is closely linked to advances in AI. For years, China has been catching up to the US. XI wants to promote economic growth through “new quality productivity” that includes advanced technology.

Many in Washington fear that the US lead is narrowing. One of the main tools in the US arsenal controls a critical part of the supply chain of semiconductors, the microchips used to train advanced AI models. The US has restricted exports of its most sophisticated chips to China. This is part of a strategy that former national security adviser Jake Sullivan described as “highfence” with the most strategically valuable technologies in the United States.

However, in January, a previously unknown Chinese company called Deepseek sets the Chinese technological scene and releases the R1, a massive linguistic inference model, to perform at a price that leads its US competitors. The model wiped out 1TN from Wall Street’s main technology index, causing a stock market crash as investors feared that US pole positions in high-tech races were no longer guaranteed.

“You can’t stress too much how crazy it is,” says MA.

Since then, China’s AI industry has been filled with optimism. As an answer to China’s long-term and sustainable growth, it was already being promoted by the government, and now the people are beginning to believe it, says Ma.

Meituan drones remove orders at the “Airdrop Cabinet” in Shenzhen, China. Photo: Meituan

Li Shuhao, a Guangzhou-based high-tech entrepreneur who founded AI marketing company TEC-DO in 2017, was in the US when the Deepseek moment happened. Suddenly, he says, “It was much easier to arrange an interview and a meeting with other AI scientists.”

“Deepseek is like a symbol of the oriental way of doing business,” says Li, a confessed “metal head” surrounded by electric guitars and drum kits in her Guangzhou office. He mentions a strategy by Deepseek founder Liang Wenfeng, who will fund it through his own hedge funds rather than seeking external venture capital funds. “This is how a typical Chinese entrepreneur thinks: survive first and then do something new.”

Deepseek has published the work as open source. This is a principle that the government has long supported and a move that encouraged the widespread adoption of the model. Robotics is a special beneficiary.

Technology is the top priority

The robot supply chain can be roughly divided into three areas: brain, body, and application of technology in the real world. China has long been confident in its capabilities in the latter two regions. The advanced supply chains of other high-tech industries, such as electric vehicles and autonomous drones, show that China has both the ability to produce large-scale industrial components and the ability to assemble them into complex commodities. However, once you solved the most difficult part of the puzzle, it was elusive to create a robotic brain that could learn human-like behavior and movements. You need sophisticated AI.

Deepseek’s R1 model is changing the game and hosting ways for domestic humanoid robot companies to keep up with their international competitors, Goldman Sachs analysts said in a recent memo. The fact that Deepseek’s open source model uses less advanced chips can help level the playing field for Chinese companies.

Engineers will train humanoid robots at the Humanoid Robot Innovation Centre in Shogan Park, Beijing, China on March 28, 2025. Photo: Beijing Youth Daily/VCG/Getty Images

The industry still has its challenges. AI models require a large amount of data to train. While LLM, used for things like chatbots, can draw out a vast universe of content, the Internet, the data in robotic AI models is relatively scarce info on how to physically move spaces and interact with objects and people.

Another sector in which China is focused, the car should be able to navigate six axes, or “degrees of freedom”, forward and rear, left and right, up and down, and rotations between these positions. The same goes for general robotics, such as Meituan’s food delivery loan. To enable humanoid robots to mimic humans on everyday tasks such as cooking, they need up to 60 degrees of freedom. There is a 27 H1 model of Unitree that caused a splash on Spring Gala.

A robot does not have to be completely humanoids to be useful. A wheel or humanoid robot with limited movement can take on automated ripe tasks, such as dangerous or repetitive factory work. Based in Shenzhen, Ubtech has already supplying humanoid robots to its car factories. With a shrinking workforce, China is keen to find ways to automate as much as possible.

The organizers, which was the Boao Forum for Asia last month, business meeting, were keen to cook Jianbing, Delicious pancakes are typical Chinese street food made with the robot arms of the booth, similar to the claws that collect toys packed in an arcade (the resulting snacks weren’t as crisp as humans made). Beijing’s parks have increased their surveillance capabilities by pasting cameras into autonomous buggies running along the path.

