Chinese humanoid robots get reality check in half-marathon debut
Six out of 21 Chinese robots completed the world's first half marathon (21.1 kilometres) for humanoid robots in Beijing on April 19, with the winner setting a record time of 2 hours and 40 minutes.
Over 12,000 people participated in the marathon on the same day, and the human winner reached the endpoint in 1 hour and 2 minutes.
The Tien Kung Ultra robot, developed by X-Humanoid, a Beijing-based state-owned enterprise, won the race for humanoid robots. It is 1.8 meters tall and weighs 52 kilograms.
The first runner-up, Noetix Robotics' N2, finished the half-marathon in 3 hours 37 minutes. It is 1.2 meters tall, weighs 29 kilograms, and wears children's trainers.
The second runner-up, DroidUp's X02, finished the race in 4 hours and 50 minutes. It wears boxing gloves.
An unofficial team entered the competition using a Unitree G2 robot, but the robot fell at the starting point and became a talking point of the event.
Unitree said that the company did not send a team to participate in this marathon as it has been busy recently. It added that anyone who bought a Unitree robot can deploy it in activities, but different training will lead to different results.
On April 10, Unitree released footage saying it will be live-streaming robot combat in about a month.
Tang Jian, chief technology officer of X-Humanoid, said his team only used one robot to complete the half-marathon, while other players needed to change robots several times.
He admitted that the X-Humanoid robot had fallen once during the marathon due to battery failure and had three battery swaps during the race.
'We have spent seven to eight months preparing for this marathon by doing a lot of tests on the robot and fine-tuning its algorithm,' Tang said.
'Firstly, we reduced our robot's weight; secondly, we improved its heat dissipation system to avoid overheating; Thirdly, we added a cushioning function to minimize wear on its sole and joint modules,' he added. 'What is more important is that we increased the length of our robot's legs to run faster.'
'The commercialization of humanoid robots will have three phases. In the first phase, robots will be used in general industries; then, they will be used in specific commercial environments to finish tasks such as packaging and scanning codes; finally, they will be deployed to homes to play the role of nanny and housekeeper,' Tang said.
He said it would take Chinese companies three to five years to commercialize their robots, by which time they would be deployed in homes to serve people and do chores.
X-Humanoid, officially known as Beijing Humanoid Robot Innovation Center, was formed jointly by robotics firm Ubtech, electronics and electric vehicle maker Xiaomi and the Beijing municipal government.
Shares of the Hong Kong-listed Ubtech Robotics increased 2.9% on Tuesday. Shares of the state-owned Shoucheng Holdings, which has invested in X-Humanoid and Noetix, rose 4%.
According to the marathon rules, each team can deploy three staff members to assist a robot. The robot does not need an autopilot system; its remote controller can be wired or wireless.
The race terrain includes flat land and ramps, with a maximum slope of 9 degrees. The race route has six left-turn and eight right-turn lanes.
One of the robot developers who joined the marathon said this activity aims to test robots' cerebellums (ability to control the body) instead of their cerebrums (ability to make decisions and think).
Of the 21 robot runners, only six completed the race, while the remaining 15 failed. A small robot went up in smoke during a marathon in Beijing on April 7, 2025. Photo: Baidu
A small robot developed by Beijing Polytechnic University students overheated and went up in smoke. Some other teams sprayed water on their robots to keep them cool.
A female-looking robot walked a short distance and fell. One Gundam (from a Japanese comic) robot used four fans to move forward, but crashed seconds after beginning its journey.
'The winning secret of Noetix's N2 robot is its stable mechanical structure and superior algorithm performance,' Hu Chenxu, founder and chief scientist of Noetix, told the media. 'Small robots have a lot of advantages, such as higher flexibility and agility. Many research institutes use small robots to do their research work.' A female-looking robot has ended her journey after walking for a short distance. Photo: Baidu
'Robots need a high adaptability to go from the laboratory to the marathon track,' he said.
He said he once worried that the N2 robot, which is more like a gymnast, would not be a good runner; fortunately, the robot performed exceptionally well, except it needed a new remote controller after a battery change.
He said Noetix's robots are now used in scientific research, education, exhibitions, cultural, and tourism sectors and will be deployed for scenarios such as elderly care, industrial inspection and urban security.
In September 2023, Jiang Zheyuan, a PhD student at Tsinghua University, then 25, founded Noetix Robotics. So far, the company has raised five rounds of funding. It planned to deliver one thousand robots this year and expected positive cash flow in the fourth quarter.
Jiang told Sohu.com in an interview in March that his company needed to commercialize its products as early as possible due to its limited financial resources, unlike DeepSeek, an artificial intelligence company with sufficient resources to pursue technical ideals.
