Have you seen two robots play tug-of-war? Not the brute-force kind where they anchor the rope to their bodies and yank — but like human tug-of-war athletes: lowering their center of gravity, adjusting their stance, sensing changes in rope tension, surging at the exact moment the opponent flags.

That scene becomes reality in Beijing in two weeks. On August 5, the organizing committee of the 2nd World Humanoid Robot Games unveiled this year's event lineup: competition runs August 22–26 at the National Speed Skating Oval, the "Ice Ribbon," with 30+ events. Most notably: robot tug-of-war, robot ping pong, and humanoid robot full-contact combat — demonstration events at last year's inaugural games — are officially joining the competitive program this year.

The Truth About Tug-of-War: It's About Algorithms, Not Force

Tug-of-war technical representative Feng Baoyun said something revealing in an interview with Beijing Youth Daily: "Humanoid robot tug-of-war isn't about strength — it's about intelligent control."

That gets at the fundamental difference between human sports and robot competitions. Human tug-of-war is about power, weight, teamwork; robot tug-of-war is about balance control algorithms, real-time force feedback, gait adjustment strategy, dynamic center-of-gravity optimization. Whichever robot maintains a lower, more stable center of gravity under changing rope tension, more accurately senses fluctuations in the opponent's pull, adjusts its steps and force direction at the right moment — that robot wins. Motor power barely matters.

This is precisely embodied AI's core technical challenge today. Robots walking on flat lab floors is no longer a problem, but maintaining balance during complex physical interactions, responding to unpredictable external forces, adjusting motion strategies in real time — these are the real hard problems. Tug-of-war, ping pong, and combat essentially package these technical challenges into spectator-friendly sports competitions.

Take full-contact combat: two robots squaring off isn't about whose punch hits harder — it's about who can more accurately read the opponent's movement intent, make evasion and counterattack decisions faster, recover balance quickly after being hit. This requires end-to-end coordination of visual perception, motion planning, balance control, and real-time decision-making — orders of magnitude harder than simply walking or grasping.

Ping pong is similar. Professional table tennis balls travel at tens of meters per second with complex spin variations. A robot must track the high-speed ball in real time, predict its trajectory, adjust arm position and racket angle, control return spin and placement — extremely demanding on vision system latency, manipulator precision, and control algorithm response time.

From Demos to the Arena: Embodied AI's Public Exam

The significance of the Humanoid Robot Games goes far beyond a tech showcase. It's becoming the embodied AI industry's annual "public exam."

The past year has seen a flood of humanoid robotics funding: Unitree Robotics' IPO, Ant LingBot's 1.5 billion yuan round, LimX Dynamics' 2 billion yuan Pre-IPO, Delta Intelligence's $500 million angel round — capital is pouring in, but public perception of actual robot capabilities still lags at short videos of walking, dancing, and somersault demos.

The games provide a public, fair, comparable competitive environment. At last year's inaugural games, many robots struggled to walk steadily; falling was normal. This year's new adversarial events signal that organizers and participants have ratcheted up their confidence in robot capabilities by a full notch. From walking to running, from independent locomotion to physical competition, embodied AI's progress is visible to the naked eye.

Holding the games at the National Speed Skating Oval — the "Ice Ribbon" — is also symbolic. This was the speed skating venue for the Beijing Winter Olympics, where human athletes pushed speed and passion; two weeks from now, it hosts the "Winter Olympics" of humanoid robots in the physical world — competing not on citius, altius, fortius but on stability, precision, and intelligence.

Objectively, today's humanoid robots still have a long way to go before truly entering homes, factories, and daily life. Winning tug-of-war doesn't mean they can haul heavy loads across rough terrain; winning ping pong doesn't mean they can cook in a cluttered kitchen. But arenas are the best catalyst — competition exposes problems, problems drive R&D, R&D delivers progress.

At last year's inaugural games, some viewers joked that "robots falling is more entertaining than robots walking." This year, let's see if the robots can hold their ground at the Ice Ribbon.

See you tomorrow.