DEPLOYDatabase

Brain

Agility Motor Cortex

Agility Robotics' proprietary whole-body control foundation model for the Digit humanoid. An LSTM neural network (<1M params) trained in NVIDIA Isaac Sim for decades of simulated time. Functions like a 'motor cortex' — takes high-level commands and generates safe, stable locomotion and manipulation. Transfers zero-shot from simulation to real world.

Foundation model · Maturity: Production · Closed

Machine-readable surfaces

Architecture

LSTM neural network, <1M params, trained in NVIDIA Isaac Sim, zero-shot sim-to-real transfer

Key facts

Whole-body control foundation model (primary Agility)
Agility Robotics' Aug 28 2025 post describes a whole-body control foundation model for Digit that it says functions much like a motor cortex, taking higher-level signals to control voluntary movement and fine motor skills while keeping the robot safe and stable. The model is described as a relatively small LSTM with fewer than one million parameters, trained in NVIDIA Isaac Sim for decades of simulated time over three or four days, learned purely in simulation and transferring zero-shot to hardware ([Agility Robotics](https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model))
Prompting interface and use (primary Agility)
The same post says the motor cortex can be prompted with dense free-space position and orientation objectives for the arms and torso to walk, pick, and place heavy objects, with downstream dexterous skills and LLM coordination built on top. It says an early version was deployed at NVIDIA GTC for a grocery-shopping demo prompted by open-vocabulary object detections. Treat sim-to-real and always-on safety-layer language as company-stated ([Agility Robotics](https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model))
Whole-body control model
Agility Robotics describes Motor Cortex as a whole-body control foundation model for Digit that manages balance, movement, and limb coordination ([Training a whole-body control foundation model](https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model)).
Simulation-trained control
Agility says Motor Cortex is a lightweight recurrent model trained with reinforcement learning in NVIDIA Isaac Sim and transferred to physical robots ([Agility and AI](https://www.agilityrobotics.com/content/agility-and-ai)).
NVIDIA learning stack
Agility says its Digit learning workflow uses NVIDIA Isaac Sim and Isaac Lab to train and test robot behaviors, and that its whole-body control model is a base layer of physical intelligence for movement, balance, and arm and leg coordination ([Agility and NVIDIA](https://www.agilityrobotics.com/content/agility-robotics-expands-relationship-with-nvidia)).

Developed by (1)

Common questions

What is Agility Motor Cortex?
Agility Robotics' proprietary whole-body control foundation model for the Digit humanoid. An LSTM neural network (<1M params) trained in NVIDIA Isaac Sim for decades of simulated time. Functions like a 'motor cortex' — takes high-level commands and generates safe, stable locomotion and manipulation. Transfers zero-shot from simulation to real world.
Who developed Agility Motor Cortex?
Agility Motor Cortex is credited to Agility Robotics on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is Agility Motor Cortex open source?
No. Agility Motor Cortex is recorded as proprietary on the DEPLOY registry. Model weights and source are not publicly available.
What type of AI is Agility Motor Cortex?
Agility Motor Cortex is a foundation model, built on a LSTM neural network, <1M params, trained in NVIDIA Isaac Sim, zero-shot sim-to-real transfer architecture on the DEPLOY registry.
What is Agility Motor Cortex's maturity stage?
Agility Motor Cortex is at the production stage on the DEPLOY maturity ladder. Production stage means high-volume sustained commercial operations.
Which robots run on Agility Motor Cortex?
No robot models on the DEPLOY registry are recorded as running Agility Motor Cortex. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Sources (3)

  1. https://www.agilityrobotics.com/content/training-a-whole-body-control-foundation-model
  2. https://www.agilityrobotics.com/content/agility-and-ai
  3. https://www.agilityrobotics.com/content/agility-robotics-expands-relationship-with-nvidia
Methodology: Verified · 3 sources (no primary) · last reviewed 2026-09-29

Verification posture

Verified

Low confidence

Review state

Stable

Last reviewed 2026-09-29

Maturity + lifecycle

Maturity stage: production

Sources by quality tier

3
unclassified
Unclassified source

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for Agility Motor Cortex.

Canonical ID cb79e1e2-4559-46ee-ae66-e5b420278f19