Zhicheng AI CEO Takes on Generalization Challenges in Embodied Intelligence With World Models

According to a report by leaderobot.com on Jan. 6, 2026, Zhicheng AI CEO Hu Luhui outlined a plan to achieve generalization across environments, robot bodies and tasks, centered on world models, at the Sixth China Robotics Industry Annual Conference, held in Hangzhou on Dec. 18-19.

The remarks matter because they set out a plan to use world models to test whether the burden of preparing separate models for each environment, robot body and task can be reduced for those developing and deploying robots for factories and homes. Hu said insufficient generalization was behind the fact that current robots are “expensive” and “difficult to use,” according to the company’s assessment of the problem. The approach aims to address adaptation to homes and factories, use of the same model across different robot bodies, and multiple tasks such as cooking, transport and inspection together, unlike conventional task-by-task optimization.