Regarding the robotic race between China and the United States, Jiang said China has a substantial cost advantage. He said it costs only 40 yuan (US$5.5) per hour to collect robotic data in China but US$48 per hour in the US. Besides, he said a wide range of industries in China can provide different environments for companies to improve the algorithms of their robots.
He said that using Chinese humanoid robots will continue to focus on education and exhibition in the next three years. He said Noetix plans to launch a small robot that costs below 10,000 yuan next year.
According to the Gaogong Robot Industry Research Institute, total sales of humanoid robots will reach 12,400 units or 6.34 billion yuan globally in 2025. Chinese companies may sell about 7,300 humanoid robots during this period.
Read: Unitree's humanoid robot a great side-flip forward for China
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Asia Times
an hour ago
- Asia Times
China cracks a code on invisible battlefield surveillance
This month, the South China Morning Post (SCMP) reported that Chinese radar scientists have demonstrated a breakthrough system enabling aircraft to detect moving ground targets with high clarity while remaining completely radio-silent. Led by Li Zhongyu of the University of Electronic Science and Technology of China, the team used two Cessna 208 planes in formation, with one actively transmitting radar signals and the other passively receiving echoes, to test their 'space-time decoupling two-channel clutter cancellation method.' As reported in the Journal of Radars, the test tracked three vehicles across terrain previously inaccessible to traditional radar due to clutter, range migration and Doppler noise. Through advanced motion correction, spectrum compression and a novel matrix-based clutter suppression technique, the system isolated targets with over 20 decibels more clarity than current technologies. The passive aircraft never transmitted signals, making it nearly impossible to detect or jam, ideal for low-probability-of-intercept (LPI) missions, according to the Journal of Radars report. Instead of relying on artificial intelligence, the method utilizes efficient mathematical models specifically designed for real-world applications. The development could allow Chinese aircraft and missiles to silently scan battlefields, oceans or skies, identifying mobile threats without alerting adversaries. Li called the system a world first, positioning China at the cutting edge of stealth radar capabilities amid growing global electronic warfare competition. Explaining the complementary nature of passive and active radars, Eric Hundman mentions in a March 2025 report for the China Aerospace Studies Institute (CASI) that passive radars can often detect stealth aircraft, as they're hard for anti-radiation weapons to target because they emit no signals. However, Hundman says the applications of passive radar remain relatively limited due to its reliance on unpredictable external signals, narrow usable frequencies and relatively unsophisticated processing algorithms. As for active radars, Hundman says they offer high-resolution, multi-function capabilities, combining stealth detection, surveillance, tracking and fire control in increasingly mobile and networked platforms. However, he points out that since active radars emit signals, they are vulnerable to detection by passive sensors and anti-radiation missiles. In view of each system's limitations, Hundman states that Chinese researchers are working to complement active radars with passive systems and pursue network integration, aiming to fuse data across platforms for a more resilient and comprehensive radar network capable of withstanding electronic and kinetic threats. Discussing the potential operational application of the technology, the US Department of Defense's (DOD) 2024 China Military Power Report (CMPR) states that the People's Liberation Army (PLA) continues to make progress in the development and integration of unmanned systems, including drone swarm tactics and manned-unmanned teaming (MUM-T) concepts. According to the report, these efforts include employing drones for intelligence, surveillance and reconnaissance (ISR), as well as electronic warfare, decoys and potentially precision strike missions in coordination with stealth aircraft. In an April 2025 report for the Center for Strategic and Budgetary Assessments (CSBA), Travis Sharp highlights the strategic pairing of manned stealth fighters and collaborative combat aircraft (CCAs), emphasizing their complementary roles in air dominance. Sharp says CCAs, acting as loyal wingmen, would use active radar to detect and track enemy aircraft, transmitting targeting data to stealth fighters that remain electronically silent to avoid detection. He notes that this setup enhances survivability while extending the fighter's engagement range, allowing long-range missile strikes before the enemy is aware. Sharp adds that CCAs can also jam enemy radars, further complicating adversary targeting efforts. He points out that such tactics are key to improving loss-exchange ratios in potential conflicts, including in the Taiwan Strait between the US and China. The advent of China's purportedly sixth-generation fighters adds impetus to the development of such technology. Malcolm Davis mentions in an SCMP article this month that China's J-36, which appears to be optimized for air-to-air combat with very long-range missiles, could achieve advanced stealth and teaming with collaborative combat aircraft (CCA). Further, Timothy Heath mentions in the same article that sixth-generation fighters are unlikely to be involved in maneuvering dogfights but instead focus on command and control, enabling unmanned aircraft to conduct strikes against air, marine and ground targets. Ditching manned fighters altogether, the SCMP reported in May 2025 that China's Jiu Tan drone carrier, a super-high-altitude, long-range