The humanoid robot will perform at the opening ceremony of the 2025 Zhongguancun Forum (ZGC Forum) Annual Meeting held in Beijing, China’s capital, on March 27, 2025. Photo: Xinhua/Rex/Shutterstock

“Robots

Source: www.theguardian.com

Observing a humanoid robot driving a car at a very slow pace

Humanoid robots that can drive cars may one day be used as chauffeurs, but their creators acknowledge that this could be at least 50 years away.

Most driverless cars work completely differently than a human driver, using artificial intelligence and custom mechanical systems to directly control the steering wheel and pedals. This approach is much more efficient and simpler than using a humanoid robot to drive, but it needs to be customized for each specific car.

Kento Kawarazuka Professor Takeru Sato of the University of Tokyo and his team have developed a humanoid robot called “Musashi” that can drive a car just like a human. Musashi has a human-like “skeleton” and “muscles,” and is equipped with cameras in both eyes and force sensors in its limbs. An artificial intelligence system determines the movements required to drive the car and responds to events such as changes in the color of traffic lights and people cutting in front of the car.

Currently, robots can only perform a limited range of driving tasks, such as going straight or turning right, at speeds of around 5km per hour on non-public roads. “The pedal speed and car speed are not high, and the car handling is also not as fast as a human,” Kawarazuka said.

Musashi is a humanoid robot that operates cars just like a human would.

Kento Kawarazuka et al. 2024

But Kawarazuka hopes that as the system is improved it could be used in any car, which could be useful when humanoid robots are routinely produced. “I’m not looking 10 or 20 years out, I’m looking 50 or 100 years out,” he says.

“This research could be of interest to people developing humanoid robots, but it doesn’t tell us much about autonomous driving.” Jack Stilgoe “Self-driving cars cannot and should not drive like humans. Because the technology doesn’t need to rely on limbs and eyes, it can rely on digital maps and dedicated infrastructure to find safer, more convenient ways to navigate the world,” say researchers at University College London.

topic:

Source: www.newscientist.com

Large-scale language models are used by Agility to communicate with humanoid robots

I’ve spent most of the past year discussing generative AI and large-scale language models with robotics experts. It is becoming increasingly clear that this type of technology is poised to revolutionize the way robots communicate, learn, look, and program.

Therefore, many leading universities, research institutes, and companies are exploring the best ways to leverage these artificial intelligence platforms. Agility, a well-funded Oregon-based startup, has been experimenting with the technology for some time with its bipedal robot Digit.

Today, the company is showcasing some of its accomplishments in a short video shared across its social channels.

“[W]We were curious to see what we could accomplish by integrating this technology into Digit,” the company said. “The physical embodiment of artificial intelligence created a demonstration space with a series of numbered towers of several heights and three boxes with multiple features. Digit has We were given information about the environment, but we were not given any specific information about the task, just to see if we could execute natural language commands of varying complexity.”

In the video example, Digit is instructed to pick up a box colored “Darth Vader’s Lightsaber” and move it to the tallest tower. As you might expect from early demos, the process is not instantaneous, but rather slow and methodical. However, the robot performs the task as described.

Agility says: “Our innovation team developed this interactive demo to show how LLM can make robots more versatile and faster to deploy. In this demo, people can use natural language to communicate with Digit. You can talk to it and ask it to perform tasks, giving you a glimpse into the future.”


Want the top robotics news in your inbox every week? Sign up for Actuator here.


Natural language communication is an important potential application of this technology, along with the ability to program systems through low-code and no-code technologies.

On my Disrupt panel, Gill Pratt explained how Toyota Research Institute is using generative AI to accelerate robot learning.

We figured out how to do something. It uses the latest generative AI techniques that allow humans to demonstrate both position and force, essentially teaching the robot from just a handful of examples. The code hasn’t changed at all. What is this based on? There is a popularization policy. This is a study we conducted in collaboration with Columbia and MIT. We have taught 60 different skills so far.

MIT CSAIL’s Daniela Russ also told me recently: “Generative AI turns out to be very powerful in solving even motion planning problems. It provides much faster solutions and more fluid and human-like control solutions than using model prediction solutions. I think this is very powerful because the robots of the future will be much less robotic. Their movements will be more fluid and human-like.”

The potential applications here are wide and exciting. And Digit, as an advanced commercial robotic system being piloted in Amazon fulfillment centers and other real-world locations, seems like a prime candidate. If robots are to work alongside humans, they will also need to learn to listen to us.

Source: techcrunch.com