- Thirty robots from 16 companies enroll as inaugural students
- Program aims to equip machines with job-specific intelligence and industry certification
China’s first school dedicated to training robots has opened in Hangzhou, marking an effort to address what researchers see as the biggest bottleneck in embodied AI: developing intelligent “brains” rather than increasingly capable hardware.
The Robotics Institute of Zhejiang University, together with the Zhejiang Institute of Quality Science and industry partners, officially launched the Hangzhou Robot School on June 29 in Hangzhou’s Yuhang District.
Thirty robots from 16 companies have enrolled as the first class.
Structured vocational training
Representing industries including manufacturing, services, security and entertainment, they will undergo structured vocational training, learn occupation-specific skills and receive professional certification before entering commercial deployment.
The initiative addresses a common weakness across the robotics industry. While many companies have become adept at building increasingly sophisticated robot bodies, relatively few possess the AI capabilities needed to develop adaptable cognitive systems.
This shortcoming leaves many machines reliant on pre-programmed behaviors that struggle outside carefully controlled environments.
“The industry doesn’t lack robots that can move—it lacks robots that can think,” said Zhu Shiqiang, president of the Robotics Institute of Zhejiang University. “Many companies can build capable hardware, but that doesn’t necessarily translate into good products.”
“The missing piece is the ability to develop an AI brain that can adapt to specific application scenarios,” he added.
Reasoning layer
According to the institute, fewer than 10% of robotics startups globally survive beyond three years, with many failing to bridge the gap between working prototypes and commercially viable products.
At the core of the school’s curriculum is the institute’s self-developed WJ-Brain cognitive architecture, which introduces a reasoning layer to the conventional vision-language-action (VLA) framework.
Rather than simply recalling learned behaviors, the system is designed to infer physical rules, evaluate task constraints and predict the consequences of actions before executing them.
The research team has developed separate versions for humanoid and quadruped robots, known respectively as WJ-Human Brain and WJ-Dog Brain.
In less than two weeks
One early example is a guide robot named Xiaoji. Just weeks ago, it functioned primarily as a mobile platform.
After being equipped with WJ-Human Brain, researchers said it learned tour guiding, visitor reception, equipment operation, multilingual communication and open-ended question answering in less than two weeks.
The robot is expected to serve as a docent at the China Painting Heritage Museum.
Standardized training pathway
Mirroring China’s vocational education system, the robot school has established a standardized training pathway.
Each robot first undergoes a technical assessment covering hardware performance and software compatibility before entering one of several specialized tracks—including industrial, healthcare, arts and sports robotics.
Upon completing training, robots that pass independent evaluations will receive a specialized skills certificate and a unique digital identity, allowing them to be deployed with verified professional qualifications.
