Equally important is a “cerebellum“ that can achieve functions such as grasping, skill breakdown and error handling, and is responsible for tasks like full-body control, bimanual collaboration, stable walking and mobile navigation.

　　The robot can first have its tasks planned by the “brain,“ then order the “cerebellum“ to execute specific actions, and pass the execution feedback back to the “brain,“ forming a task loop to complete autonomous decision-making and motion execution in complex environments.

　　Xiong Youjun, the general manager of the center, said that embodied AI is still in its infancy. The industry is in need of a universal intelligent platform that is compatible with different body types and scenarios, and can perform general tasks.

　　At a launch event on Wednesday, the platform demonstrated real-machine operations in four scenarios: industrial sorting, blocks building, desktop cleaning, and logistics packaging. Through various means such as voice interaction and direct connection via an APP, users can easily interact with the robots.

　　For example, in the industrial sorting task, the operator directly connected through the HuiSi KaiWu APP, while robotic arms accurately interpreted the voice commands and completed the sorting operation through bimanual collaboration.

　　Tang said that the “APP plus robot“ model packages complex technical capabilities such as reasoning, planning and skill invocation into simple procedures, lowering the threshold for users. Meanwhile, the platform supports the rapid addition of customized models and skills, which can flexibly meet the needs of different scenarios and provide convenient, efficient and intelligent solutions for industrial automation.