From 22 to 26 August 2026, the National Speed Skating Oval in Beijing, nicknamed “ Ice Ribbon ”, hosts the second World Humanoid Robot Games. On the programme : 2 056 robots, 666 teams from 16 countries and 51 events, compared to 26 events in 2025. That makes 1 301 matches, compared to 487 last year. Tickets start at 98 yuan, approximately 14.50 $ (Mashable, Global Times).

These 51 events are divided into two families : 30 sporting competitions and 21 scenario-based events, the latter recreating a real location — an apartment, a hotel, a library. And for the first time, they are doing the cleaning.

It is more interesting than a robot race. To distinguish between 38 teams on chores, a scoring system had to be written. And this scoring system says out loud what most household robot videos leave unspoken.

A scoring system that pays more for laundry than for the living room

The domestic event takes place in a Beijing reference apartment named Habit@. It is scored out of 300 points, distributed across three very specific items (Global Times) :

  • 75 points : tidying the living room 
  • 100 points : folding laundry in the bedroom 
  • 125 points : washing and then hanging the laundry, in the bathroom and on the balcony.

The hierarchy speaks for itself. Laundry is worth more than the living room, because you have to wash, and then hang, between the bathroom and the balcony. In the middle of the chores, a surprise task comes to disrupt everything : on a voice command, the robot must stop, bring in a delivered package, and then return to finish the cleaning. And every misstep is penalised, with points deducted for any collision, any dropped object and any crossing of a limit. Finishing is not enough : you have to finish properly.

A round piloted by a human counts for only half

Here is the real breakthrough in the rules. A round accomplished in decisional autonomy, that is to say when the machine decides for itself what it does, is weighted 1. A teleoperated round, that is to say piloted remotely by a human, is weighted 0.5. Better still : if someone intervenes during the run, it is not just the gesture that is voided. It is the entire round that is reclassified as remote piloting and falls back to the 0.5 coefficient.

On the track, the rules for this edition announce full autonomy, except for the 100 metres and the 400 metres hurdles. In the details of the events, the human is nevertheless still allowed in the long jump and in kickboxing, whereas breakdance, the pom-pom girls and the 4 × 100 metres relay are contested without any intervention.

Someone has finally put a figure on the difference between “ the robot does it ” and “ a human makes it do it ”. According to sector analyst Liu Dingding, cited by the Global Times, another person is still necessary at the exploitation stage to pilot or assist a teleoperated machine — and the robot then loses the value it is supposed to bring.

A hotel room to remake in 30 minutes

The scenario-based events are not limited to the reference apartment, and some took place before the opening. On Sunday 16 August 2026, the hotel event was held at the Beijing Continental Grand Hotel, in Beijing. Each robot had 30 minutes to move a client’s luggage into an assigned room, restock towels, slippers and water bottles, remove dirty laundry and make the bed. Again, the cleanliness of the result is scored, not just the fact of finishing within the allotted time. A bed made in a hurry will not pass.

The same Sunday, another event was taking place in a working library : 30 minutes to collect returns, shelve the books and correct misplaced volumes. This is not a demonstration platform, it is a place where people work.

A word of honesty, nonetheless : at this stage, no results have been published. What we know are the rules and the scoring systems. Not the scores.

What does a robot designed for the home look like ?

The TianYi 2.0 from Beijing manufacturer X-Humanoid, entered in all scenario-based events, gives a good idea. It does not walk, it rolls. And above all, it varies the height of its torso between 130 and 165 centimetres : enough to reach both a towel rack and a bedside table. It lasts more than two hours on a charge.

These two hours, that is the battery. Not the decisional autonomy, the one that counts as coefficient 1. Two different notions, two different words.

And does it really work ?

Not always, and let us say so. On the eve of the Games, a video went viral. It shows, during training on the track, an unidentified humanoid hitting a padded wall, falling backwards onto its head and breaking at the waist while throwing sparks. A second clip shows Unitree’s Superman going out of its lane and hitting electrical cabinets. This Superman is a prototype unveiled on 17 August 2026, announced by its manufacturer at 12.66 m/s (45.6 km/h) — a figure that no independent measurement has validated. It is a training session, not the final, but it puts things back in perspective.

And what about us ?

A German survey published on 24 August 2026 measures that 48 % of those surveyed say they are open to the idea of an AI-piloted humanoid robot in their home. In detail : 18 % “ definitely ” and 30 % “ rather yes ”. On the other side, 47 % refuse (heise). A country split in two.

Read with its technical sheet : the survey was conducted by Bitkom Research for Bitkom, the German digital federation, among 1 159 people aged 16 and over, during the summer of 2026. It was published just before the IFA technology trade fair. A poll commissioned by the industry, on the eve of its major event : you know that before reading the figures.

And for what purpose ? For those whom the idea interests, the list of chores to delegate looks very much like the Beijing scoring system :

  • tidying up : 87 % 
  • putting away groceries : 77 % 
  • cleaning : 73 % 
  • helping in the kitchen : 60 % 
  • looking after pets : 58 % 
  • assisting a dependent person : 55 % 
  • DIY and repairs : 54 %.

So the demand exists, and it is perfectly mundane : we want someone to tidy up for us. What remains is the hurdle to cross, the one that Beijing now quantifies at 0.5.

Because what is bought today for these chores is precisely at 0.5. The cleaning billed at 30 $ an hour by Tau Robotics, in San Francisco, is piloted live by a human operator during each intervention. The NEO from 1X is sold for 20 000 $, or around 500 $ a month on subscription. During the Wall Street Journal test, it did not execute any task alone : everything went through a remote operator, virtual reality headset on its head. As for the Isaac 1 from Weave Robotics, 7 999 $, presented as autonomous, it hands back control to a remote human as soon as it gets stuck. If you are looking at prices, our article on the price of humanoids completes the picture. And before the next video of a robot folding a shirt, ask the only question that matters : is it deciding, or is someone piloting it ?