In the humanoid robot sports games, there is no real winner.

Perhaps, this sports event was never meant to find the "champion". "Use competition rules to drive technological breakthroughs and use market attention to nurture industrial development." When the whistle of this sports event sounds, humanoid robot companies will find their own implementation positions in the industrial coordinate system.

The first humanoid robot sports event has many underlying intentions.

Let's first look at the opening ceremony. All the programs were jointly performed by humans and humanoid robots. Hu Bing walked the runway with a group of humanoid robots, martial arts teenagers and young opera actors performed on the same stage with humanoid robots, etc. Many programs presented the resonance and harmony between humans and humanoid robots, composing the prelude to the "era of human - machine collaboration".

Of course, there were also many incidents during the opening ceremony. Some robots stumbled and fell, while others malfunctioned and had to be carried away. There may also be mistakes during the competition, but these are not "negative news". Instead, by exposing "uncertainties" or bad cases, we can find and solve problems in real - world environments and promote long - term development.

This is not just a competition; it is a window through which we can glimpse the future.

Image source: Humanoid Robot Sports Event

This grand event was held at the National Speed Skating Oval from August 15th to 17th. There were as many as 127 brands participating, with more than 500 humanoid robots gathered. It is reported that 280 teams from 16 countries on five continents around the world will participate, including 192 university teams and 88 enterprise teams.

Different from a marathon, this sports event does not have long - distance events. Compared with a half - marathon that tests endurance, this event focuses more on hardware and body structure design. The popular competitive events in this sports event are mainly short - distance and group events, such as the 100 - meter dash, obstacle - crossing, long - jump, etc., which rely on explosive power and body coordination. The multi - robot football game tests the group collaboration and intelligence among robots.

The exhibition games and scenario - based competitions mainly examine the practical skills and intelligence levels of humanoid robots to explore their implementation directions.

Looking back, competitive sports events like football games in the sports field correspond to the "sports - event - based economy"; dance performances are the earliest implementation scenarios for robots and are relatively mature. In addition, application - oriented scenarios such as factory operations, hospital work, and hotel services correspond to more specific industrial, medical, and service - industry scenarios.

This multi - dimensional humanoid robot event covering "sports + art + application" is not about determining rankings. The real intention is to anchor technology to scenarios. On the eve of large - scale mass production, it aims to find "useful applications" for robots.

From the competition results, there is no real winner in each event. The first - place runner in a race does not necessarily have the best gait, and the first robot to complete a cleaning task may not be the one that cleans the most thoroughly. Each participant has its own unique advantages, and crossing the finish line first is just one of the evaluation criteria.

Perhaps, this sports event was never meant to find the "champion".

"Use competition rules to drive technological breakthroughs and use market attention to nurture industrial development." When the whistle of this sports event sounds, humanoid robot companies will find their own implementation positions in the industrial coordinate system.

Two Core Questions: Unitree and Remote Control

Will remote control be allowed? Will the Unitree team participate?

The answers are both yes.

Let's first talk about remote control. The competition rules clearly state that the control method of the robot can be manual remote control, but only wireless remote control is allowed. The remote - control operator must conduct the operation within a designated area. It can also be fully autonomous. The event regulations state that for the free - gymnastics, single - robot dance, and group - dance events, the control method of the robots must be fully autonomous.

Now, let's talk about Unitree. Unitree officially announced before the event that it would participate in the first World Humanoid Robot Sports Event. Unitree also emphasized that in addition to its own team, there are also multiple teams using Unitree's hardware (with algorithms developed independently by each team).

Before the event, warm - up videos of robots from Unitree, Tianguang, etc. were released. A robot suspected to be from Unitree's H1 series ran at a sprint speed faster than humans, which was in sharp contrast to the slow pace in the half - marathon. Unitree Technology has an opportunity to prove itself. As of the time of writing, Unitree Technology has won the first place in the first group of the 1500 - meter final.

