Continuously computes the Zero Moment Point (ZMP) or utilizes Whole-Body Control (WBC) at high frequencies to adjust leg joint angles in real-time, which enables stable walking, running, and jumping, while coping with uneven terrain, external pushes, or even slips. It also handles emergency situations like fall protection and autonomous recovery. The current industry standard typically utilizes ARM or x86 CPU-based solutions.
Compared to wheeled robots, such systems place far greater demands on self-posture control and spatial recognition for constantly facing challenges of kinematic stability during operation. They possess higher sensor integration and processing capabilities, enabling long-horizon task planning (e.g., "traverse multiple rooms to locate and close a valve") along with 3D reconstruction and navigation in complex environments. The current industry standard typically utilizes high-performance GPUs, such as the Orin developer kit. However, developer kits often face limitations in sensor integration capabilities and bandwidth.
Cerebellum: The T80 utilizes a high-performance ARM; the N204 utilizes a high-performance x86.
Brain: The N201, T300, and T500 series provide high-compute Brain solutions supporting multi-sensor integration.