Summary of Key Points
The most striking takeaway from this World Humanoid Robot Games is that robots have surpassed human elite performance in "extreme sports under specific conditions," yet their ability to perform tasks reliably in real-world scenarios is still far from that of humans. On one hand, robots have broken various human world records in events such as the 100-meter dash, high jump, and long jump. On the other hand, they frequently experience issues like falling, freezing in place, and crashing into walls, making them slow and clumsy when working in real-life settings such as homes and supermarkets. At the same time, the industry is beginning to diverge: some teams are focusing on achieving extreme performance, while others are转向 mass production and practical applications.
1. Extreme Performance: Robots Outperform Humans
The athletics events were the most impressive, with robots leaving human champions behind:
- 100-meter Dash: The Tian Gong Ultra robot from the Beijing Humanoid Robot Innovation Center ran 8.64 seconds, faster than Usain Bolt's record of 9.58 seconds; the Honor "Lightning" also achieved 8.83 seconds.
- High Jump: The Tian Zhuo team's robot jumped 3.4 meters without any assistance, while the human male high jump record is only 2.45 meters (the difference is even greater when considering the difference between standing start and running start).
- Long Jump: The Tian Zhuo team's robot jumped 4.83 meters, significantly improving from last year's 1.25 meters.
These advancements are not accidental; they are the result of substantial improvements in:
- Hardware: More powerful joint motors, lighter bodies, and "streamlining designs" to reduce running resistance (similar to racing cars).
- Algorithms: Optimized for different events, such as enhanced acceleration at the start of the 100-meter dash and improved curve control for the 400-meter dash.
- Autonomous Learning: The Tian Gong Omni robot's running posture (with arms tucked in) was not designed by engineers; it was the optimal solution it found through simulation training—this posture reduces wind resistance and stabilizes its center of gravity, showing that robots are beginning to use their own logic to improve efficiency.
More importantly, most of these robots have moved away from being controlled remotely and can now complete the entire process from starting to finishing on their own, demonstrating significant progress in intelligence.
2. Challenges and Shortcomings: Robots’ Strengths and Weaknesses
Behind the impressive performances, there are several shortcomings that highlight three major issues:
1. Reliability:
- Many robots dropped out of the competition; in the 400-meter preliminaries, only 1-2 robots completed each of the 5 lanes.
- Difficulty in completing the race: Some robots froze in place upon hearing the start signal, others fell while turning, and some crashed into walls upon reaching the finish line (even the champion, Tian Gong Ultra, did this).
- Reasons: Difficulty in maintaining balance at high speeds (complex calculations for turning), motor overheating (triggering safety mechanisms and resulting in loss of power), and challenges with braking for both legs (both slowing down and staying stable).
2. Perception and Coordination:
- Free fighting events were a source of humor: Robots either swung their fists at the air or fell while trying to kick their opponents, sometimes even tripping over the ropes of the ring.
- Reasons: Most robots are still remotely controlled (operators cannot see depth and distance, leading to signal delays); if they were fully autonomous, they would need to instantly perceive their opponents, make decisions, and counterattacks, but their "brains" are not yet capable of handling this.
3. Industry Divisions:
- Last year's champion, Yu Shu, performed moderately this year (12.41 seconds in the 100-meter dash, ranking low in its group). Could it be because it focused on mass-producing products and reduced its investment in the competition?
- In contrast, Tian Gong is pursuing extreme performance, while Yu Shu focuses on mass production. The industry is dividing into two paths: should robots be made to run the fastest or be more marketable?
3. Real-World Challenges: Everyday Tasks Prove Difficult
The games included 21 scenario-based events in various settings (homes, supermarkets, hotels, etc.). Tasks that seem simple (opening windows, retrieving items, nailing nails) proved challenging for the robots:
- Galaxy General: The robot completed a full house cleaning task autonomously (organizing, washing clothes, picking up deliveries), but humans could finish this in 10 minutes.
- You Li Qi: Won the hotel scenario championship for two consecutive years and collaborates with hotels; however, hotel environments are relatively standardized, making it difficult for robots to adapt to more chaotic home environments.
- Zhi Yuan: Used the same hardware for multiple scenarios (firefighting, libraries), and its dexterous hands can perform basic tasks like building blocks and weighing powder, but precise operations (such as picking up a bottle) are still very basic.
The biggest obstacle is generalization ability: Robots can learn to open a Coke bottle but not a mineral water bottle; they may also struggle with simple tasks like moving a table lamp from one side of a table to the other. Solving these issues is much more difficult than simply improving running speed.
4. Industry Insights: The First Half is Over, the Real Battle Lies Ahead
This competition marks the end of the "first half" of the humanoid robot development phase (breaking performance limits). The next "half" will focus on solving three key problems:
1. Stability over Speed: Being able to run 10 seconds without crashing is more practical than running 8 seconds but crashing frequently.
2. Generalization Ability: Robots need to be capable of performing tasks in various scenarios, not just repeating fixed actions.
3. Balancing Extreme Performance with Practicality: Achieving extreme performance is impressive, but mass production and practical applications are essential for real profitability.
In short, robots have already surpassed humans in speed, but there is still a long way to go before they can outperform humans in practical tasks. After all, in the real world, no one needs a robot that can run faster than Usain Bolt but can't even hold a bottle steadily.
This competition serves as a mirror, reflecting both the progress of humanoid robots and the real challenges faced by the industry. Breaking records is just the beginning; the real test of success is whether robots can be useful in everyday life.