Mainly manages the agile movement, braking, and obstacle avoidance of the wheeled base, while simultaneously controlling the upper body's Balance. Current industry solutions for the wheeled Cerebellum typically utilize an embedded multi-core ARM CPU as the core. Some ARM cores are dedicated to running EtherCAT/CAN FD tasks and driving motors for operation, while the remaining cores run navigation algorithms to enable the robot's mobility.
The core mission of wheeled robots lies in executing practical, actionable tasks, coupled with robust human-robot interaction capabilities (e.g., voice dialogue, expression recognition), scene understanding, and service task execution. Their Brain must perform whole-body task planning, coordinate actions between the mobile base and manipulator arms, meanwhile enable end-to-end large model perception, decision-making, and execution for specific tasks. Current industry standard Brain solutions often utilize Jetson AGX Orin, x86 + RTX 4090, or wireless connections to GPU servers to facilitate voice interaction and the operation of LLM/VLM/VLA models.
Cerebellum: Utilizes the N40, an embedded multi-core ARM-based Cerebellum solution, supporting Real-Time Operating System, CAN FD, EtherCAT, and 485 communication protocols.