The first World Humanoid Robot Conference concluded, showcasing intense competition across various race categories. Yushu (宇树) achieved remarkable success, securing gold medals in all track races, including the 1500m, 400m, and 4x100m events.

In the 100-meter race, TsiangGong Ultra (天工Ultra) stood out by winning the gold medal with its innovative autonomous navigation strategy, relying on technologies like LiDAR, surround-view cameras, and advanced algorithms. Tang Jian from the Beijing Humanoid Robot Innovation Center stated that this strategy aims to shed the perception of robots as mere "big toys." He highlighted that the challenge in this field lies in the complex control of over 30 joints, making it more demanding than autonomous driving systems.

Yushu's founder, Wang Xingxing, revealed that their H1 robot achieved an average speed of 3.8 meters per second in the 1500m race, utilizing a remote controller to further "squeeze" the robot's performance. Other participating teams also reported significant improvements in their robots' performance during the short preparation period for the competition. Notably, MagicBot Z1 from Magic Atom (魔法原子) secured the second position in the 100m race. Chen Chunyu, the company's R&D head, confirmed that enhancements in reinforcement learning algorithms, including adjusting reward functions and optimizing joint logic, have increased the robots' maximum speed by 1 meter per second.

The 100-meter hurdle race highlighted the challenges of this category. Yushu's subsidiary, Leying Technology (灵翌科技), completed the race in 38.36 seconds, while the second-place team took 7 minutes and 22 seconds. Six out of eight participants failed to finish the race, indicating a completion rate of only 25%. This underscores the current pain points in the humanoid robotics industry, particularly concerning algorithm robustness, performance stability, and the coordination of perception and motion.

In other categories such as material handling, Tianzhi 2.0 (天轶2.0) from the Beijing Humanoid Robot Innovation Center secured second and third place by completing tasks with full autonomy. In the hotel cleaning races (hotel cleaning), Ulysse (优理奇) clinched the first and second positions. Yang Fengyu, CEO of Ulysse, reported that only three out of approximately twenty participating teams managed to complete tasks autonomously, with most companies still relying on traditional programming. He observed that robots struggle to adapt to minor changes, such as switching beverage brands, exposing their limitations in perception and generalization.

This championship demonstrates that humanoid robots still require significant advancements in perception, adaptability, and generalization to move from demonstration-level to widespread applications. While robots may not be fully "ready" for real-world applications, the challenges and failures encountered by teams in this competition offer valuable insights into the actual progress of this industry.

Regional Perspective

The inaugural World Humanoid Robot Conference, hosted in China, serves as a critical platform for showcasing technological advancements in the field. The competitions, particularly with the participation of Chinese firms like Yushu and TsiangGong Ultra, have highlighted the growing capabilities of humanoid robots in complex tasks. However, the low completion rates in hurdle races reveal significant gaps in robustness and adaptability, areas that demand substantial investment in research and development. The results suggest that Chinese companies are leading innovation in robot performance, but there is an urgent need to improve autonomous capabilities and generalization for broad applications. Investors and companies in this sector can focus on teams demonstrating progress in these key areas, keeping in mind that success in this competition may not necessarily reflect full readiness for commercial applications.

Analysis

The first World Humanoid Robot Conference is a strong indicator of the humanoid robotics industry's evolution, yet it also exposes the significant challenges that remain. Yushu's stellar performance in track races, averaging 3.8 meters per second, showcases progress in motion dynamics and control. However, the 25% completion rate in the hurdle race serves as a stark reminder of the fragility of current systems when faced with unexpected environments. This directly implies that the current focus on improving speed and performance under ideal conditions may not automatically translate to reliability in real-world scenarios. The failure of many robots to adapt to minor changes, like differing beverage brands, suggests that perception and generalization capabilities are still in their nascent stages. For investors, this means that valuations of humanoid robotics companies should look beyond mere competition achievements and focus more on the strength of their machine learning algorithms, sensing capabilities, and adaptability to changing environments. Investment opportunities lie with companies directly addressing these challenges, particularly in areas such as robust reinforcement learning, advanced computer vision, and autonomous control systems. In the long term, companies that can build humanoid robots that are not just fast but also reliable and adaptable will capture market share.