Perception
· Visual: Humanoid binocular/wrist vision, high-prec 3D recon & object recognition
· Force: Optional 6-axis wrist force sensor, compliant & safe
· Speech: LLM-integrated, hyper-human synthesis, real-time lifelike chat
· Navigation: SLAM-led (VSLAM-assisted), adapts to harsh environments
Brain + Cerebellum
· Brain-Cerebellum Joint Control (powered by large models)
· Sim2Real VLA Pipeline: Massive data online gen & training (via EmbodiChain)
· Total system computing power ≥ 300 TOPS
· Dual-PC hardware support for enhanced computing performance
Execution
· Max total DOF: 40pcs
· Arm: High flexibility/reach, 10kg max load (single)
· Torso: Lift/rotate/pitch
· Chassis: Optional manual/differential/omni
· Gripper: Dexterous/2-finger, fits home/commercial/industrial
· High-freq control: 1000Hz
· High-speed comms: 100Mbps (power unit)

6cm Baseline Humanoid Binocular Vision
Interactive Voice Model
Force-Torque Sensor
Wrist-Mounted Cameras×2
LiDAR for SLAM Navigation and Obstacle Avoidance×2
Camera for VSLAM Navigation and Obstacle Avoidance×2
Sim2Real VLA
PC2
PC1
Neck: 2 DOF
Arm: 7 DOF per arm
Waist: 2 DOF
End-Effector:
6 DOF per dexterous hand
2 DOF per two-finger gripper
6 DOF per dexterous hand
2 DOF per two-finger gripper
Leg: 2 DOF
Chassis:
8 DOF for omnidirectional chasis
2 DOF for differential chassis
8 DOF for omnidirectional chasis
2 DOF for differential chassis














