RobotGuide

The industry's current main architecture

Brain and cerebellum: why a robot needs two «brains»

However strong a large model is, it cannot directly control a 1 kHz joint. Understanding the brain–cerebellum split is the highest-leverage piece of knowledge for entering robotics today — it explains where the algorithm, architecture, and embedded-control roles each sit.

One metaphor

The brain is navigation software: it tells you where to drive, and being a few seconds late is fine. The cerebellum is your own hands and feet: 1 ms late and the car is in the ditch. Safety reflexes are «pull the hand back when it burns» — they never pass through the brain at all.

The split in one table

LayerTypical instanceFrequencyWhat it ownsLatency requirement
BrainVLA / multimodal large models1–10HzUnderstand the scene, parse the instruction, decide «what to do»seconds, tolerable
Middle layerSkill library / trajectory planning / WBC10–100HzTurn intent into executable trajectories and torque targetsstable within tens of ms
CerebellumReal-time control (joint servo / whole-body control)500Hz–1kHz
(current loop 10kHz+)
Track trajectories, keep balance, close the torque loopjitter < 100µs, above that it falls
Reflex layerHardware protection in MCU / hard limitsmicrosecondsOvercurrent, overspeed, e-stop — no model involvedmust be hardware-level

The table mirrors the human brain: the prefrontal cortex is slow but smart, spinal reflexes are fast and dumb. Not a coincidence — it is a division of labor forced by the physics of latency.

Three integration patterns

Cloud brain + onboard cerebellum

The large model runs in the cloud; the robot fetches instructions over the network. Fast to deploy, but a network drop is paralysis and latency jitter is uncontrolled.

Fits: demos and remote orchestration

Dual-system on board (current mainstream)

Slow system (VLA) + fast system (real-time control) both on the robot's own GPU, joined by one intent vector. Examples: Figure Helix, GR00T N1.

Examples: Helix · GR00T

End-to-end single model (research stage)

One large model maps pixels straight to joint commands. The most elegant, but high-frequency control latency and safety proofs remain unsolved.

Status: many papers, few products

Whatever the school, the bottom layer has one constant

Safety reflexes never pass through a model: overcurrent protection, hard limits, e-stop loops live in MCU hardware. The model may err; the reflex may not — which is why embedded control engineers are scarcer, not less needed, in the «large model era».

How to learn into each layer

Layer you wantMatching roleCore topicsEntry
Brain layer (models)Embodied-AI algorithmsC6 C7 C8 + F1 F2 F4 (data)C8 · model survey
Middle layer (glue)Systems software architectureD8 D5 D6 + C-domain literacyD8
Cerebellum layer (real-time)Embedded / control electronicsD4 D7 D11 D12 + E2 E3 E8D4 · PID lab

One judgment: the brain layer's gate is papers and compute, the middle layer's gate is experience, the cerebellum layer's gate is measurement data — of the three, cerebellum people are scarcest and hardest to replace with a large model.

Keep reading

The model survey → VLA anatomy: four components → The data engine → Deployment and inference → The truth about evaluation → Six company teardowns → Kalman lab: how prediction + measurement fuse → PID lab: the cerebellum's smallest unit →