Overall, humanoid robot manufacturers generally try to distinguish their products from the label of "big toys", and "remote control" is a sensitive topic in the industry.

In fact, there is no clear superiority between remote control and autonomous decision - making. There is no absolute right or wrong in technical routes. The former tests the robot's body, while the latter tests its "brain".

The core of autonomous decision - making is "machines replacing human decision - making". Through algorithms, robots can autonomously plan their actions based on environmental perception, which is suitable for unstructured, highly repetitive, or low - risk scenarios. For example, in commercial scenarios, robots can rely on visual recognition to pick up food packages, place them in the cargo hold, sense the dynamic environment in real - time, and autonomously plan delivery routes without human intervention.

The core of remote control is "human - machine collaboration" and "human - led decision - making". Through remote control, robots can extend human operating capabilities.

For example, employees of Yuejiang in Shenzhen can remotely control the humanoid robot atom in Shandong through a VR helmet to make steaks; the Robody nursing humanoid robot of the German robot startup Devanthro allows nurses to remotely control the robot to take care of the daily lives of the elderly.

Autonomous decision - making represents the ultimate goal, but it does not mean that remote control is useless. What we need are embodied intelligent products that can truly solve problems.

Both are essentially about constructing solutions around scenario pain points. In scenarios with different requirements for embodied intelligence, there is no need to worry about the "superiority" of technical routes. We only need to aim to meet the actual needs of the scenarios and ultimately form a virtuous cycle of "scenarios defining technical directions and technology serving scenario pain points", so that robot technology can truly take root in the practice of solving problems.

Don't Just Focus on the "Champion"

In this humanoid robot sports event, individual events like the 100 - meter dash and the 100 - meter obstacle race are as eye - catching as their human counterparts. This type of competition best reflects the performance of robots, including lightweight design, high - torque motors, and motion strategies such as step frequency and gait.

It can be seen that this event will have a similar "event spill - over effect". Referring to Songyan Power, the outstanding performer in the April marathon event, it received a large number of orders after the event due to the excellent performance of its N2 model. It is reported that it received 2000 commercial orders in the first half of this year.

Image source: Humanoid Robot Sports Event

In addition, Jiasu Dongli's T1 and K1 models won championships at the 2025 RoboCup Brazilian Robot Football World Cup in June, which also led to a record - high product delivery volume in July. More than 120 units of the two championship models were delivered in total. It is reported that more than 95% of the teams in this event will use their robot bodies to participate.

In group events, 3V3 or 5V5 football games and group dances fully test the group collaboration ability of robots. Each robot needs to sense the positions of other robots in real - time, coordinate with each other, and plan the goal - scoring route.

The core goals of multi - robot collaboration are "information sharing, task allocation, path planning, and action coordination". With each additional robot on the field, the decision - making dimension of the robots will increase significantly.

When Figure AI released the Helix model, in the video, two robots cooperated with each other to deliver ingredients and put them into the refrigerator. They solved the smallest unit problem of "group collaboration" in the home - service scenario; from dozens of UBTECH Walker S1 robots being stationed in the Geely 5G factory for multi - robot, multi - scenario, and multi - task collaborative training, to the cross - model ecological collaboration of multiple hardware products of Youibot and Qinglang, this is the prototype of "group intelligence".

Of course, this group collaboration ability will enable robots to form a "1 + 1>2" synergy effect. On a small scale, they can link with other intelligent agents to form a service network; on a large scale, they can form a more extensive productive force and even give rise to a "production relationship" similar to that in human society.

Of course, whether it is an individual event or a group - collaboration competition, there is no absolute "winner" in the robot field, just like in the human 100 - meter race, the fastest runner may not have the most scientific gait.

In this event, the 100 - meter champion model may be unparalleled in straight - line speed, but in a group dance, it may not be able to adapt to the rhythm of its partners; the champion team in a football game may have a single robot with less explosive power than the top performers in individual events, but it excels in precise collaborative decision - making. There is no such thing as a "one - move - to - win" in robot competitions.

