Summary of Key Highlights
The 2026 World Robot Conference (WRC) was an unprecedented success, with 373 companies showcasing 3,000 products. Robots have moved beyond the stage of being merely “cool demonstrations” and are now being integrated into real work scenarios—making coffee, doing household chores, and performing industrial sorting tasks. However, the industry has not yet reached a consensus on the technical approaches for embodied intelligence, with disagreements ranging from data training methods to model architectures. While companies are seizing the opportunity to deploy these technologies, they also face uncertainties about rapid technological advancements. Nevertheless, industry insiders believe that this “chaos” is an essential part of the innovation process.
1. The Conference’s Success: Robots Moving from Showoff to Practical Use
The most notable change at this year’s WRC was the shift in robot functionality from mere performances (such as boxing and playing the piano) to actual work tasks:
- Service Applications: Robots from Zhi Ping Fang can make coffee, ice cream, and even work as shop assistants, engaging in rock-paper-scissors games with visitors.
- Home Applications: Robots can organize clothes, deliver food, and manage cat litter; they can also handle packages intelligently—lifting light ones, pushing heavy ones, and flattening soft ones.
- Industrial Applications: Yu Bi Xuan’s humanoid robots can load and unload car parts, while highly skilled robotic arms can assemble themselves.
- Collaborative Applications: Xian Gong Zhi Neng’s “robot brains” can coordinate the work of humanoid robots, delivery vehicles, and forklifts, with services available in 43 countries.
The overwhelming attendance and the widespread sharing of videos on social media demonstrate that robots are becoming increasingly integrated into everyday life—no longer just a product of science fiction, but practical tools that solve real problems.
2. Technological Divisions: The Industry is Still in a State of Trial and Error
Despite the excitement around applications, there are significant disagreements within the industry regarding technical approaches:
- Data Training: Some companies use “human-first perspective data” (e.g., DynaRobotics’ Dyna-2 model, which was trained with 1 million hours of real human activity videos), while others combine virtual simulation data with real-world data (e.g., Da Xiao Robot’s three-tier training method). Some argue that the diversity of use cases is more important than the amount of data collected (e.g., data from hospitals and ultra-clean environments is particularly valuable).
- Model Capacity: Models like GEN-1.5 can learn new tasks quickly (e.g., by observing a 3–12-second demonstration). While this ability is similar to that of large language models, it’s still uncertain whether it will become the standard.
In simple terms, it’s like a group of chefs learning to cook; some use traditional methods, some use modern tools, and no consensus has been reached on the “optimal recipe.”
3. Companies’ Dual Concerns: Both Deployment and Technological Advancement
While exhibitors are highlighting their applications, they are also grappling with dual challenges:
- Deployment Pressure: They need to develop profitable products quickly. For example, JD.com is establishing robot maintenance centers with plans to expand to 50 cities within three years, aiming to ensure that robots are both useful and easily maintained.
- Technological Uncertainty: Industry forums are filled with discussions about the potential obsolescence of current technologies. For instance, a CEO from a Shenzhen-based company stated, “Technology will inevitably converge in five years, but what we do now may not be the standard. However, this is the only way forward.”
It’s akin to students preparing for exams while simultaneously learning new concepts, fearing to fall behind but unable to skip the current learning process.
4. Chaos as a Path to Innovation: Embrace the Uncertainty
Industry insiders see this chaos as a positive:
- Similar to the transition from convolutional neural networks (CNN) to transformers in the AI field, current disagreements are temporary, and a unified framework will eventually emerge.
- The current period of trial and error is not a waste but a critical step in the innovation process. Robots trained with different methods may help identify the most effective approaches for the industry.
In other words, the current “chaos” is a necessary phase for the robot industry to mature.
Conclusion
The robotics industry is at a stage where applications are booming, but the underlying technologies are still evolving. On one hand, robots are becoming an integral part of our lives, solving practical problems; on the other hand, there is intense debate about the best technical approaches. For consumers, this means that more intelligent robots will soon be available. For companies, it presents both opportunities and challenges—those who can navigate this period of experimentation and identify the most effective solutions will emerge as leaders in the industry.