> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cadenzalabs.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# The physical AI stack

> The layers of a modern physical-AI system, from models down to software infrastructure.

Building a robot that learns is a layered problem. A working physical-AI system is
not one model — it's a **stack**, where each layer depends on the ones beneath it.
Understanding those layers makes it clear where the hard parts are and how the
pieces fit together.

This page is a general map of that stack. The rest of the Concepts section zooms
into the layers that matter most for learning robots.

## The layers

```mermaid theme={null}
flowchart TD
    M["Models<br/>VLA · world models · perception"] --> P["Policies<br/>map observations to actions"]
    P --> C["Control & actions<br/>gaits, manipulation, sequencing"]
    C --> S["Simulation & physics<br/>the world the robot acts in"]
    S --> D["Data & training<br/>rollouts, rewards, fine-tuning"]
    D --> I["Software infrastructure<br/>orchestration, logging, deploy"]
    D -.feeds back.-> M
```

<CardGroup cols={2}>
  <Card title="Models" icon="brain">
    The learned components — vision-language-action (VLA) models, world models,
    and perception. They turn raw sensory input and intent into a plan.
  </Card>

  <Card title="Policies" icon="route">
    The decision layer: a function from observation to action. A policy is what
    you actually train and deploy.
  </Card>

  <Card title="Control & actions" icon="hand">
    The vocabulary of motion — gaits, grasps, balance, recovery — composed into
    the behaviors a policy can call on.
  </Card>

  <Card title="Simulation & physics" icon="atom">
    The environment the robot acts in. High-fidelity physics is where policies
    are trained and validated before touching hardware.
  </Card>

  <Card title="Data & training" icon="database">
    Rollouts, rewards, and the pipelines that turn experience into better
    policies through RL and fine-tuning.
  </Card>

  <Card title="Software infrastructure" icon="server">
    The connective tissue — orchestration, logging, evaluation, and deployment
    drivers that make the whole loop reproducible.
  </Card>
</CardGroup>

## Models vs the rest of the stack

It's tempting to think physical AI is "just" a model problem. It isn't. A
state-of-the-art model is useless without:

* a **simulator** good enough to train and test it safely,
* a **control layer** that turns its outputs into stable motion,
* a **data pipeline** that captures and labels what happened,
* and **infrastructure** to run, evaluate, and deploy it.

Most of the engineering effort — and most of the time lost — lives in the layers
*around* the model. That's exactly the gap Cadenza is built to close.

## How Cadenza maps onto the stack

| Layer                       | In Cadenza                                                            |
| --------------------------- | --------------------------------------------------------------------- |
| **Models**                  | Bring your own VLA / world model, or use the built-in inference stack |
| **Policies**                | Trained and fine-tuned with RL + LoRA                                 |
| **Control & actions**       | The reusable, phase-aware action library                              |
| **Simulation & physics**    | The MuJoCo-based simulator with Go1 / G1 robots                       |
| **Data & training**         | Automatic rollout logging, scoring, and fine-tuning                   |
| **Software infrastructure** | The SDK foundation + the CLI workflow, plus deploy drivers            |

<Card title="Next: Models & policies" icon="arrow-right" href="/labs/models">
  Zoom into the model layer — VLA models, world models, and what a policy is.
</Card>