China's Humanoid Robots Advance Toward Commercialization

Daily Technology

06/04/2026

China's humanoid robot sector is transitioning from a phase of technological innovation to one of widespread commercial application. Central to this progress is the enhancement of the robot's "brain"—the sophistication of its underlying algorithms.

The Power of Advanced Algorithms

A key technological driver is the Vision-Language-Action (VLA) model. This single, end-to-end architecture enables robots to achieve precise control, translating visual perception directly into complex joint movements. With VLA, humanoid robots can undertake highly complex, long-sequence tasks, such as making a bed, demonstrating a significant leap in autonomous capability.

Diverse Training Methods and Applications

Different companies are employing varied strategies to train robots for commercial use. Beijing-based startup Spirit AI focuses on learning by imitation. In one scenario, a robot learns to prepare hot pot ingredients by precisely mimicking the movements of a staff member wearing a set of wearable devices.

In contrast, Shenzhen's X Square Robot utilizes in-situ training. Its robots are deployed alongside cleaning staff to provide household services. Their primary task is to assist with organizing sundries and collecting garbage. Through this practical housework, the robots continuously train and optimize their models, improving their performance over time.

Another approach is seen with Zerith Robotics, which focuses on data-driven generalization for tasks like sorting commodities in supermarkets. Using a pre-trained model, the robot requires only a small amount of data—about 20 pieces collected over half an hour—to learn a new category of goods. The goal is for the robot to draw inferences by analogy, generalizing from one type of potato chip to all types, a capability achieved through large model data-driven methods.

Projections for Market Entry

The industry outlook suggests a phased commercial rollout. While achieving multi-dimensional, human-like general artificial intelligence may take five to ten years, significant breakthroughs are expected in specific commercial scenarios much sooner. According to industry experts, pioneering application cases in closed-loop environments are predicted to emerge within the next one to two years. This indicates a near-term future where specialized humanoid robots become a common sight in targeted commercial settings.

2026-03-30

Beijing's 'Robot School' Cultivates Smarter Humanoid Companions

Discover how Beijing's 'Robot School' is training humanoid robots for future applications in manufacturing, homes, and services, leveraging advanced data collection and motion capture technology.