Published : 30/04/2026
At the 2026 Beijing E-Town Half Marathon and Humanoid Robot Half Marathon (hereinafter referred to as the "Robot Marathon"), the winning robot's time was two-thirds shorter than last year's champion's, and even surpassed the men's human half-marathon record.
So, in what areas have humanoid robots improved over the past year?
Champion's time two-thirds shorter than last year
What improvements have been made in this year's Robot Marathon compared to the first event last year? Not only have the race results improved by leaps and bounds, but the participation and influence of the competition have also expanded significantly.
The most obvious change comes from the huge increase in speed. This year's Robot Marathon champion is Honor's "Lightning" autonomous navigation robot, with a time of 50 minutes and 26 seconds, which is two-thirds shorter than last year's champion, "Tiangong Ultra", which finished in 2 hours, 40 minutes, and 42 seconds.
This result even surpassed the men's human half-marathon world record of 57 minutes and 31 seconds, setting a new historical precedent.
Additionally, the number of participating teams surged from over 20 last year to more than 100 this year, an almost five-fold expansion. The level of internationalisation has also increased; for the first time, five international teams from countries such as Germany, France, and Portugal participated.
In terms of completion rate, a total of 47 teams successfully finished the race this year, with a completion rate of over 45%, whereas last year, only 6 teams were able to complete the entire course.
Read more: How do China's industrial robots accelerate smart manufacturing?
What are the advancements in robot technology?
The increase in the completion rate indicates that the performance and technology of the robots have significantly improved in all aspects over the past year. In what specific areas is this manifested?
If last year's competition was primarily about the basic running capabilities of the hardware, then this year's core breakthrough lies in the evolution of the "brain".
Last year, the vast majority of robots still required manual remote control, whereas this year, nearly 40% of the participating teams used autonomous navigation technology.
This means that robots can navigate on their own, plan their routes, and avoid obstacles without relying entirely on human instructions.
This year's competition innovatively established two participation methods: autonomous navigation and remote control, with both groups competing on the same track and ranked together.
However, the autonomous navigation group and the remote control group's results were calculated using weighting factors of 1.0 and 1.2, respectively, meaning the remote control group's completion time was multiplied by 1.2 to determine their final score.
This measure was intended to encourage participating teams to use the autonomous navigation method as much as possible.
The second significant improvement is battery life. Honor's "Lightning" only had its battery changed once, with a single battery lasting for more than 10 kilometres, a substantial leap from the previous common endurance level of 4-5 kilometres.
In response to the joint overheating issues exposed in last year's competition, the teams have made extensive optimisations. This year's robots are generally equipped with water-cooling/air-cooling systems, and the joint temperature has been reduced from the previous 70-80℃ to 60℃.
The battery replacement time has also been shortened from three or four minutes to 10 seconds, and there is no need to restart the system, significantly reducing time loss.
Why did the robot from phone manufacturer win the championship?
When you see the champion "Honor" mentioned above, does it make you think of the well-known mobile phone brand HONOR? That's right, they are the same company.
First-time participant Honor defeated established robot manufacturers such as Unitree and Neotix, becoming the dark horse of this competition.
Honor not only won the first, second and third place in this robot marathon, but also secured the fourth, fifth, and sixth places, as well as the Best Gait Award. For Honor, participating the marathon is just a means to an end, its real goal is to build an AI ecosystem.
In the broader context of advancements in humanoid robots, the cross-industry company Honor has leveraged its deep accumulation of consumer electronics technology over many years to meticulously plan its path in robotics.
The simulation laboratory that Honor built from scratch has accumulated a database of over 1,000 types of materials. With the support of artificial intelligence algorithms, it can iterate nearly 30,000 design schemes in 3 months, enabling rapid verification and rapid optimisation, which made it possible to complete development and take shape in just a few months.
What technological breakthroughs and industrial forces are behind the champion?
Looking at the overall development of humanoid robots, their rapid progress—from stumbling last year to running steadily this year—hides three core secrets: algorithm training, the physical body, and perception and decision-making.
At the same time, this competition is also a major review of China's robotics industry chain, including upstream high-precision sensors and lidar, midstream motors, batteries, and controllers, and downstream chips and edge computing.
The participating teams used their self-developed technologies, utilised open-source platforms to quickly verify algorithms, and skipped ground-up development, saving time and energy to focus on technological innovation.
Read more: How does China maintain its position as world's 1st robot producer?
Why let robots run marathons? What else can be improved?
Everyone hopes that robots can enter daily life sooner to serve humanity, so why train robots to run marathons?
From a technological innovation perspective, the competition forces robots out of the laboratory to verify their capabilities, identify problems, and optimise algorithms on real road conditions, compelling them to iterate and upgrade.
From a social awareness perspective, the marathon itself is an excellent popular science activity. It not only increases the popularity of robots but also attracts young people's interest in them, preparing the way for cultivating talent for the robotics industry.
Although the performance of the robots running the marathon this year has shown great improvement compared to last year, there is still clear room for improvement, such as enhancing autonomous navigation and anti-interference capabilities, and continuing to improve cooling solutions and fall recovery strategies.
In addition, the racecourses for the past two years have been flat roads, and the robots have not yet been tested on complex terrains such as slopes and stairs.
From "whether they can walk" to "how well they walk", humanoid robots are advancing in a more stable, more intelligent, and more user-friendly direction. Using the robot marathon as a training ground, every small step on the race track is a big step for humanoid robots towards practical application.