Humanoid robots ran faster and jumped higher than last year at the World Humanoid Robot Games in Beijing, but the event also showed why athletic demonstrations are a poor substitute for reliable autonomy. Sprinting machines crashed into barriers, broke apart and, in several cases, caught fire after crossing the finish line.
Researchers attributed the improved races to advances in whole-body control. Yet many robots were optimized for a single event and could not perform a basic follow-up action such as slowing down safely. Some competitions also permitted remote human control, with a scoring penalty for teleoperation.
The new practical events offered a clearer test. Robots folded laundry, restocked hotel rooms, shelved books and manipulated small objects. In a firefighting course, teams had 30 minutes to identify hazards, close three different valves, locate a fire and use an extinguisher. Only three of 12 teams completed the full sequence.
These results suggest that specialized motion is advancing faster than general, dependable work. Real deployments require repeated success, recovery from interruptions and safe operation around people—not one optimized run. Many competitors also depended on cloud computing rather than carrying all required processing onboard.