Three Clear Scenarios and a Closed - Loop Workflow

As a multi - dimensional humanoid robot event covering "sports + art + application", it can correspond to implementation scenarios such as sports events, exhibitions, and factory, service - industry, and family scenarios.

Image source: Humanoid Robot Sports Event

In a previous article The Person Who Sends Humanoid Robots to Compete Must Be a Genius by the Embodied Learning Club, it was proposed that "use competition rules to drive technological breakthroughs and use market attention to nurture industrial development". Through intense competition and the outcome of the games, enterprises are stimulated to iterate their technologies. The impactful on - site performances are far more effective than demo promotions and technical white papers and can bring real orders to enterprises.

Furthermore, through a series of operations, sports events can bring revenue to enterprises. In a previous interview, Zhao Weichen, the vice - president of Jiasu Jinhua, specifically mentioned a similar view. He pointed out that on the one hand, sports - event scenarios and education scenarios resonate with each other. The "research + competition" combination is strongly linked. Its football - robot courses have entered many universities and middle schools, and the education - scenario market is larger. On the other hand, there is also the possibility of revenue from the IP - based operation of sports events. Recently, its "Robot Super League" has achieved commercial success - "700 tickets sold, many sponsors, and over 300 million exposures".

The "art" dimension can be regarded as a general term for exhibition scenarios. Currently, its main functions are to attract customers to shopping malls and liven up events. This type of scenario can be compared to drone shows, and the market for it actually exists.

Moreover, if we expand the imagination of the "art" dimension: robots can play roles that require precise and repetitive actions on the drama stage, serve as tour guides in museums who can adjust their tones according to the stories of the exhibits, or even cooperate with human artists to create installation art that combines mechanical aesthetics and humanistic thinking.

There are already such "art - savvy" products in the industry. The PNDbotics humanoid robot Adam can either play the guitar on stage with human singers at a music festival; or appear in the same frame as the historical buildings and traditional skills on the Beijing Central Axis; or witness the aesthetics created by humans in the CAFA Art Museum... At this moment, mechanical aesthetics truly shines.

Finally, there is the "application" dimension. The scenarios shown here are closer to daily life compared to sports and art and most test the embodied intelligence of the robots. Their performance in tasks is related to whether robots can become a new form of productive force and enter people's lives and homes.

But when you carefully look at the competition items in the scenario - based events, you will find that the "workflow" is closed - loop. From material handling to material sorting in factories, from drug classification to drug splitting and packaging in hospitals, from greeting guests to cleaning in hotels, these competition items are not scattered, independent - unit - based skill assessments but a closed - loop test of a whole set of work processes.

Taking the factory environment as an example, Wang He, the founder and CTO of Galaxy Universal, admitted in a recent interview: "Handling is just the first step. In addition to handling, the ability to complete the closed - loop of palletizing is also required. Only when both handling and palletizing are completed in a closed - loop can the robot truly be competent for the whole set of work processes. Otherwise, if it only completes half of the task, the effect is not ideal."

The closed - loop of the workflow means that robots are no longer just a gimmick for scenarios but have the possibility of performing a whole set of operations. In essence, it marks the paradigm shift of robots from the "tool attribute" to the "system role". In other words, if a robot only serves as a substitute for a single process, it still has to compete with automated equipment. After the closed - loop workflow, the operation content and ROI of the robot become clearer, and it will truly play its value in a systematic and embedded way.

This sports event constructs a "technology - scenario" coordinate system. Humanoid robot companies will clarify the positioning of their products in the competition and find the direction for technological iteration in the game. This "military review" will further accelerate the maturity of the industry.

This article is from the WeChat official account "Embodied Learning Club", author: Peng Kunfang, editor: Lü Xinyi. It is published by 36Kr with permission.