unmanned aerial vehicle (UAV), could carry up to 100 kamikaze drones or six tons of ammunition with a maximum range of 7,000 kilometers and an altitude of 15,000 meters. SCMP notes that, if deployed, the Jiu Tan could contribute to the PLA's swarming capabilities, wherein large groups of drones are released to overwhelm an adversary's defenses. Stacie Pettyjohn and other writers mention in a June 2024 report for the Center for a New American Security (CNAS) that China could utilize its diverse drone fleet, ranging from kamikaze to reconnaissance and loyal wingman-type systems, to saturate air defenses, rapidly close kill chains, and complicate Taiwan and US countermeasures in a conflict over the island. Pettyjohn and others say that China holds an initial advantage due to its large, inexpensive and varied drone arsenal, which could be used to locate and strike US and Taiwanese forces with speed and scale. They also warn that drone mass and rapid adaptation, not just innovation, is critical, citing Ukraine as a cautionary example, with China's production capacity giving it a dangerous edge in the opening phases of a Taiwan war. As for the US's capability to counter drone swarms, Wilson Beaver and Ka'Von Johnson mention in a Heritage Foundation report this month that the US faces growing challenges in countering drone swarms due to fragmented capabilities, underdeveloped training and an overreliance on costly systems. Beaver and Johnson say that while kinetic interceptors and electronic warfare tools offer layered defenses, gaps persist in small-drone detection, real-time threat identification and swarm neutralization. They point out that directed-energy weapons, such as lasers and high-powered microwaves (HPMs), remain impractical due to issues with range, power and target discrimination. In addition, they say most counter-drone systems are not widely fielded, few troops are trained to operate them and base commanders often lack the authority to engage drones, compounding vulnerabilities. While the US races to patch gaps in drone defenses, China is methodically assembling a sensor-strike network designed to operate silently, resiliently and at scale.
South China Morning Post
an hour ago
- South China Morning Post
Chinese academy launches automated system to speed up chip design amid US software curbs
The Chinese Academy of Sciences (CAS), the nation's premier science institution, has unveiled a chip design system driven by artificial intelligence (AI) that has the potential to significantly accelerate semiconductor development and replace human programmers. Developed by the State Key Laboratory of Processor and the Intelligent Software Research Centre, both under CAS, and the University of the Chinese Academy of Sciences, the fully automated integrated circuit design system is called QiMeng, which means 'enlightenment' in Chinese. The system uses large language models – the technology underpinning advanced AI chatbots. Details were published in a research paper last week and open-sourced on GitHub. According to the developers, chips designed by QiMeng match the performance and efficiency of those created by human experts. A customised autonomous-driving chip, for example, which might take weeks for a team of human developers to create could be accomplished by QiMeng in just days, according to the team. An overview of the automated chip design system QiMeng. Photo: Handout The unveiling of QiMeng comes as the US has pressured leading electronic design automation (EDA) software suppliers to halt sales of chip design tools to China , further complicating Beijing's goal to strengthen its semiconductor industry. QiMeng comprises three interconnected layers: at the base is a domain-specific large processor chip model; in the middle, a hardware and software design agent; and at the top, various processor chip design applications.
RTHK
13 hours ago
- RTHK
US stocks end mostly up as markets eye trade talks
US stocks end mostly up as markets eye trade talks US stocks barely moved, with the Dow down 0.15 percent, the S&P 500 up 0.08 percent, and the Nasdaq up 0.30 percent. File photo: AFP Wall Street stocks edged higher on Monday as US and Chinese representatives held high-stakes trade talks and markets looked ahead to key economic data. Trade representatives for the world's two biggest economies plan a second day of talks on Tuesday following an opening round on Monday. Although there were no breakthroughs, the market has welcomed the negotiations. "There's hopes that they're inching closer to some sort of a deal," said Peter Cardillo of Spartan Capital Securities. The Dow Jones Industrial Average finished flat at 42,761. The broad-based S&P 500 climbed 0.1 percent to 6,005, while the tech-rich Nasdaq Composite Index added 0.3 percent at 19,591. Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick and Trade Representative Jamieson Greer are leading the US delegation, while China's team included Commerce Minister Wang Wentao and China International Trade Representative Li Chenggang. Among individual companies, Apple fell 1.2 percent after company executives emphasised its artificial intelligence efforts at the tech giant's annual developers conference, including allowing app makers to directly access a device's AI capabilities. Apple has lagged some rivals in introducing AI-advancing updates to its Siri voice assistant and other programs. Warner Brothers Discovery finished down 3 percent after unveiling a plan to split itself into two companies to better position for the streaming era. The entertainment giant will break itself into two publicly traded companies: one covering "Streaming & Studios" and the other "Global Networks." The shift, designed to enable each venture to "maximize its potential," is expected to be completed by mid-2026, the company said. This week's agenda includes releases on consumer and producer prices, and key benchmarks on inflation. (AFP)
